Update README.md

Added the correct discord link

.dockerignore

@@ -0,0 +1,42 @@
+# Git
+.git
+.gitignore
+.github
+
+# Rust
+target/
+
+# Documentation
+*.md
+docs/
+CHANGELOG.md
+SECURITY.md
+CODE_OF_CONDUCT.md
+CONTRIBUTING.md
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+*~
+
+# OS
+.DS_Store
+Thumbs.db
+
+# Docker
+docker/
+Dockerfile*
+docker-compose*.yml
+.dockerignore
+
+# Test artifacts
+*.log
+*.tmp
+
+# Build artifacts
+*.o
+*.so
+*.dylib
+*.dll

.editorconfig

@@ -0,0 +1,24 @@
+# EditorConfig helps maintain consistent coding styles
+# https://editorconfig.org
+
+root = true
+
+[*]
+charset = utf-8
+end_of_line = lf
+insert_final_newline = true
+trim_trailing_whitespace = true
+indent_style = space
+indent_size = 4
+
+[*.rs]
+indent_size = 4
+
+[*.{yml,yaml}]
+indent_size = 2
+
+[*.md]
+trim_trailing_whitespace = false
+
+[Makefile]
+indent_style = tab

.gitattributes

@@ -0,0 +1,28 @@
+# Auto detect text files and perform LF normalization
+* text=auto
+
+# Rust source files
+*.rs text eol=lf
+*.toml text eol=lf
+
+# Documentation
+*.md text eol=lf
+*.txt text eol=lf
+
+# Scripts
+*.sh text eol=lf
+
+# Binary files
+*.mv2 binary
+*.mv2e binary
+*.pdf binary
+*.png binary
+*.jpg binary
+*.jpeg binary
+*.gif binary
+*.ico binary
+*.woff binary
+*.woff2 binary
+
+# Prevent showing Cargo.lock diffs
+Cargo.lock linguist-generated=true

.github/FUNDING.yml

@@ -0,0 +1,3 @@
+# These are supported funding model platforms
+
+github: [memvid]

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,45 @@
+---
+name: Bug Report
+about: Report a bug to help us improve Memvid
+title: '[BUG] '
+labels: 'bug'
+assignees: ''
+---
+
+## Bug Description
+A clear and concise description of the bug.
+
+## Steps to Reproduce
+1.
+2.
+3.
+
+## Expected Behavior
+What you expected to happen.
+
+## Actual Behavior
+What actually happened.
+
+## Environment
+- **OS**: [e.g., macOS 14.0, Ubuntu 22.04, Windows 11]
+- **Rust Version**: [e.g., 1.85.0]
+- **Memvid Version**: [e.g., 2.0.0]
+- **Features Enabled**: [e.g., lex, vec, clip]
+
+## Minimal Reproducible Example
+```rust
+// Code to reproduce the issue
+```
+
+## Error Output
+```
+// Paste any error messages or stack traces here
+```
+
+## Additional Context
+Any other context about the problem (screenshots, logs, etc.)
+
+## Checklist
+- [ ] I have searched existing issues for duplicates
+- [ ] I have tested with the latest version
+- [ ] I can reproduce this consistently

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1,11 @@
+blank_issues_enabled: false
+contact_links:
+  - name: Questions & Discussions
+    url: https://github.com/memvid/memvid/discussions
+    about: Ask questions and discuss ideas with the community
+  - name: Documentation
+    url: https://docs.memvid.com
+    about: Read the official documentation
+  - name: Discord Community
+    url: https://discord.gg/2mynS7fcK7
+    about: Join our Discord for real-time chat

.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,32 @@
+---
+name: Feature Request
+about: Suggest a new feature for Memvid
+title: '[FEATURE] '
+labels: 'enhancement'
+assignees: ''
+---
+
+## Feature Description
+A clear and concise description of the feature you'd like.
+
+## Problem Statement
+What problem does this feature solve? Why is it needed?
+
+## Proposed Solution
+Describe how you envision this feature working.
+
+## Example Usage
+```rust
+// How would this feature be used?
+```
+
+## Alternatives Considered
+What alternative solutions or features have you considered?
+
+## Additional Context
+Any other context, mockups, or examples.
+
+## Checklist
+- [ ] I have searched existing issues/discussions for similar requests
+- [ ] This feature aligns with Memvid's goals (portable, single-file AI memory)
+- [ ] I am willing to help implement this feature

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,42 @@
+## Description
+Brief description of the changes in this PR.
+
+## Related Issue
+Fixes #(issue number)
+
+## Type of Change
+- [ ] Bug fix (non-breaking change that fixes an issue)
+- [ ] New feature (non-breaking change that adds functionality)
+- [ ] Breaking change (fix or feature that would cause existing functionality to change)
+- [ ] Documentation update
+- [ ] Performance improvement
+- [ ] Code refactoring (no functional changes)
+
+## Changes Made
+-
+-
+-
+
+## Testing
+- [ ] I have added tests that prove my fix/feature works
+- [ ] All existing tests pass (`cargo test`)
+- [ ] I have tested on my local machine
+
+## Documentation
+- [ ] I have updated relevant documentation
+- [ ] I have added doc comments for new public APIs
+- [ ] CHANGELOG.md has been updated (if applicable)
+
+## Checklist
+- [ ] My code follows the project's style guidelines
+- [ ] I have performed a self-review of my code
+- [ ] I have commented my code where necessary
+- [ ] My changes generate no new warnings
+- [ ] I have run `cargo clippy` and addressed any issues
+- [ ] I have run `cargo fmt` to format my code
+
+## Screenshots (if applicable)
+Add screenshots to help explain your changes.
+
+## Additional Notes
+Any additional information reviewers should know.

.github/dependabot.yml.disable

@@ -0,0 +1,31 @@
+version: 2
+updates:
+  - package-ecosystem: cargo
+    directory: /
+    schedule:
+      interval: weekly
+    cooldown: # applies only to version-updates (not security-updates)
+      default-days: 7
+    ignore:
+      - dependency-name: "bincode"
+        versions: [">=3.0.0"] # unsupported since v3
+  - package-ecosystem: docker
+    directories:
+      - docker/cli
+      - docker/core
+    schedule:
+      interval: weekly
+    cooldown:
+      default-days: 7
+  - package-ecosystem: docker-compose
+    directory: docker/core
+    schedule:
+      interval: weekly
+    cooldown:
+      default-days: 7
+  - package-ecosystem: github-actions
+    directory: /
+    schedule:
+      interval: weekly
+    cooldown: # https://blog.yossarian.net/2025/11/21/We-should-all-be-using-dependency-cooldowns
+      default-days: 7

.github/workflows/ci.yml

@@ -0,0 +1,63 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+env:
+  CARGO_TERM_COLOR: always
+  RUST_BACKTRACE: 1
+
+jobs:
+  test:
+    name: Test
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-latest, macos-latest, windows-latest]
+        rust: [stable]
+
+    steps:
+      - uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8 # v6.0.1
+
+      - name: Install Rust
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          toolchain: ${{ matrix.rust }}
+
+      - name: Cache cargo
+        uses: actions/cache@9255dc7a253b0ccc959486e2bca901246202afeb # v5.0.1
+        with:
+          path: |
+            ~/.cargo/bin/
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+          key: ${{ runner.os }}-cargo-${{ hashFiles('**/Cargo.toml') }}
+
+      - name: Build
+        run: cargo build --verbose
+
+      - name: Run tests
+        run: cargo test --verbose ${{ runner.os == 'Windows' && '-- --test-threads=1' || '' }}
+
+  lint:
+    name: Lint
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8 # v6.0.1
+
+      - name: Install Rust
+        uses: dtolnay/rust-toolchain@stable
+
+      - name: Install components
+        run: rustup component add rustfmt clippy
+
+      - name: Check formatting
+        run: cargo fmt --all -- --check
+
+      - name: Clippy
+        run: cargo clippy -- -D warnings -A clippy::non_std_lazy_statics

.github/workflows/docker-release.yml

@@ -0,0 +1,50 @@
+name: Docker Release
+
+on:
+  push:
+    tags:
+      - 'v*'
+  workflow_dispatch:
+
+env:
+  REGISTRY: docker.io
+  IMAGE_NAME: memvid/cli
+
+jobs:
+  build-and-push:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8 # v6.0.1
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@c7c53464625b32c7a7e944ae62b3e17d2b600130 # v3.7.0
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@8d2750c68a42422c14e847fe6c8ac0403b4cbd6f # v3.12.0
+
+      - name: Login to Docker Hub
+        uses: docker/login-action@5e57cd118135c172c3672efd75eb46360885c0ef # v3.6.0
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
+      - name: Extract version
+        id: version
+        run: echo "VERSION=${GITHUB_REF#refs/tags/v}" >> $GITHUB_OUTPUT
+
+      - name: Build and push
+        uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83 # v6.18.0
+        with:
+          context: ./docker/cli
+          platforms: linux/amd64,linux/arm64
+          push: true
+          tags: |
+            ${{ env.IMAGE_NAME }}:latest
+            ${{ env.IMAGE_NAME }}:${{ steps.version.outputs.VERSION }}
+          cache-from: type=gha
+          cache-to: type=gha,mode=max

.gitignore

@@ -1,9 +1,49 @@
-.env
+doc.mv2
+*.mv2
+*.mv2-wal
+*.mv2-shm
+*.mv2-lock
+*.mv2-journal
+*.mv2.*
+*.mv2-wal.*
+*.mv2-shm.*
+*.mv2-lock.*
+*.mv2-journal.*
+
 .DS_Store
-.memvid/
-output/
-prompt.memvid
+*.bk
+.idea
 
-USAGE.md
 
-CLAUDE.md
+/target
+**/*.rs.bk
+# Cargo.lock - now committed for reproducible CI builds
+.idea
+
+
+entry2.json
+entry1.txt
+vector_entry.txt
+vector_embedding.txt
+
+*.pdf
+*.txt
+# Exception: SymSpell dictionary files needed for PDF text cleanup
+!data/*.txt
+*.docx
+*.doc
+*.pptx
+*.ppt
+*.xlsx
+*.xls
+*.csv
+*.json
+
+models/
+*.DS_Store
+
+issue_overview.md
+
+feature_100x_memvid.md
+
+plan_model_robust.md

CHANGELOG.md

@@ -0,0 +1,50 @@
+# Changelog
+
+All notable changes to Memvid will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+### Added
+- Initial public release of Memvid core library
+- Single-file `.mv2` format for portable AI memory
+- Full-text search with BM25 ranking (Tantivy)
+- Vector similarity search with HNSW
+- PDF, DOCX, XLSX document ingestion
+- CLIP visual embeddings for image search
+- Whisper audio transcription
+- Timeline queries for chronological browsing
+- Crash-safe WAL-based writes
+- Blake3 checksums for data integrity
+- Ed25519 signatures for authenticity
+- Optional AES-256-GCM encryption
+
+### Security
+- Embedded WAL prevents data corruption
+- Atomic commits ensure consistency
+- File locking prevents concurrent write conflicts
+
+## [2.0.0] - 2026-01-05
+
+### Added
+- Complete rewrite in Rust for performance and safety
+- New `.mv2` file format (single-file, no sidecars)
+- Append-only frame-based architecture
+- Built-in full-text and vector search
+- Cross-platform support (macOS, Linux, Windows)
+
+### Changed
+- Migrated from Python to Rust
+- New API design focused on simplicity
+- Improved memory efficiency
+
+### Removed
+- Legacy Python implementation
+- QR code video encoding (replaced with efficient binary format)
+
+---
+
+[Unreleased]: https://github.com/memvid/memvid/compare/v2.0.0...HEAD
+[2.0.0]: https://github.com/memvid/memvid/releases/tag/v2.0.0

CLAUDE.md

@@ -1,62 +1,119 @@
 # CLAUDE.md
 
-This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+This file provides guidance to Claude Code and other AI assistants working with this repository.
 
 ## Project Overview
 
-Memvid is a Python library for QR code video-based AI memory that enables:
-- Chunking and encoding text data into QR code videos
-- Fast semantic search and retrieval from QR videos
-- Conversational AI interface with context-aware memory
+Memvid is a Rust library that provides a single-file memory layer for AI agents. It packages documents, embeddings, search indices, and metadata into a portable `.mv2` file.
 
-## Key Architecture
+## Architecture
 
 ### Core Components
-- **MemvidEncoder** (memvid/encoder.py): Handles text chunking and QR video creation
-- **MemvidRetriever** (memvid/retriever.py): Fast semantic search, QR frame extraction, context assembly
-- **MemvidChat** (memvid/chat.py): Manages conversations, context retrieval, and LLM interface
-- **IndexManager** (memvid/index.py): Embedding generation, storage, and vector search
 
-### Data Flow
-1. Text chunks → Embeddings → QR codes → Video frames
-2. Query → Semantic search → Frame extraction → QR decode → Context
-3. Context + History → LLM → Response
+```
+src/
+├── lib.rs              # Public API exports
+├── memvid/             # Main Memvid struct and operations
+│   ├── mod.rs          # Memvid implementation
+│   ├── mutation.rs     # Write operations (put, commit, delete)
+│   ├── search/         # Search implementations
+│   └── ask.rs          # RAG query handling
+├── io/                 # File I/O layer
+│   ├── header.rs       # File header (4KB)
+│   ├── wal.rs          # Write-ahead log
+│   └── time_index.rs   # Chronological index
+├── lex.rs              # Full-text search (Tantivy)
+├── vec.rs              # Vector search (HNSW)
+├── clip.rs             # CLIP embeddings
+├── whisper.rs          # Audio transcription
+└── types/              # Type definitions
+```
+
+### File Format (.mv2)
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │
+├────────────────────────────┤
+│ Data Segments              │
+├────────────────────────────┤
+│ Lex Index (Tantivy)        │
+├────────────────────────────┤
+│ Vec Index (HNSW)           │
+├────────────────────────────┤
+│ Time Index                 │
+├────────────────────────────┤
+│ TOC (Footer)               │
+└────────────────────────────┘
+```
 
 ## Development Commands
 
 ```bash
-# Create and activate virtual environment
-python -m venv .memvid
-source .memvid/bin/activate  # On macOS/Linux
+# Build
+cargo build
+cargo build --release
 
-# Install dependencies
-pip install -r requirements.txt
+# Test
+cargo test
+cargo test --test lifecycle
+cargo test -- --nocapture
 
-# Run tests
-pytest tests/
+# Lint
+cargo clippy
+cargo fmt
 
-# Run specific test
-pytest tests/test_encoder.py::TestSpecificFunction
+# Run examples
+cargo run --example basic_usage
+cargo run --example pdf_ingestion
 
-# Install package in development mode
-pip install -e .
+# Benchmarks
+cargo bench
 ```
 
-## Key Dependencies
-- qrcode, Pillow: QR generation
-- opencv-python: Video processing
-- pyzbar: QR decoding
-- sentence-transformers: Semantic embeddings
-- numpy: Vector operations
-- openai: LLM integration (pluggable)
+## Key APIs
 
-## Performance Requirements
-- Retrieval (search + QR decode) must be < 2 seconds for 1M chunks
-- Use batching and parallel processing for frame extraction
-- Implement caching for hot frames and common queries
+```rust
+// Create/Open
+let mut mem = Memvid::create("file.mv2")?;
+let mut mem = Memvid::open("file.mv2")?;
 
-## Implementation Notes
-- Vector DB options: FAISS, Annoy, or Chroma for scalability
-- LLM backend should be pluggable (OpenAI, Claude, Gemini, local)
-- Thread/process pools for parallel QR decoding
-- Disk-based index for large-scale deployments
+// Write
+mem.put_bytes(content)?;
+mem.put_bytes_with_options(content, options)?;
+mem.commit()?;
+
+// Search
+mem.search(SearchRequest { query, top_k, .. })?;
+mem.timeline(TimelineQuery::default())?;
+
+// Verify
+Memvid::verify("file.mv2", deep)?;
+```
+
+## Feature Flags
+
+| Feature | Purpose |
+|---------|---------|
+| `lex` | Full-text search (default) |
+| `vec` | Vector similarity search |
+| `clip` | CLIP image embeddings |
+| `whisper` | Audio transcription |
+| `encryption` | AES-256-GCM encryption |
+
+## Code Style
+
+- Follow Rust idioms and best practices
+- Use `thiserror` for error types
+- Use `tracing` for logging
+- Add doc comments for public APIs
+- Write tests for new functionality
+
+## Important Notes
+
+- Single-file design: Never create sidecar files
+- Crash safety: All writes go through WAL
+- Append-only: Frames are immutable once committed
+- No async: Library is synchronous for simplicity

CODE_OF_CONDUCT.md

@@ -0,0 +1,45 @@
+# Code of Conduct
+
+## Our Pledge
+
+We as members, contributors, and leaders pledge to make participation in our community a harassment-free experience for everyone, regardless of age, body size, visible or invisible disability, ethnicity, sex characteristics, gender identity and expression, level of experience, education, socio-economic status, nationality, personal appearance, race, caste, color, religion, or sexual identity and orientation.
+
+We pledge to act and interact in ways that contribute to an open, welcoming, diverse, inclusive, and healthy community.
+
+## Our Standards
+
+**Examples of behavior that contributes to a positive environment:**
+
+- Using welcoming and inclusive language
+- Being respectful of differing viewpoints and experiences
+- Gracefully accepting constructive criticism
+- Focusing on what is best for the community
+- Showing empathy towards other community members
+
+**Examples of unacceptable behavior:**
+
+- The use of sexualized language or imagery, and sexual attention or advances of any kind
+- Trolling, insulting or derogatory comments, and personal or political attacks
+- Public or private harassment
+- Publishing others' private information without explicit permission
+- Other conduct which could reasonably be considered inappropriate in a professional setting
+
+## Enforcement Responsibilities
+
+Community leaders are responsible for clarifying and enforcing our standards of acceptable behavior and will take appropriate and fair corrective action in response to any behavior that they deem inappropriate, threatening, offensive, or harmful.
+
+## Scope
+
+This Code of Conduct applies within all community spaces, and also applies when an individual is officially representing the community in public spaces.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be reported to the community leaders responsible for enforcement at:
+
+**conduct@memvid.com**
+
+All complaints will be reviewed and investigated promptly and fairly.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant](https://www.contributor-covenant.org/), version 2.1.

CONTRIBUTING.md

@@ -2,125 +2,158 @@
 
 Thank you for your interest in contributing to Memvid! We welcome contributions from everyone.
 
-## 🚀 Getting Started
+## Getting Started
+
+### Prerequisites
+
+- **Rust 1.85.0+** — Install from [rustup.rs](https://rustup.rs)
+- **Git** — For version control
+
+### Setup
 
 1. **Fork the repository** on GitHub
-2. **Clone your fork** locally:
+
+2. **Clone your fork**:
    ```bash
    git clone https://github.com/YOUR_USERNAME/memvid.git
    cd memvid
    ```
-3. **Create a virtual environment**:
+
+3. **Build the project**:
    ```bash
-   python -m venv .memvid
-   source .memvid/bin/activate  # On Windows: .memvid\Scripts\activate
-   ```
-4. **Install development dependencies**:
-   ```bash
-   pip install -e ".[dev]"
-   pip install PyPDF2  # For PDF support
+   cargo build
    ```
 
-## 🔧 Development Workflow
-
-1. **Create a new branch** for your feature or fix:
+4. **Run tests**:
    ```bash
-   git checkout -b feature/your-feature-name
+   cargo test
    ```
 
-2. **Make your changes** and ensure:
-   - Code follows existing style patterns
-   - New features have tests
-   - All tests pass: `pytest tests/`
-   - Code is properly formatted
+## Development Workflow
 
-3. **Run tests**:
-   ```bash
-   pytest tests/
-   pytest --cov=memvid tests/  # With coverage
-   ```
+### Creating a Branch
 
-4. **Commit your changes**:
-   ```bash
-   git add .
-   git commit -m "Add: brief description of your changes"
-   ```
+```bash
+git checkout -b feature/your-feature-name
+# or
+git checkout -b fix/your-bug-fix
+```
 
-5. **Push to your fork**:
+### Making Changes
+
+1. Write your code following the [code style guidelines](#code-style)
+2. Add tests for new functionality
+3. Ensure all tests pass: `cargo test`
+4. Run clippy: `cargo clippy`
+5. Format code: `cargo fmt`
+
+### Committing
+
+Write clear, concise commit messages:
+
+```bash
+git commit -m "feat: add support for XYZ"
+git commit -m "fix: resolve issue with ABC"
+git commit -m "docs: update README examples"
+```
+
+### Submitting a Pull Request
+
+1. Push to your fork:
    ```bash
    git push origin feature/your-feature-name
    ```
 
-6. **Create a Pull Request** on GitHub
+2. Open a Pull Request on GitHub
 
-## 📝 Code Style
+3. Fill out the PR template completely
 
-- Follow PEP 8 guidelines
-- Use meaningful variable and function names
-- Add docstrings to all functions and classes
-- Keep functions focused and small
-- No commented-out code
+4. Wait for review and address feedback
 
-## 🧪 Testing
+## Code Style
 
-- Write tests for new features
-- Ensure all tests pass before submitting PR
-- Aim for high test coverage
-- Test edge cases
+### Rust Guidelines
 
-## 📚 Documentation
+- Follow standard Rust idioms and conventions
+- Use `rustfmt` for formatting (`cargo fmt`)
+- Use `clippy` for linting (`cargo clippy`). We maintain a **zero-warning policy**.
+- Prefer explicit types for public APIs
+- Use `thiserror` for error definitions
 
-- Update README.md if adding new features
-- Add docstrings following Google style
-- Include usage examples for new functionality
-- Update CLAUDE.md if changing core architecture
+### Linting & Safety
 
-## 🐛 Reporting Issues
+We enforce strict linting to ensure safety and portability:
 
-When reporting issues, please include:
-- Python version
+1.  **Zero Warnings**: CI will fail on any warning. Run `cargo clippy --workspace --all-targets -- -D warnings` locally.
+2.  **No Panics**: `unwrap()` and `expect()` are **denied** in library code. Use `Result` propagation (`?`) or graceful error handling. They are allowed in `tests/`.
+3.  **No Truncation**: `cast_possible_truncation` is denied. Use `try_from` when converting `u64` to `usize`/`u32`.
+4.  **Exceptions**: We allow pragmatic lints (e.g., `cast_precision_loss` for ML math) in `src/lib.rs`. Do not add global `#![allow]` without discussion.
+
+### Documentation
+
+- Add doc comments (`///`) to all public functions, structs, and modules
+- Include examples in doc comments where helpful
+- Keep comments concise and up-to-date
+
+### Testing
+
+- Write unit tests for new functionality
+- Place tests in the same file using `#[cfg(test)]` module
+- Integration tests go in the `tests/` directory
+- Aim for high coverage of edge cases
+
+## Project Structure
+
+```
+memvid/
+├── src/              # Source code
+│   ├── lib.rs        # Public API
+│   ├── memvid/       # Core implementation
+│   ├── io/           # File I/O
+│   └── types/        # Type definitions
+├── tests/            # Integration tests
+├── examples/         # Example code
+├── benchmarks/       # Benchmarks
+└── data/             # Required data files
+```
+
+## Feature Flags
+
+When adding new functionality, consider if it should be behind a feature flag:
+
+```toml
+[features]
+my_feature = ["dep:some-dependency"]
+```
+
+This keeps the default build lean and fast.
+
+## Reporting Issues
+
+When reporting bugs, please include:
+
+- Rust version (`rustc --version`)
 - Operating system
-- Complete error messages
-- Steps to reproduce
+- Memvid version
+- Minimal code to reproduce
 - Expected vs actual behavior
 
-## 💡 Feature Requests
+## Translations
 
-We love new ideas! When proposing features:
-- Explain the use case
-- Describe the expected behavior
-- Consider backward compatibility
-- Discuss implementation approach
+Interested in translating Memvid's documentation? See [Contributing Translations](docs/i18n/CONTRIBUTING_TRANSLATIONS.md) for guidelines on translating the README and other documentation.
 
-## 🤝 Code of Conduct
+## Getting Help
 
-- Be respectful and inclusive
-- Welcome newcomers and help them get started
-- Focus on constructive feedback
-- Celebrate diversity of ideas and approaches
+- Open a [Discussion](https://github.com/memvid/memvid/discussions) for questions
+- Check existing [Issues](https://github.com/memvid/memvid/issues) for similar problems
+- Email: contact@memvid.com
 
-## 📞 Getting Help
+## Recognition
 
-- Open an issue for bugs or features
-- Join discussions in existing issues
-- Tag @olow304 for urgent matters
-
-## ✅ Checklist Before Submitting PR
-
-- [ ] Tests pass locally
-- [ ] Code follows project style
-- [ ] Commit messages are clear
-- [ ] Documentation is updated
-- [ ] PR description explains changes
-- [ ] Related issue is linked (if any)
-
-## 🎉 Recognition
-
-Contributors will be:
-- Added to the project's contributor list
-- Mentioned in release notes
-- Part of the amazing Memvid community!
+Contributors are:
+- Listed in release notes
+- Part of the Memvid community
 
 ---
 
-**Thank you for making Memvid better!** 🙏
+**Thank you for making Memvid better!**

Cargo.toml

@@ -0,0 +1,164 @@
+[package]
+name = "memvid-core"
+version = "2.0.140"
+edition = "2024"
+rust-version = "1.85.0"
+license = "Apache-2.0"
+description = "Core library for Memvid v2, a crash-safe, deterministic, single-file AI memory."
+repository = "https://github.com/memvid/memvid"
+documentation = "https://docs.memvid.com"
+readme = "README.md"
+keywords = ["ai", "memory", "search", "vector", "embeddings"]
+categories = ["database", "data-structures"]
+include = ["src/**/*", "data/**/*", "Cargo.toml", "README.md"]
+
+[dependencies]
+once_cell = "1.19.0"
+serde = { version = "1.0.228", features = ["derive"] }
+bincode = { version = "2.0.1", features = ["serde"] }
+blake3 = "1.5.1"
+uuid = { version = "1.10.0", features = ["v4", "serde"] }
+log = "0.4.22"
+thiserror = "2.0.17"
+fs2 = "0.4.3"
+zstd = "0.13.1"
+lz4_flex = "0.12.0"
+tracing = "0.1.41"
+serde_json = "1.0.145"
+ed25519-dalek = { version = "2.2.0", features = ["std"] }
+base64 = "0.22.1"
+sha2 = "0.10.9"
+hex = "0.4.3"
+# extractous uses GraalVM native compilation which doesn't work on Windows ARM or WSL2 on ARM
+# Enable with --features extractous for full document extraction (PDF, DOCX, etc.)
+extractous = { version = "0.3", optional = true }
+regex = "1.11.1"
+time = { version = "0.3.36", features = ["formatting", "parsing"] }
+chrono = { version = "0.4.42", features = ["serde"] }
+interim = { version = "0.2.1", optional = true }
+lopdf = "0.39"
+# Pure Rust PDF text extraction - used as primary extractor when extractous is disabled
+pdf-extract = { version = "0.10", optional = true }
+# High-accuracy PDF text extraction with perfect word spacing (2025)
+pdf_oxide = { version = "0.3", optional = true }
+unicode-normalization = "0.1"
+unicode-segmentation = "1.11"
+zip = { version = "7.1", default-features = false, features = ["deflate"] }
+quick-xml = "0.31"
+calamine = "0.22"
+pdfium-render = { version = "0.8.28", optional = true }
+tempfile = "3.10.1"
+num_cpus = { version = "1.17", optional = true }
+crossbeam-channel = { version = "0.5.13", optional = true }
+memmap2 = "0.9"
+memchr = "2.7"
+same-file = "1.0"
+fs-err = "3.2"
+atomic-write-file = "0.3"
+dirs-next = "2.0"
+smallvec = { version = "1.13", features = ["serde", "union", "const_generics", "write"] }
+tantivy = { version = "0.25.0", optional = true, default-features = false, features = ["mmap"] }
+ort = { version = "=2.0.0-rc.10", optional = true }
+hnsw = { version = "0.11.0", optional = true, features = ["serde"] }
+jsonwebtoken = { version = "10.0.0", optional = true, features = ["rust_crypto"] }
+image = { version = "0.25", optional = true, default-features = false, features = ["jpeg", "png", "webp"] }
+ndarray = { version = "0.16", optional = true }
+rayon = { version = "1.10", optional = true }
+tokenizers = { version = "0.22", optional = true }
+symphonia = { version = "0.5.3", optional = true, default-features = false, features = ["aac", "mp3", "flac", "isomp4", "ogg", "wav", "pcm"] }
+rubato = { version = "0.15", optional = true }
+rustfft = { version = "6.2", optional = true }
+
+# Encryption capsules (.mv2e) - feature-gated
+argon2 = { version = "0.5", optional = true }
+aes-gcm = { version = "0.10", optional = true }
+rand = { version = "0.8", optional = true, features = ["serde1"] }
+rand_pcg = { version = "0.3", optional = true, features = ["serde1"] }
+zeroize = { version = "1.7", optional = true }
+
+# Candle ML framework for Whisper transcription
+candle-core = { version = "0.9", optional = true }
+candle-nn = { version = "0.9", optional = true }
+candle-transformers = { version = "0.9", optional = true }
+hf-hub = { version = "0.4", optional = true, features = ["tokio"] }
+byteorder = { version = "1.5", optional = true }
+
+# SymSpell for PDF text cleanup - fixes broken word spacing from PDF extraction
+symspell = { version = "0.4", optional = true }
+# SIMD acceleration for vector distance calculations
+wide = { version = "1.1", optional = true }
+space = { version = "0.17", optional = true }
+
+# HTTP client for API-based embedding providers (OpenAI, etc.)
+reqwest = { version = "0.12", optional = true, default-features = false, features = ["blocking", "json", "rustls-tls"] }
+
+# Platform-specific: libc for stderr suppression on macOS
+[target.'cfg(target_os = "macos")'.dependencies]
+libc = "0.2"
+
+[features]
+default = ["lex", "pdf_extract", "simd"]
+# pdf_oxide disabled - cff-parser panics on ligature fonts (uniFB01/uniFB02)
+# symspell_cleanup - enables robust PDF text repair (requires `make download-models` for dictionaries)
+lex = ["dep:tantivy"]
+extractous = ["dep:extractous"]
+# Pure Rust PDF extraction - faster than lopdf for text extraction, cross-platform
+pdf_extract = ["dep:pdf-extract"]
+# High-accuracy PDF extraction with perfect word spacing (recommended for 2025+)
+pdf_oxide = ["dep:pdf_oxide"]
+vec = ["dep:ort", "dep:hnsw", "dep:ndarray", "dep:tokenizers", "dep:space", "dep:rand", "dep:rand_pcg"]
+clip = ["vec", "dep:image", "dep:ndarray", "dep:rayon", "dep:tokenizers"]
+mmap = []
+pdfium = ["dep:pdfium-render"]
+temporal_track = []
+temporal_enrich = ["dep:interim"]
+parallel_segments = ["dep:num_cpus", "dep:crossbeam-channel"]
+# Logic-Mesh: entity-relationship graph with NER extraction using DistilBERT-NER ONNX
+logic_mesh = ["dep:ort", "dep:ndarray", "dep:tokenizers"]
+# Whisper: audio transcription with Candle inference
+whisper = ["dep:symphonia", "dep:rubato", "dep:tokenizers", "dep:candle-core", "dep:candle-nn", "dep:candle-transformers", "dep:hf-hub", "dep:byteorder"]
+# GPU acceleration for Whisper (optional)
+metal = ["candle-core/metal", "candle-nn/metal", "candle-transformers/metal"]
+cuda = ["candle-core/cuda", "candle-nn/cuda", "candle-transformers/cuda"]
+accelerate = ["candle-core/accelerate", "candle-nn/accelerate", "candle-transformers/accelerate"]
+# Time-travel replay for agent sessions
+replay = []
+# Password-based encryption capsules (.mv2e)
+encryption = ["dep:argon2", "dep:aes-gcm", "dep:rand", "dep:zeroize"]
+# SymSpell-based PDF text cleanup - fixes broken word spacing
+symspell_cleanup = ["dep:symspell"]
+# API-based embedding providers (OpenAI, Anthropic, etc.) - requires network
+api_embed = ["dep:reqwest"]
+# SIMD acceleration for vector distance calculations
+simd = ["dep:wide"]
+hnsw_bench = ["dep:hnsw", "dep:rand", "dep:space", "dep:rand_pcg"]
+
+[dev-dependencies]
+fastrand = "2.0"
+tempfile = "3.10.1"
+criterion = { version = "0.8.1", features = ["html_reports"] }
+
+[[example]]
+name = "text_embedding"
+required-features = ["vec"]
+
+[[example]]
+name = "text_embed_cache_bench"
+required-features = ["vec"]
+
+[[example]]
+name = "openai_embedding"
+required-features = ["api_embed"]
+
+[[example]]
+name = "simd_benchmark"
+required-features = ["simd"]
+
+[[bench]]
+name = "search_precision_benchmark"
+harness = false
+
+[[bench]]
+name = "vec_search_benchmark"
+harness = false
+required-features = ["hnsw_bench"]

LICENSE

@@ -1,21 +1,190 @@
-MIT License
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
 
-Copyright (c) 2025 Olow304
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
 
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
+   1. Definitions.
 
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
 
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to the Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   Copyright 2026 Memvid
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

MANIFEST.in

@@ -1,9 +0,0 @@
-include README.md
-include LICENSE
-include requirements.txt
-recursive-include memvid *.py
-recursive-include examples *.py
-recursive-include tests *.py
-global-exclude __pycache__ *.pyc *.pyo *.swp .DS_Store
-exclude data/*
-exclude output/*

MV2_SPEC.md

@@ -0,0 +1,226 @@
+# MV2 File Format Specification
+
+Version 2.1
+
+## Overview
+
+MV2 is a single-file format for AI memory storage. Everything lives in one file: header, write-ahead log, data segments, search indices, and metadata. No sidecar files.
+
+```
+┌─────────────────────────────────────────────────────────────┐
+│                        .mv2 FILE                            │
+├─────────────────────────────────────────────────────────────┤
+│ Header                 │ 4 KB                               │
+├─────────────────────────────────────────────────────────────┤
+│ Embedded WAL           │ 1-64 MB (capacity-dependent)       │
+├─────────────────────────────────────────────────────────────┤
+│ Data Segments          │ Variable                           │
+│   - Frame payloads                                          │
+│   - Compressed content                                      │
+├─────────────────────────────────────────────────────────────┤
+│ Lex Index Segment      │ Tantivy index (optional)           │
+├─────────────────────────────────────────────────────────────┤
+│ Vec Index Segment      │ HNSW vectors (optional)            │
+├─────────────────────────────────────────────────────────────┤
+│ Time Index Segment     │ Chronological ordering             │
+├─────────────────────────────────────────────────────────────┤
+│ TOC (Footer)           │ Segment catalog + checksums        │
+└─────────────────────────────────────────────────────────────┘
+```
+
+## Header (4096 bytes)
+
+The header occupies the first 4 KB of the file.
+
+| Offset | Size | Field | Description |
+|--------|------|-------|-------------|
+| 0 | 4 | `magic` | `MV2\0` (0x4D 0x56 0x32 0x00) |
+| 4 | 2 | `version` | Format version (little-endian) |
+| 6 | 1 | `spec_major` | Spec major version (2) |
+| 7 | 1 | `spec_minor` | Spec minor version (1) |
+| 8 | 8 | `footer_offset` | Byte offset to TOC |
+| 16 | 8 | `wal_offset` | Byte offset to WAL (always 4096) |
+| 24 | 8 | `wal_size` | WAL region size in bytes |
+| 32 | 8 | `wal_checkpoint_pos` | Last checkpointed sequence |
+| 40 | 8 | `wal_sequence` | Current WAL sequence number |
+| 48 | 32 | `toc_checksum` | SHA-256 of TOC segment |
+| 80 | 4016 | reserved | Zero-filled, reserved for future use |
+
+All multi-byte integers are little-endian.
+
+## Write-Ahead Log (WAL)
+
+The embedded WAL provides crash recovery. It starts at byte 4096 and has a capacity determined by the file's target size:
+
+| File Capacity | WAL Size |
+|---------------|----------|
+| < 100 MB | 1 MB |
+| < 1 GB | 4 MB |
+| < 10 GB | 16 MB |
+| >= 10 GB | 64 MB |
+
+### WAL Entry Format
+
+```
+┌──────────────────────────────────────┐
+│ sequence    │ 8 bytes (u64 LE)       │
+│ entry_type  │ 1 byte                 │
+│ payload_len │ 4 bytes (u32 LE)       │
+│ payload     │ variable               │
+│ checksum    │ 4 bytes (CRC32)        │
+└──────────────────────────────────────┘
+```
+
+Entry types:
+- `0x01` - Frame append
+- `0x02` - Frame update
+- `0x03` - Frame delete (tombstone)
+- `0x04` - Index update
+
+### Checkpoint Behavior
+
+- Checkpoint triggers at 75% WAL occupancy or every 1,000 transactions
+- Checkpoint flushes WAL entries to data segments
+- `seal()` forces immediate checkpoint
+- Recovery replays entries with `sequence > wal_checkpoint_pos`
+
+## Frame Structure
+
+Each frame represents a single piece of content.
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `frame_id` | u64 | Unique identifier (monotonic) |
+| `uri` | String | Hierarchical path (`mv2://path/to/doc`) |
+| `title` | String? | Optional display title |
+| `created_at` | u64 | Unix timestamp (seconds) |
+| `encoding` | u8 | Content encoding (see below) |
+| `payload` | bytes | Compressed content |
+| `payload_checksum` | [u8; 32] | SHA-256 of uncompressed payload |
+| `tags` | Map<String, String> | User-defined key-value pairs |
+| `status` | u8 | 0=active, 1=tombstoned |
+
+### Encoding Types
+
+| Value | Name | Description |
+|-------|------|-------------|
+| 0 | Raw | Uncompressed bytes |
+| 1 | Zstd | Zstandard compression |
+| 2 | Lz4 | LZ4 compression |
+
+## Data Segments
+
+Frames are grouped into segments for efficient storage and retrieval.
+
+### Segment Header
+
+```
+┌──────────────────────────────────────┐
+│ magic         │ 4 bytes              │
+│ version       │ 2 bytes              │
+│ segment_type  │ 1 byte               │
+│ frame_count   │ 4 bytes              │
+│ compressed    │ 1 byte (bool)        │
+│ checksum      │ 32 bytes             │
+└──────────────────────────────────────┘
+```
+
+Segment types:
+- `0x01` - Data segment (frames)
+- `0x02` - Lex index segment
+- `0x03` - Vec index segment
+- `0x04` - Time index segment
+
+## Time Index
+
+The time index enables chronological queries and time-travel.
+
+### Time Index Entry
+
+| Field | Size | Description |
+|-------|------|-------------|
+| `frame_id` | 8 | Frame identifier |
+| `timestamp` | 8 | Unix timestamp |
+| `offset` | 8 | Byte offset in data segment |
+
+Magic: `MVTI` (0x4D 0x56 0x54 0x49)
+
+## Lex Index (Full-Text Search)
+
+When the `lex` feature is enabled, the file contains a Tantivy index segment.
+
+Indexed fields:
+- `body` - Full text content
+- `title` - Document title
+- `uri` - Document URI
+- `tags` - Flattened tag values
+
+Supports:
+- BM25 ranking
+- Phrase queries
+- Boolean operators
+- Date range filters
+
+## Vec Index (Vector Search)
+
+When the `vec` feature is enabled, the file contains an HNSW index segment.
+
+| Parameter | Value |
+|-----------|-------|
+| Dimensions | 384 (BGE-small) |
+| Distance | Cosine similarity |
+| M | 16 |
+| ef_construction | 200 |
+
+## Table of Contents (TOC)
+
+The TOC is the final segment, pointed to by `footer_offset` in the header.
+
+```
+┌──────────────────────────────────────┐
+│ magic         │ "MVTC"               │
+│ version       │ 2 bytes              │
+│ segment_count │ 4 bytes              │
+│ segments[]    │ SegmentDescriptor[]  │
+│ manifests     │ IndexManifests       │
+│ checksum      │ 32 bytes             │
+└──────────────────────────────────────┘
+```
+
+### Segment Descriptor
+
+| Field | Size | Description |
+|-------|------|-------------|
+| `segment_type` | 1 | Type identifier |
+| `offset` | 8 | Byte offset in file |
+| `length` | 8 | Segment size in bytes |
+| `checksum` | 32 | SHA-256 of segment |
+
+## URI Scheme
+
+All content is addressable via `mv2://` URIs:
+
+```
+mv2://[track/][path/]name
+```
+
+Examples:
+- `mv2://meetings/2024-01-15`
+- `mv2://docs/api/reference.md`
+- `mv2://media/photo.png`
+
+## Invariants
+
+1. **Single-file guarantee**: No `.wal`, `.shm`, `.lock`, or other sidecar files
+2. **Append-only frames**: Existing frames are never modified in place
+3. **Determinism**: Same API calls produce identical bytes
+4. **Crash safety**: WAL ensures durability across unexpected termination
+5. **Self-describing**: TOC contains all metadata needed to parse the file
+
+## Version History
+
+| Version | Changes |
+|---------|---------|
+| 2.1 | Current version. Embedded WAL, temporal track support |
+| 2.0 | Single-file format, removed external indices |
+| 1.x | Legacy format (deprecated) |

Makefile

@@ -0,0 +1,236 @@
+.PHONY: help build build-release test test-verbose clean fmt fmt-check clippy clippy-fix doc doc-open check install run-example docker-build docker-test docker-push pre-commit tree outdated bloat test-doc test-release package publish-dry-run coverage watch
+
+# Default target
+.DEFAULT_GOAL := help
+
+# Variables
+CARGO := cargo
+RUST_VERSION := 1.85.0
+FEATURES := lex,pdf_extract
+
+# Colors for output
+CYAN := \033[0;36m
+GREEN := \033[0;32m
+YELLOW := \033[0;33m
+NC := \033[0m # No Color
+
+help: ## Show this help message
+	@echo "$(CYAN)Memvid Makefile Commands:$(NC)"
+	@echo ""
+	@grep -E '^[a-zA-Z_-]+:.*?## .*$$' $(MAKEFILE_LIST) | sort | awk 'BEGIN {FS = ":.*?## "}; {printf "  $(GREEN)%-20s$(NC) %s\n", $$1, $$2}'
+	@echo ""
+
+install: ## Install Rust toolchain and dependencies
+	@echo "$(CYAN)Installing Rust toolchain...$(NC)"
+	@rustup toolchain install $(RUST_VERSION) || true
+	@rustup default $(RUST_VERSION) || true
+	@echo "$(CYAN)Installing cargo dependencies...$(NC)"
+	@$(CARGO) fetch
+
+download-models: ## Download required models and dictionaries
+	@echo "$(CYAN)Downloading SymSpell dictionaries...$(NC)"
+	@mkdir -p data
+	@curl -L https://raw.githubusercontent.com/wolfgarbe/SymSpell/master/SymSpell/frequency_dictionary_en_82_765.txt -o data/frequency_dictionary_en_82_765.txt
+	@curl -L https://raw.githubusercontent.com/wolfgarbe/SymSpell/master/SymSpell/frequency_bigramdictionary_en_243_342.txt -o data/frequency_bigramdictionary_en_243_342.txt
+
+
+check: ## Check code without building
+	@echo "$(CYAN)Checking code...$(NC)"
+	@$(CARGO) check --features $(FEATURES)
+
+build: ## Build in debug mode
+	@echo "$(CYAN)Building in debug mode...$(NC)"
+	@$(CARGO) build --features $(FEATURES)
+
+build-release: ## Build in release mode (optimized)
+	@echo "$(CYAN)Building in release mode...$(NC)"
+	@$(CARGO) build --release --features $(FEATURES)
+
+build-verbose: ## Build with verbose output
+	@echo "$(CYAN)Building with verbose output...$(NC)"
+	@$(CARGO) build --verbose --features $(FEATURES)
+
+build-release-verbose: ## Build release with verbose output
+	@echo "$(CYAN)Building release with verbose output...$(NC)"
+	@$(CARGO) build --release --verbose --features $(FEATURES)
+
+build-all-features: ## Build with all features enabled
+	@echo "$(CYAN)Building with all features...$(NC)"
+	@$(CARGO) build --release --all-features
+
+test: ## Run tests
+	@echo "$(CYAN)Running tests...$(NC)"
+	@$(CARGO) test --features $(FEATURES)
+
+test-verbose: ## Run tests with output
+	@echo "$(CYAN)Running tests with output...$(NC)"
+	@$(CARGO) test --features $(FEATURES) -- --nocapture
+
+test-release: ## Run tests in release mode
+	@echo "$(CYAN)Running tests in release mode...$(NC)"
+	@$(CARGO) test --release --features $(FEATURES)
+
+test-doc: ## Run documentation tests only
+	@echo "$(CYAN)Running documentation tests...$(NC)"
+	@$(CARGO) test --doc --features $(FEATURES)
+
+test-all-targets: ## Run all test targets (lib, bins, tests, examples)
+	@echo "$(CYAN)Running all test targets...$(NC)"
+	@$(CARGO) test --all-targets --features $(FEATURES)
+
+test-no-fail-fast: ## Run all tests even if one fails
+	@echo "$(CYAN)Running all tests (no fail fast)...$(NC)"
+	@$(CARGO) test --features $(FEATURES) -- --no-fail-fast
+
+test-integration: ## Run integration tests only
+	@echo "$(CYAN)Running integration tests...$(NC)"
+	@$(CARGO) test --test lifecycle --test search --test mutation --test crash_recovery --test doctor_recovery --test encryption_capsule --test replay_integrity --test single_file --features $(FEATURES)
+
+test-unit: ## Run unit tests only
+	@echo "$(CYAN)Running unit tests...$(NC)"
+	@$(CARGO) test --lib --features $(FEATURES)
+
+fmt: ## Format code
+	@echo "$(CYAN)Formatting code...$(NC)"
+	@$(CARGO) fmt --all
+
+fmt-check: ## Check code formatting
+	@echo "$(CYAN)Checking code formatting...$(NC)"
+	@$(CARGO) fmt --all -- --check
+
+clippy: ## Run clippy linter
+	@echo "$(CYAN)Running clippy...$(NC)"
+	@$(CARGO) clippy --all-targets --features $(FEATURES) -- -D warnings
+
+clippy-fix: ## Run clippy and auto-fix issues
+	@echo "$(CYAN)Running clippy with auto-fix...$(NC)"
+	@$(CARGO) clippy --fix --all-targets --features $(FEATURES) -- -D warnings
+
+doc: ## Generate documentation
+	@echo "$(CYAN)Generating documentation...$(NC)"
+	@$(CARGO) doc --features $(FEATURES) --no-deps
+
+doc-open: ## Generate and open documentation
+	@echo "$(CYAN)Generating and opening documentation...$(NC)"
+	@$(CARGO) doc --features $(FEATURES) --no-deps --open
+
+clean: ## Clean build artifacts
+	@echo "$(CYAN)Cleaning build artifacts...$(NC)"
+	@$(CARGO) clean
+
+clean-all: clean ## Clean everything including target directory
+	@echo "$(CYAN)Cleaning all artifacts...$(NC)"
+	@rm -rf target/
+
+run-example-basic: ## Run basic_usage example
+	@echo "$(CYAN)Running basic_usage example...$(NC)"
+	@$(CARGO) run --example basic_usage --features $(FEATURES)
+
+run-example-pdf: ## Run pdf_ingestion example (usage: make run-example-pdf PDF_PATH=/path/to/pdf)
+	@if [ -z "$(PDF_PATH)" ]; then \
+		echo "$(YELLOW)Usage: make run-example-pdf PDF_PATH=/path/to/pdf$(NC)"; \
+		echo "$(YELLOW)Example: make run-example-pdf PDF_PATH=examples/1706.03762v7.pdf$(NC)"; \
+		exit 1; \
+	fi
+	@echo "$(CYAN)Running pdf_ingestion example with $(PDF_PATH)...$(NC)"
+	@$(CARGO) run --example pdf_ingestion --features $(FEATURES) -- $(PDF_PATH)
+
+run-example-clip: ## Run clip_visual_search example (usage: make run-example-clip PDF_PATH=/path/to/pdf)
+	@if [ -z "$(PDF_PATH)" ]; then \
+		echo "$(YELLOW)Usage: make run-example-clip PDF_PATH=/path/to/pdf$(NC)"; \
+		echo "$(YELLOW)Example: make run-example-clip PDF_PATH=examples/document.pdf$(NC)"; \
+		exit 1; \
+	fi
+	@echo "$(CYAN)Running clip_visual_search example with $(PDF_PATH)...$(NC)"
+	@$(CARGO) run --example clip_visual_search --features $(FEATURES),clip -- $(PDF_PATH)
+
+run-example-whisper: ## Run test_whisper example (usage: make run-example-whisper AUDIO_PATH=/path/to/audio)
+	@if [ -z "$(AUDIO_PATH)" ]; then \
+		echo "$(YELLOW)Usage: make run-example-whisper AUDIO_PATH=/path/to/audio$(NC)"; \
+		echo "$(YELLOW)Example: make run-example-whisper AUDIO_PATH=examples/call_sale.mp3$(NC)"; \
+		exit 1; \
+	fi
+	@echo "$(CYAN)Running test_whisper example with $(AUDIO_PATH)...$(NC)"
+	@$(CARGO) run --example test_whisper --features $(FEATURES),whisper -- $(AUDIO_PATH)
+
+lint: fmt-check clippy ## Run all linting checks
+
+verify: check lint test ## Run all verification checks (check, lint, test)
+
+ci: verify build-release ## Run CI pipeline (verify + release build)
+
+docker-build: ## Build Docker image
+	@echo "$(CYAN)Building Docker image...$(NC)"
+	@cd docker/cli && docker build -t memvid/cli:latest .
+
+docker-test: ## Test Docker image
+	@echo "$(CYAN)Testing Docker image...$(NC)"
+	@cd docker/cli && ./test.sh
+
+docker-push: ## Push Docker image to registry (requires login)
+	@echo "$(CYAN)Pushing Docker image...$(NC)"
+	@docker push memvid/cli:latest
+
+docker-tag: ## Tag Docker image with version
+	@echo "$(CYAN)Tagging Docker image...$(NC)"
+	@VERSION=$$(grep "^version" Cargo.toml | cut -d'"' -f2); \
+	docker tag memvid/cli:latest memvid/cli:$$VERSION; \
+	echo "Tagged as memvid/cli:$$VERSION"
+
+bench: ## Run benchmarks (if available)
+	@echo "$(CYAN)Running benchmarks...$(NC)"
+	@$(CARGO) bench --features $(FEATURES) || echo "$(YELLOW)No benchmarks found$(NC)"
+
+update: ## Update dependencies
+	@echo "$(CYAN)Updating dependencies...$(NC)"
+	@$(CARGO) update
+
+audit: ## Audit dependencies for security vulnerabilities
+	@echo "$(CYAN)Auditing dependencies...$(NC)"
+	@$(CARGO) audit || echo "$(YELLOW)cargo-audit not installed. Install with: cargo install cargo-audit$(NC)"
+
+version: ## Show version information
+	@echo "$(CYAN)Version Information:$(NC)"
+	@$(CARGO) --version
+	@rustc --version
+	@echo ""
+	@echo "$(CYAN)Project version:$(NC)"
+	@grep "^version" Cargo.toml
+
+tree: ## Show dependency tree
+	@echo "$(CYAN)Dependency tree:$(NC)"
+	@$(CARGO) tree --features $(FEATURES)
+
+tree-duplicates: ## Show duplicate dependencies
+	@echo "$(CYAN)Duplicate dependencies:$(NC)"
+	@$(CARGO) tree --duplicates --features $(FEATURES)
+
+outdated: ## Check for outdated dependencies
+	@echo "$(CYAN)Checking for outdated dependencies...$(NC)"
+	@$(CARGO) outdated || echo "$(YELLOW)cargo-outdated not installed. Install with: cargo install cargo-outdated$(NC)"
+
+bloat: ## Analyze binary size bloat
+	@echo "$(CYAN)Analyzing binary size...$(NC)"
+	@$(CARGO) bloat --release --features $(FEATURES) || echo "$(YELLOW)cargo-bloat not installed. Install with: cargo install cargo-bloat$(NC)"
+
+package: ## Create a package for publishing
+	@echo "$(CYAN)Creating package...$(NC)"
+	@$(CARGO) package
+
+publish-dry-run: ## Dry run of publish (check if ready)
+	@echo "$(CYAN)Running publish dry-run...$(NC)"
+	@$(CARGO) publish --dry-run
+
+coverage: ## Generate test coverage report
+	@echo "$(CYAN)Generating test coverage...$(NC)"
+	@$(CARGO) tarpaulin --features $(FEATURES) --out Html || echo "$(YELLOW)cargo-tarpaulin not installed. Install with: cargo install cargo-tarpaulin$(NC)"
+
+watch: ## Watch for changes and run tests
+	@echo "$(CYAN)Watching for changes...$(NC)"
+	@$(CARGO) watch -x "test --features $(FEATURES)" || echo "$(YELLOW)cargo-watch not installed. Install with: cargo install cargo-watch$(NC)"
+
+watch-build: ## Watch for changes and rebuild
+	@echo "$(CYAN)Watching for changes and rebuilding...$(NC)"
+	@$(CARGO) watch -x "build --features $(FEATURES)" || echo "$(YELLOW)cargo-watch not installed. Install with: cargo install cargo-watch$(NC)"
+
+pre-commit: fmt-check clippy test ## Run pre-commit checks (fmt, clippy, test)

README.md

@@ -1,371 +1,514 @@
-# Memvid - Video-Based AI Memory 🧠📹
-
-**The lightweight, game-changing solution for AI memory at scale**
-
-[![PyPI version](https://badge.fury.io/py/memvid.svg)](https://pypi.org/project/memvid/)
-[![Downloads](https://pepy.tech/badge/memvid)](https://pepy.tech/project/memvid)
-[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
-[![Python 3.8+](https://img.shields.io/badge/python-3.8+-blue.svg)](https://www.python.org/downloads/)
-[![Code style: black](https://img.shields.io/badge/code%20style-black-000000.svg)](https://github.com/psf/black)
-
-Memvid revolutionizes AI memory management by encoding text data into videos, enabling **lightning-fast semantic search** across millions of text chunks with **sub-second retrieval times**. Unlike traditional vector databases that consume massive amounts of RAM and storage, Memvid compresses your knowledge base into compact video files while maintaining instant access to any piece of information.
-
-## ✨ Key Features
-
-- 🎥 **Video-as-Database**: Store millions of text chunks in a single MP4 file
-- 🔍 **Semantic Search**: Find relevant content using natural language queries
-- 💬 **Built-in Chat**: Conversational interface with context-aware responses
-- 📚 **PDF Support**: Direct import and indexing of PDF documents
-- 🚀 **Fast Retrieval**: Sub-second search across massive datasets
-- 💾 **Efficient Storage**: 10x compression compared to traditional databases
-- 🔌 **Pluggable LLMs**: Works with OpenAI, Anthropic, or local models
-- 🌐 **Offline-First**: No internet required after video generation
-- 🔧 **Simple API**: Get started with just 3 lines of code
-
-## 🎯 Use Cases
-
-- **📖 Digital Libraries**: Index thousands of books in a single video file
-- **🎓 Educational Content**: Create searchable video memories of course materials
-- **📰 News Archives**: Compress years of articles into manageable video databases
-- **💼 Corporate Knowledge**: Build company-wide searchable knowledge bases
-- **🔬 Research Papers**: Quick semantic search across scientific literature
-- **📝 Personal Notes**: Transform your notes into a searchable AI assistant
-
-## 🚀 Why Memvid?
-
-### Game-Changing Innovation
-- **Video as Database**: Store millions of text chunks in a single MP4 file
-- **Instant Retrieval**: Sub-second semantic search across massive datasets
-- **10x Storage Efficiency**: Video compression reduces memory footprint dramatically
-- **Zero Infrastructure**: No database servers, just files you can copy anywhere
-- **Offline-First**: Works completely offline once videos are generated
-
-### Lightweight Architecture
-- **Minimal Dependencies**: Core functionality in ~1000 lines of Python
-- **CPU-Friendly**: Runs efficiently without GPU requirements
-- **Portable**: Single video file contains your entire knowledge base
-- **Streamable**: Videos can be streamed from cloud storage
-
-## 📦 Installation
-
-### Quick Install
-```bash
-pip install memvid
-```
-
-### For PDF Support
-```bash
-pip install memvid PyPDF2
-```
-
-### Recommended Setup (Virtual Environment)
-```bash
-# Create a new project directory
-mkdir my-memvid-project
-cd my-memvid-project
-
-# Create virtual environment
-python -m venv venv
-
-# Activate it
-# On macOS/Linux:
-source venv/bin/activate
-# On Windows:
-venv\Scripts\activate
-
-# Install memvid
-pip install memvid
-
-# For PDF support:
-pip install PyPDF2
-```
-
-## 🎯 Quick Start
-
-### Basic Usage
-```python
-from memvid import MemvidEncoder, MemvidChat
-
-# Create video memory from text chunks
-chunks = ["Important fact 1", "Important fact 2", "Historical event details", ...]
-encoder = MemvidEncoder()
-encoder.add_chunks(chunks)
-encoder.build_video("memory.mp4", "memory_index.json")
-
-# Chat with your memory
-chat = MemvidChat("memory.mp4", "memory_index.json")
-chat.start_session()
-response = chat.chat("What do you know about historical events?")
-print(response)
-```
-
-### Building Memory from Documents
-```python
-from memvid import MemvidEncoder
-import os
-
-# Load documents
-encoder = MemvidEncoder(chunk_size=512, overlap=50)
-
-# Add text files
-for file in os.listdir("documents"):
-    with open(f"documents/{file}", "r") as f:
-        encoder.add_text(f.read(), metadata={"source": file})
-
-# Build optimized video
-encoder.build_video(
-    "knowledge_base.mp4",
-    "knowledge_index.json",
-    fps=30,  # Higher FPS = more chunks per second
-    frame_size=512  # Larger frames = more data per frame
-)
-```
-
-### Advanced Search & Retrieval
-```python
-from memvid import MemvidRetriever
-
-# Initialize retriever
-retriever = MemvidRetriever("knowledge_base.mp4", "knowledge_index.json")
-
-# Semantic search
-results = retriever.search("machine learning algorithms", top_k=5)
-for chunk, score in results:
-    print(f"Score: {score:.3f} | {chunk[:100]}...")
-
-# Get context window
-context = retriever.get_context("explain neural networks", max_tokens=2000)
-print(context)
-```
-
-### Interactive Chat Interface
-```python
-from memvid import MemvidInteractive
-
-# Launch interactive chat UI
-interactive = MemvidInteractive("knowledge_base.mp4", "knowledge_index.json")
-interactive.run()  # Opens web interface at http://localhost:7860
-```
-
-### Complete Example: Chat with a PDF Book
-```bash
-# 1. Create a new directory and set up environment
-mkdir book-chat-demo
-cd book-chat-demo
-python -m venv venv
-source venv/bin/activate  # On Windows: venv\Scripts\activate
-
-# 2. Install dependencies
-pip install memvid PyPDF2
-
-# 3. Create book_chat.py
-cat > book_chat.py << 'EOF'
-from memvid import MemvidEncoder, chat_with_memory
-import os
-
-# Your PDF file
-book_pdf = "book.pdf"  # Replace with your PDF path
-
-# Build video memory
-encoder = MemvidEncoder()
-encoder.add_pdf(book_pdf)
-encoder.build_video("book_memory.mp4", "book_index.json")
-
-# Chat with the book
-api_key = os.getenv("OPENAI_API_KEY")  # Optional: for AI responses
-chat_with_memory("book_memory.mp4", "book_index.json", api_key=api_key)
-EOF
-
-# 4. Run it
-export OPENAI_API_KEY="your-api-key"  # Optional
-python book_chat.py
-```
-
-
-## 🔧 API Reference
-
-### MemvidEncoder
-```python
-encoder = MemvidEncoder(
-    chunk_size=512,      # Characters per chunk
-    overlap=50,          # Character overlap between chunks
-    model_name='all-MiniLM-L6-v2'  # Sentence transformer model
-)
-
-# Methods
-encoder.add_chunks(chunks: List[str], metadata: List[dict] = None)
-encoder.add_text(text: str, metadata: dict = None)
-encoder.build_video(video_path: str, index_path: str, fps: int = 30, qr_size: int = 512)
-```
-
-### MemvidRetriever
-```python
-retriever = MemvidRetriever(
-    video_path: str,
-    index_path: str,
-    cache_size: int = 100  # Number of frames to cache
-)
-
-# Methods
-results = retriever.search(query: str, top_k: int = 5)
-context = retriever.get_context(query: str, max_tokens: int = 2000)
-chunks = retriever.get_chunks_by_ids(chunk_ids: List[int])
-```
-
-### MemvidChat
-```python
-chat = MemvidChat(
-    video_path: str,
-    index_path: str,
-    llm_backend: str = 'openai',  # 'openai', 'anthropic', 'local'
-    model: str = 'gpt-4'
-)
-
-# Methods
-chat.start_session(system_prompt: str = None)
-response = chat.chat(message: str, stream: bool = False)
-chat.clear_history()
-chat.export_conversation(path: str)
-```
-
-## 🛠️ Advanced Configuration
-
-### Custom Embeddings
-```python
-from sentence_transformers import SentenceTransformer
-
-# Use custom embedding model
-custom_model = SentenceTransformer('sentence-transformers/all-mpnet-base-v2')
-encoder = MemvidEncoder(embedding_model=custom_model)
-```
-
-### Video Optimization
-```python
-# For maximum compression
-encoder.build_video(
-    "compressed.mp4",
-    "index.json",
-    fps=60,  # More frames per second
-    frame_size=256,  # Smaller frames
-    video_codec='h265',  # Better compression
-    crf=28  # Compression quality (lower = better quality)
-)
-```
-
-### Distributed Processing
-```python
-# Process large datasets in parallel
-encoder = MemvidEncoder(n_workers=8)
-encoder.add_chunks_parallel(massive_chunk_list)
-```
-
-## 🐛 Troubleshooting
-
-### Common Issues
-
-**ModuleNotFoundError: No module named 'memvid'**
-```bash
-# Make sure you're using the right Python
-which python  # Should show your virtual environment path
-# If not, activate your virtual environment:
-source venv/bin/activate  # On Windows: venv\Scripts\activate
-```
-
-**ImportError: PyPDF2 is required for PDF support**
-```bash
-pip install PyPDF2
-```
-
-**OpenAI API Key Issues**
-```bash
-# Set your API key (get one at https://platform.openai.com)
-export OPENAI_API_KEY="sk-..."  # macOS/Linux
-# Or on Windows:
-set OPENAI_API_KEY=sk-...
-```
-
-**Large PDF Processing**
-```python
-# For very large PDFs, use smaller chunk sizes
-encoder = MemvidEncoder()
-encoder.add_pdf("large_book.pdf", chunk_size=400, overlap=50)
-```
-
-## 🤝 Contributing
-
-We welcome contributions! Please see our [Contributing Guide](CONTRIBUTING.md) for details.
-
-```bash
-# Run tests
-pytest tests/
-
-# Run with coverage
-pytest --cov=memvid tests/
-
-# Format code
-black memvid/
-```
-
-## 🆚 Comparison with Traditional Solutions
-
-| Feature | Memvid | Vector DBs | Traditional DBs |
-|---------|--------|------------|-----------------|
-| Storage Efficiency | ⭐⭐⭐⭐⭐ | ⭐⭐ | ⭐⭐⭐ |
-| Setup Complexity | Simple | Complex | Complex |
-| Semantic Search | ✅ | ✅ | ❌ |
-| Offline Usage | ✅ | ❌ | ✅ |
-| Portability | File-based | Server-based | Server-based |
-| Scalability | Millions | Millions | Billions |
-| Cost | Free | $$$$ | $$$ |
-
-## 🗺️ Roadmap
-
-- [ ] **v0.2.0** - Multi-language support
-- [ ] **v0.3.0** - Real-time memory updates
-- [ ] **v0.4.0** - Distributed video sharding
-- [ ] **v0.5.0** - Audio and image support
-- [ ] **v1.0.0** - Production-ready with enterprise features
-
-## 📚 Examples
-
-Check out the [examples/](examples/) directory for:
-- Building memory from Wikipedia dumps
-- Creating a personal knowledge base
-- Multi-language support
-- Real-time memory updates
-- Integration with popular LLMs
-
-## 🆘 Getting Help
-
-- 📖 [Documentation](https://github.com/olow304/memvid/wiki) - Comprehensive guides
-- 💬 [Discussions](https://github.com/olow304/memvid/discussions) - Ask questions
-- 🐛 [Issue Tracker](https://github.com/olow304/memvid/issues) - Report bugs
-- 🌟 [Show & Tell](https://github.com/olow304/memvid/discussions/categories/show-and-tell) - Share your projects
-
-## 🔗 Links
-
-- [GitHub Repository](https://github.com/olow304/memvid)
-- [PyPI Package](https://pypi.org/project/memvid)
-- [Changelog](https://github.com/olow304/memvid/releases)
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)"
+     src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
 
+<div style="height: 16px;"></div>
 
-## 📄 License
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
 
-MIT License - see [LICENSE](LICENSE) file for details.
+<p align="center">
+  <strong>Memvid is a single-file memory layer for AI agents with instant retrieval and long-term memory.</strong><br/>
+  Persistent, versioned, and portable memory, without databases.
+</p>
 
-## 🙏 Acknowledgments
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
 
-Created by [Olow304](https://github.com/olow304) and the Memvid community.
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
 
-Built with ❤️ using:
-- [sentence-transformers](https://www.sbert.net/) - State-of-the-art embeddings for semantic search
-- [OpenCV](https://opencv.org/) - Computer vision and video processing
-- [qrcode](https://github.com/lincolnloop/python-qrcode) - QR code generation
-- [FAISS](https://github.com/facebookresearch/faiss) - Efficient similarity search
-- [PyPDF2](https://github.com/py-pdf/pypdf) - PDF text extraction
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/7RUve6Zrrv"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
 
-Special thanks to all contributors who help make Memvid better!
+
+## Benchmark Highlights
+
+**🚀 Higher accuracy than any other memory system :** +35% SOTA on LoCoMo, best-in-class long-horizon conversational recall & reasoning
+
+**🧠 Superior multi-hop & temporal reasoning:**  +76% multi-hop, +56% temporal vs. the industry average
+
+**⚡ Ultra-low latency at scale** 0.025ms P50 and 0.075ms P99, with 1,372× higher throughput than standard
+
+**🔬 Fully reproducible benchmarks:** LoCoMo (10 × ~26K-token conversations), open-source eval, LLM-as-Judge
+
+
+## What is Memvid?
+
+Memvid is a portable AI memory system that packages your data, embeddings, search structure, and metadata into a single file.
+
+Instead of running complex RAG pipelines or server-based vector databases, Memvid enables fast retrieval directly from the file.
+
+The result is a model-agnostic, infrastructure-free memory layer that gives AI agents persistent, long-term memory they can carry anywhere.
+
+    
+## What are Smart Frames?
+
+Memvid draws inspiration from video encoding, not to store video, but to **organize AI memory as an append-only, ultra-efficient sequence of Smart Frames.**
+
+A Smart Frame is an immutable unit that stores content along with timestamps, checksums and basic metadata.
+Frames are grouped in a way that allows efficient compression, indexing, and parallel reads.
+
+This frame-based design enables:
+
+-   Append-only writes without modifying or corrupting existing data
+-   Queries over past memory states
+-   Timeline-style inspection of how knowledge evolves
+-   Crash safety through committed, immutable frames
+-   Efficient compression using techniques adapted from video encoding
+
+The result is a single file that behaves like a rewindable memory timeline for AI systems.
+
+
+## Core Concepts
+
+-   **Living Memory Engine**
+    Continuously append, branch, and evolve memory across sessions.
+
+-   **Capsule Context (`.mv2`)**
+    Self-contained, shareable memory capsules with rules and expiry.
+
+-   **Time-Travel Debugging**
+    Rewind, replay, or branch any memory state.
+
+-   **Smart Recall**
+    Sub-5ms local memory access with predictive caching.
+
+-   **Codec Intelligence**
+    Auto-selects and upgrades compression over time.
+
+
+## Use Cases
+
+Memvid is a portable, serverless memory layer that gives AI agents persistent memory and fast recall. Because it's model-agnostic, multi-modal, and works fully offline, developers are using Memvid across a wide range of real-world applications.
+
+-   Long-Running AI Agents
+-   Enterprise Knowledge Bases
+-   Offline-First AI Systems
+-   Codebase Understanding
+-   Customer Support Agents
+-   Workflow Automation
+-   Sales and Marketing Copilots
+-   Personal Knowledge Assistants
+-   Medical, Legal, and Financial Agents
+-   Auditable and Debuggable AI Workflows
+-   Custom Applications
+
+
+## SDKs & CLI
+
+Use Memvid in your preferred language:
+
+| Package         | Install                     | Links                                                                                                               |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
 
 ---
 
-**Ready to revolutionize your AI memory management? Install Memvid and start building!** 🚀
+## Installation (Rust)
+
+### Requirements
+
+-   **Rust 1.85.0+** — Install from [rustup.rs](https://rustup.rs)
+
+### Add to Your Project
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Feature Flags
+
+| Feature             | Description                                                      |
+| ------------------- | ---------------------------------------------------------------- |
+| `lex`               | Full-text search with BM25 ranking (Tantivy)                     |
+| `pdf_extract`       | Pure Rust PDF text extraction                                    |
+| `vec`               | Vector similarity search (HNSW + local text embeddings via ONNX) |
+| `clip`              | CLIP visual embeddings for image search                          |
+| `whisper`           | Audio transcription with Whisper                                 |
+| `api_embed`         | Cloud API embeddings (OpenAI)                                    |
+| `temporal_track`    | Natural language date parsing ("last Tuesday")                   |
+| `parallel_segments` | Multi-threaded ingestion                                         |
+| `encryption`        | Password-based encryption capsules (.mv2e)                       |
+| `symspell_cleanup`  | Robust PDF text repair (fixes "emp lo yee" -> "employee")        |
+
+Enable features as needed:
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+
+## Quick Start
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Create a new memory file
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Add documents with metadata
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // Search
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## Build
+
+Clone the repository:
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+Build in debug mode:
+
+```bash
+cargo build
+```
+
+Build in release mode (optimized):
+
+```bash
+cargo build --release
+```
+
+Build with specific features:
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## Run Tests
+
+Run all tests:
+
+```bash
+cargo test
+```
+
+Run tests with output:
+
+```bash
+cargo test -- --nocapture
+```
+
+Run a specific test:
+
+```bash
+cargo test test_name
+```
+
+Run integration tests only:
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## Examples
+
+The `examples/` directory contains working examples:
+
+### Basic Usage
+
+Demonstrates create, put, search, and timeline operations:
+
+```bash
+cargo run --example basic_usage
+```
+
+### PDF Ingestion
+
+Ingest and search PDF documents (uses the "Attention Is All You Need" paper):
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### CLIP Visual Search
+
+Image search using CLIP embeddings (requires `clip` feature):
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Whisper Transcription
+
+Audio transcription (requires `whisper` feature):
+
+```bash
+cargo run --example test_whisper --features whisper -- /path/to/audio.mp3
+```
+
+**Available Models:**
+
+| Model                 | Size   | Speed   | Use Case                            |
+| --------------------- | ------ | ------- | ----------------------------------- |
+| `whisper-small-en`    | 244 MB | Slowest | Best accuracy (default)             |
+| `whisper-tiny-en`     | 75 MB  | Fast    | Balanced                            |
+| `whisper-tiny-en-q8k` | 19 MB  | Fastest | Quick testing, resource-constrained |
+
+**Model Selection:**
+
+```bash
+# Default (FP32 small, highest accuracy)
+cargo run --example test_whisper --features whisper -- audio.mp3
+
+# Quantized tiny (75% smaller, faster)
+MEMVID_WHISPER_MODEL=whisper-tiny-en-q8k cargo run --example test_whisper --features whisper -- audio.mp3
+```
+
+**Programmatic Configuration:**
+
+```rust
+use memvid_core::{WhisperConfig, WhisperTranscriber};
+
+// Default FP32 small model
+let config = WhisperConfig::default();
+
+// Quantized tiny model (faster, smaller)
+let config = WhisperConfig::with_quantization();
+
+// Specific model
+let config = WhisperConfig::with_model("whisper-tiny-en-q8k");
+
+let transcriber = WhisperTranscriber::new(&config)?;
+let result = transcriber.transcribe_file("audio.mp3")?;
+println!("{}", result.text);
+```
+
+
+## Text Embedding Models
+
+The `vec` feature includes local text embedding support using ONNX models. Before using local text embeddings, you need to download the model files manually.
+
+### Quick Start: BGE-small (Recommended)
+
+Download the default BGE-small model (384 dimensions, fast and efficient):
+
+```bash
+mkdir -p ~/.cache/memvid/text-models
+
+# Download ONNX model
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5.onnx
+
+# Download tokenizer
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5_tokenizer.json
+```
+
+### Available Models
+
+| Model                   | Dimensions | Size   | Best For        |
+| ----------------------- | ---------- | ------ | --------------- |
+| `bge-small-en-v1.5`     | 384        | ~120MB | Default, fast   |
+| `bge-base-en-v1.5`      | 768        | ~420MB | Better quality  |
+| `nomic-embed-text-v1.5` | 768        | ~530MB | Versatile tasks |
+| `gte-large`             | 1024       | ~1.3GB | Highest quality |
+
+### Other Models
+
+**BGE-base** (768 dimensions):
+```bash
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5.onnx
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5_tokenizer.json
+```
+
+**Nomic** (768 dimensions):
+```bash
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5.onnx
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5_tokenizer.json
+```
+
+**GTE-large** (1024 dimensions):
+```bash
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/gte-large.onnx
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/gte-large_tokenizer.json
+```
+
+### Usage in Code
+
+```rust
+use memvid_core::text_embed::{LocalTextEmbedder, TextEmbedConfig};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// Use default model (BGE-small)
+let config = TextEmbedConfig::default();
+let embedder = LocalTextEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 384);
+
+// Use different model
+let config = TextEmbedConfig::bge_base();
+let embedder = LocalTextEmbedder::new(config)?;
+```
+
+See `examples/text_embedding.rs` for a complete example with similarity computation and search ranking.
+
+### Model Consistency
+
+To prevent accidental model mixing (e.g., querying a BGE-small index with OpenAI embeddings), you can explicitly bind your Memvid instance to a specific model name:
+
+```rust
+// Bind the index to a specific model.
+// If the index was previously created with a different model, this will return an error.
+mem.set_vec_model("bge-small-en-v1.5")?;
+```
+
+This binding is persistent. Once set, future attempts to use a different model name will fail fast with a `ModelMismatch` error.
+
+
+
+## API Embeddings (OpenAI)
+
+The `api_embed` feature enables cloud-based embedding generation using OpenAI's API.
+
+### Setup
+
+Set your OpenAI API key:
+
+```bash
+export OPENAI_API_KEY="sk-..."
+```
+
+### Usage
+
+```rust
+use memvid_core::api_embed::{OpenAIConfig, OpenAIEmbedder};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// Use default model (text-embedding-3-small)
+let config = OpenAIConfig::default();
+let embedder = OpenAIEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 1536);
+
+// Use higher quality model
+let config = OpenAIConfig::large();  // text-embedding-3-large (3072 dims)
+let embedder = OpenAIEmbedder::new(config)?;
+```
+
+### Available Models
+
+| Model                    | Dimensions | Best For                   |
+| ------------------------ | ---------- | -------------------------- |
+| `text-embedding-3-small` | 1536       | Default, fastest, cheapest |
+| `text-embedding-3-large` | 3072       | Highest quality            |
+| `text-embedding-ada-002` | 1536       | Legacy model               |
+
+See `examples/openai_embedding.rs` for a complete example.
+
+
+
+## File Format
+
+Everything lives in a single `.mv2` file:
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  Magic, version, capacity
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  Crash recovery
+├────────────────────────────┤
+│ Data Segments              │  Compressed frames
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy full-text
+├────────────────────────────┤
+│ Vec Index                  │  HNSW vectors
+├────────────────────────────┤
+│ Time Index                 │  Chronological ordering
+├────────────────────────────┤
+│ TOC (Footer)               │  Segment offsets
+└────────────────────────────┘
+```
+
+No `.wal`, `.lock`, `.shm`, or sidecar files. Ever.
+
+See [MV2_SPEC.md](MV2_SPEC.md) for the complete file format specification.
+
+
+
+## Support
+
+Have questions or feedback?
+Email: contact@memvid.com
+
+**Drop a ⭐ to show support**
+
+---
+
+> **Memvid v1 (QR-based memory) is deprecated**
+>
+> If you are referencing QR codes, you are using outdated information.
+>
+> See: https://docs.memvid.com/memvid-v1-deprecation
+
+---
+
+## License
+
+Apache License 2.0 — see the [LICENSE](LICENSE) file for details.

SECURITY.md

@@ -0,0 +1,68 @@
+# Security Policy
+
+## Supported Versions
+
+| Version | Supported          |
+| ------- | ------------------ |
+| 2.x     | :white_check_mark: |
+| < 2.0   | :x:                |
+
+## Reporting a Vulnerability
+
+We take security seriously at Memvid. If you discover a security vulnerability, please report it responsibly.
+
+### How to Report
+
+**Please do NOT open a public GitHub issue for security vulnerabilities.**
+
+Instead, email us at: **security@memvid.com**
+
+Include the following in your report:
+- Description of the vulnerability
+- Steps to reproduce
+- Potential impact
+- Suggested fix (if any)
+
+### What to Expect
+
+- **Acknowledgment**: We will acknowledge your report within 48 hours
+- **Assessment**: We will assess the vulnerability and determine its severity
+- **Fix Timeline**: Critical vulnerabilities will be addressed within 7 days
+- **Disclosure**: We will coordinate with you on public disclosure timing
+- **Credit**: We will credit you in our security advisories (unless you prefer anonymity)
+
+### Scope
+
+The following are in scope:
+- Memory corruption vulnerabilities
+- Data leakage from `.mv2` files
+- Encryption bypass (when using `encryption` feature)
+- Denial of service attacks
+- Path traversal vulnerabilities
+
+### Safe Harbor
+
+We consider security research conducted in good faith to be authorized. We will not pursue legal action against researchers who:
+- Act in good faith
+- Avoid privacy violations
+- Do not access or modify other users' data
+- Report vulnerabilities promptly
+- Give us reasonable time to fix issues before disclosure
+
+## Security Best Practices
+
+When using Memvid:
+
+1. **File Permissions**: Set appropriate file permissions on `.mv2` files
+2. **Encryption**: Use the `encryption` feature for sensitive data
+3. **Validation**: Validate input before ingesting into memory
+4. **Updates**: Keep Memvid updated to the latest version
+
+## Security Features
+
+Memvid includes several security features:
+
+- **Checksums**: Blake3 checksums for data integrity
+- **Signatures**: Ed25519 signatures for authenticity
+- **Encryption**: Optional AES-256-GCM encryption (`.mv2e` capsules)
+- **Crash Safety**: WAL-based recovery prevents corruption

USAGE.md

@@ -1,711 +0,0 @@
-# Memvid Usage Guide
-
-## Table of Contents
-1. [Overview](#overview)
-2. [Installation](#installation)
-3. [Quick Start](#quick-start)
-4. [Architecture](#architecture)
-5. [File Outputs Explained](#file-outputs-explained)
-6. [Core Components](#core-components)
-7. [API Reference](#api-reference)
-8. [Advanced Usage](#advanced-usage)
-9. [Performance Optimization](#performance-optimization)
-10. [Troubleshooting](#troubleshooting)
-
-## Overview
-
-Memvid is a Python library that enables efficient storage and retrieval of text data using QR code videos. It combines:
-- **Text chunking** and semantic embeddings
-- **QR code generation** for data encoding
-- **Video creation** for compact storage
-- **Vector search** for fast retrieval
-- **Conversational AI** interface with context-aware memory
-
-### Key Benefits
-- Store millions of text chunks in a single video file
-- Fast semantic search (< 2 seconds for 1M chunks)
-- No database required - just MP4 + index files
-- Portable and shareable knowledge bases
-- Works with any LLM (OpenAI, Claude, local models)
-
-## Installation
-
-### Prerequisites
-- Python 3.8 or higher
-- FFmpeg (for video encoding)
-- libzbar0 (for QR decoding)
-
-### System Dependencies
-
-**macOS:**
-```bash
-brew install ffmpeg zbar
-```
-
-**Ubuntu/Debian:**
-```bash
-sudo apt-get update
-sudo apt-get install ffmpeg libzbar0
-```
-
-**Windows:**
-- Install FFmpeg from https://ffmpeg.org/download.html
-- Install zbar from https://sourceforge.net/projects/zbar/
-
-### Python Installation
-
-**Option 1: From source (recommended for development)**
-```bash
-# Clone the repository
-git clone https://github.com/your-repo/memvid.git
-cd memvid
-
-# Create virtual environment
-python -m venv .memvid
-source .memvid/bin/activate  # On Windows: .memvid\Scripts\activate
-
-# Install dependencies
-pip install -r requirements.txt
-
-# Install in development mode
-pip install -e .
-```
-
-**Option 2: Direct installation**
-```bash
-pip install memvid
-```
-
-### Verify Installation
-```python
-import memvid
-print(memvid.__version__)
-```
-
-## Quick Start
-
-### 1. Creating a Memory Video
-
-```python
-from memvid import MemvidEncoder
-
-# Create encoder
-encoder = MemvidEncoder()
-
-# Add individual chunks
-chunks = [
-    "Quantum computers use qubits instead of classical bits",
-    "Machine learning models can process billions of parameters",
-    "Cloud computing enables scalable infrastructure"
-]
-encoder.add_chunks(chunks)
-
-# Or add text with automatic chunking
-long_text = """Your long document text here..."""
-encoder.add_text(long_text, chunk_size=200, overlap=50)
-
-# Build video and index
-encoder.build_video("output/knowledge.mp4", "output/knowledge_index.json")
-```
-
-### 2. Searching the Memory
-
-```python
-from memvid import MemvidRetriever
-
-# Load retriever
-retriever = MemvidRetriever("output/knowledge.mp4", "output/knowledge_index.json")
-
-# Search for relevant chunks
-results = retriever.search("quantum computing", top_k=5)
-for chunk in results:
-    print(chunk)
-```
-
-### 3. Interactive Chat
-
-```python
-from memvid import MemvidChat
-
-# Initialize chat (set OPENAI_API_KEY environment variable)
-chat = MemvidChat("output/knowledge.mp4", "output/knowledge_index.json")
-chat.start_session()
-
-# Have a conversation
-response = chat.chat("What do you know about quantum computers?")
-print(response)
-```
-
-## Architecture
-
-### Data Flow Pipeline
-
-```
-1. Text Input → Chunking → Embeddings → QR Codes → Video Frames → MP4 File
-                    ↓
-                Vector Index → FAISS Index → JSON Metadata
-
-2. Query → Embedding → Vector Search → Frame Numbers → QR Decode → Text
-                             ↓
-                     Retrieved Context → LLM → Response
-```
-
-### System Components
-
-```
-memvid/
-├── encoder.py      # Text → QR Video conversion
-├── retriever.py    # Video → Text retrieval
-├── chat.py         # Conversational interface
-├── index.py        # Vector indexing & search
-├── utils.py        # QR & video utilities
-└── config.py       # Configuration management
-```
-
-## File Outputs Explained
-
-When you build a memory video, two files are created:
-
-### 1. MP4 Video File (`memory.mp4`)
-
-**What it is:** A video file where each frame is a QR code containing encoded text data.
-
-**Structure:**
-- Each frame = 1 QR code
-- Each QR code = 1 text chunk + metadata
-- Frame rate = 30 FPS (configurable)
-- Resolution = 512x512 pixels (configurable)
-
-**Contents of each QR code:**
-```json
-{
-    "id": 0,
-    "text": "The actual text chunk content...",
-    "metadata": {
-        "source": "optional source info",
-        "timestamp": "2024-01-01T00:00:00"
-    }
-}
-```
-
-**Why video?**
-- Efficient compression (H.264/H.265)
-- Portable and shareable
-- Streamable over networks
-- Supported everywhere
-
-### 2. Index Files (`memory_index.json` + `memory_index.faiss`)
-
-**What they are:** Search index files for fast retrieval.
-
-**memory_index.json structure:**
-```json
-{
-    "metadata": [
-        {
-            "chunk_id": 0,
-            "text": "Text preview (first 200 chars)...",
-            "frame": 0,
-            "char_count": 250,
-            "word_count": 45
-        }
-    ],
-    "chunk_to_frame": {
-        "0": 0,
-        "1": 1
-    },
-    "frame_to_chunks": {
-        "0": [0],
-        "1": [1]
-    },
-    "config": {
-        "embedding": {
-            "model": "all-MiniLM-L6-v2",
-            "dimension": 384
-        },
-        "index": {
-            "type": "Flat"
-        }
-    }
-}
-```
-
-**memory_index.faiss:** Binary vector index for similarity search.
-
-### File Size Comparison
-
-For 10,000 text chunks (average 200 chars each):
-- Raw text: ~2 MB
-- MP4 video: ~15-20 MB (with compression)
-- FAISS index: ~15 MB (384-dim vectors)
-- JSON metadata: ~3 MB
-
-## Core Components
-
-### 1. MemvidEncoder
-
-Handles text processing and video creation.
-
-```python
-from memvid import MemvidEncoder
-
-encoder = MemvidEncoder(config={
-    "qr": {
-        "version": 10,          # QR code version (1-40)
-        "error_correction": "H", # Error correction level (L/M/Q/H)
-        "box_size": 10,         # Pixel size of QR boxes
-        "border": 4             # Border size
-    },
-    "video": {
-        "fps": 30,              # Frames per second
-        "codec": "libx264",     # Video codec
-        "crf": 23,              # Quality (lower = better)
-        "preset": "medium"      # Encoding speed
-    }
-})
-
-# Add data
-encoder.add_chunks(["chunk1", "chunk2"])
-encoder.add_text("long text", chunk_size=200, overlap=50)
-
-# Get statistics
-stats = encoder.get_stats()
-print(f"Total chunks: {stats['total_chunks']}")
-print(f"Total size: {stats['total_characters']} chars")
-
-# Build video
-build_stats = encoder.build_video(
-    "output.mp4", 
-    "output_index.json",
-    show_progress=True
-)
-```
-
-### 2. MemvidRetriever
-
-Handles search and text extraction from videos.
-
-```python
-from memvid import MemvidRetriever
-
-retriever = MemvidRetriever(
-    "video.mp4", 
-    "index.json",
-    config={
-        "retrieval": {
-            "cache_size": 1000,      # Frame cache size
-            "batch_size": 50,        # Parallel decode batch
-            "max_workers": 4         # Thread pool size
-        }
-    }
-)
-
-# Basic search
-chunks = retriever.search("quantum computing", top_k=5)
-
-# Search with metadata
-results = retriever.search_with_metadata("AI", top_k=3)
-for result in results:
-    print(f"Score: {result['score']:.3f}")
-    print(f"Text: {result['text']}")
-    print(f"Frame: {result['frame']}")
-
-# Get specific chunk
-chunk = retriever.get_chunk_by_id(42)
-
-# Stats
-stats = retriever.get_stats()
-print(f"Cache hit rate: {stats['cache_hit_rate']:.2%}")
-```
-
-### 3. MemvidChat
-
-Conversational interface with memory.
-
-```python
-from memvid import MemvidChat
-import os
-
-# Set API key
-os.environ['OPENAI_API_KEY'] = 'your-key-here'
-
-chat = MemvidChat(
-    "video.mp4",
-    "index.json",
-    llm_model="gpt-3.5-turbo",  # or gpt-4, claude-3, etc.
-    config={
-        "chat": {
-            "context_chunks": 5,      # Chunks per query
-            "max_history": 10        # Conversation history
-        },
-        "llm": {
-            "temperature": 0.7,
-            "max_tokens": 500
-        }
-    }
-)
-
-# Start session
-chat.start_session()
-
-# Chat
-response = chat.chat("What's in this knowledge base?")
-
-# Direct search
-results = chat.search_context("specific topic", top_k=10)
-
-# Export conversation
-chat.export_session("session.json")
-
-# Get statistics
-stats = chat.get_stats()
-```
-
-## Advanced Usage
-
-### Custom Chunking Strategies
-
-```python
-from memvid import MemvidEncoder
-import re
-
-def custom_chunker(text, max_size=200):
-    """Chunk by sentences, respecting max size"""
-    sentences = re.split(r'(?<=[.!?])\s+', text)
-    chunks = []
-    current = ""
-    
-    for sentence in sentences:
-        if len(current) + len(sentence) > max_size and current:
-            chunks.append(current.strip())
-            current = sentence
-        else:
-            current += " " + sentence
-    
-    if current:
-        chunks.append(current.strip())
-    
-    return chunks
-
-encoder = MemvidEncoder()
-chunks = custom_chunker(long_text)
-encoder.add_chunks(chunks)
-```
-
-### Metadata Enrichment
-
-```python
-encoder = MemvidEncoder()
-
-# Add chunks with metadata
-chunks_with_meta = [
-    {
-        "text": "Quantum computing breakthrough...",
-        "metadata": {
-            "source": "arxiv:2024.1234",
-            "date": "2024-01-15",
-            "category": "quantum",
-            "author": "Smith et al."
-        }
-    }
-]
-
-for item in chunks_with_meta:
-    encoder.add_chunk(item["text"], metadata=item["metadata"])
-```
-
-### Custom Embedding Models
-
-```python
-from sentence_transformers import SentenceTransformer
-
-# Use a different embedding model
-encoder = MemvidEncoder(config={
-    "embedding": {
-        "model": "all-mpnet-base-v2",  # Higher quality
-        "dimension": 768,
-        "batch_size": 32
-    }
-})
-```
-
-### Batch Processing Large Datasets
-
-```python
-import os
-from pathlib import Path
-
-def process_directory(dir_path, output_prefix):
-    """Process all text files in directory"""
-    encoder = MemvidEncoder()
-    
-    for file_path in Path(dir_path).glob("*.txt"):
-        with open(file_path, 'r') as f:
-            text = f.read()
-            encoder.add_text(
-                text, 
-                chunk_size=300,
-                overlap=50,
-                metadata={"source": file_path.name}
-            )
-    
-    # Build video
-    encoder.build_video(
-        f"{output_prefix}.mp4",
-        f"{output_prefix}_index.json"
-    )
-
-# Process all documents
-process_directory("documents/", "output/knowledge_base")
-```
-
-### Multi-Video Federation
-
-```python
-class MemvidFederation:
-    """Search across multiple video memories"""
-    
-    def __init__(self):
-        self.retrievers = {}
-    
-    def add_memory(self, name, video_file, index_file):
-        self.retrievers[name] = MemvidRetriever(video_file, index_file)
-    
-    def search_all(self, query, top_k=5):
-        all_results = []
-        
-        for name, retriever in self.retrievers.items():
-            results = retriever.search_with_metadata(query, top_k)
-            for r in results:
-                r['source'] = name
-                all_results.append(r)
-        
-        # Sort by score
-        all_results.sort(key=lambda x: x['score'], reverse=True)
-        return all_results[:top_k]
-
-# Use federation
-fed = MemvidFederation()
-fed.add_memory("tech", "tech_memory.mp4", "tech_index.json")
-fed.add_memory("science", "science_memory.mp4", "science_index.json")
-results = fed.search_all("quantum mechanics")
-```
-
-## Performance Optimization
-
-### 1. Encoding Performance
-
-```python
-# Parallel encoding for large datasets
-encoder = MemvidEncoder(config={
-    "encoding": {
-        "max_workers": 8,        # Parallel QR generation
-        "batch_size": 100        # Process in batches
-    },
-    "video": {
-        "preset": "ultrafast",   # Faster encoding
-        "crf": 28               # Lower quality = faster
-    }
-})
-```
-
-### 2. Retrieval Performance
-
-```python
-# Optimize for speed
-retriever = MemvidRetriever(video_file, index_file, config={
-    "retrieval": {
-        "cache_size": 5000,      # Larger cache
-        "batch_size": 100,       # Bigger batches
-        "max_workers": 8,        # More threads
-        "preload_frames": True   # Preload common frames
-    }
-})
-
-# Warm up cache with common queries
-common_queries = ["AI", "machine learning", "data"]
-for query in common_queries:
-    retriever.search(query, top_k=10)
-```
-
-### 3. Index Optimization
-
-```python
-# Use approximate search for large datasets
-encoder = MemvidEncoder(config={
-    "index": {
-        "type": "IVF",          # Inverted file index
-        "nlist": 100,           # Number of clusters
-        "nprobe": 10           # Clusters to search
-    }
-})
-
-# Or use HNSW for better quality
-encoder = MemvidEncoder(config={
-    "index": {
-        "type": "HNSW",         # Hierarchical NSW
-        "M": 32,                # Number of connections
-        "ef_construction": 200  # Construction parameter
-    }
-})
-```
-
-### 4. Memory Management
-
-```python
-# For very large videos
-retriever = MemvidRetriever(video_file, index_file, config={
-    "retrieval": {
-        "mmap_video": True,      # Memory-map video file
-        "chunk_buffer": 1000,    # Buffer size
-        "gc_interval": 100       # Garbage collection
-    }
-})
-```
-
-## Troubleshooting
-
-### Common Issues
-
-#### 1. "huggingface/tokenizers" Warning
-```bash
-# Set before running
-export TOKENIZERS_PARALLELISM=false
-
-# Or in Python
-import os
-os.environ['TOKENIZERS_PARALLELISM'] = 'false'
-```
-
-#### 2. QR Decode Failures
-```python
-# Increase QR code quality
-encoder = MemvidEncoder(config={
-    "qr": {
-        "error_correction": "H",  # Highest correction
-        "version": 15,           # Larger QR codes
-        "box_size": 15          # Bigger pixels
-    }
-})
-```
-
-#### 3. Video Codec Issues
-```python
-# Try different codecs
-encoder = MemvidEncoder(config={
-    "video": {
-        "codec": "libx265",     # Or "h264_nvenc" for NVIDIA
-        "pixel_format": "yuv420p"
-    }
-})
-```
-
-#### 4. Memory Issues with Large Videos
-```python
-# Enable streaming mode
-retriever = MemvidRetriever(video_file, index_file, config={
-    "retrieval": {
-        "streaming_mode": True,
-        "buffer_size": 100      # Frames in memory
-    }
-})
-```
-
-#### 5. Slow Search Performance
-```python
-# Debug performance
-stats = retriever.get_stats()
-print(f"Average decode time: {stats['avg_decode_time']}ms")
-print(f"Cache hit rate: {stats['cache_hit_rate']:.2%}")
-
-# Enable profiling
-retriever.enable_profiling()
-results = retriever.search("test query")
-print(retriever.get_profile_stats())
-```
-
-### Debugging Tips
-
-```python
-# Enable detailed logging
-import logging
-logging.basicConfig(level=logging.DEBUG)
-
-# Check video integrity
-from memvid.utils import verify_video
-is_valid, stats = verify_video("output.mp4")
-print(f"Valid: {is_valid}")
-print(f"Readable frames: {stats['readable_frames']}/{stats['total_frames']}")
-
-# Test QR encoding/decoding
-from memvid.utils import create_qr_code, decode_qr
-qr_img = create_qr_code("test data")
-decoded = decode_qr(qr_img)
-assert decoded == "test data"
-```
-
-## Best Practices
-
-1. **Chunk Size**: 100-500 characters works best
-2. **Overlap**: 20-30% overlap prevents context loss
-3. **Video FPS**: 30 FPS is optimal (higher = larger files)
-4. **Cache Size**: Set to 10% of total frames
-5. **Batch Size**: 50-100 for parallel processing
-6. **Error Correction**: Use "H" for archival, "L" for speed
-
-## Example Projects
-
-### 1. Personal Knowledge Base
-```python
-# Build from markdown files
-from pathlib import Path
-
-encoder = MemvidEncoder()
-for md_file in Path("notes/").glob("**/*.md"):
-    with open(md_file) as f:
-        encoder.add_text(f.read(), metadata={"file": str(md_file)})
-
-encoder.build_video("personal_kb.mp4", "personal_kb_index.json")
-```
-
-### 2. Documentation Search
-```python
-# Create searchable docs
-encoder = MemvidEncoder()
-encoder.add_text(api_docs, metadata={"type": "api"})
-encoder.add_text(tutorials, metadata={"type": "tutorial"})
-encoder.build_video("docs.mp4", "docs_index.json")
-
-# Search with filtering
-retriever = MemvidRetriever("docs.mp4", "docs_index.json")
-api_results = [r for r in retriever.search_with_metadata("authentication") 
-               if r.get("metadata", {}).get("type") == "api"]
-```
-
-### 3. Research Paper Archive
-```python
-# Archive papers with citations
-papers = load_papers()  # Your paper loader
-encoder = MemvidEncoder()
-
-for paper in papers:
-    encoder.add_text(
-        paper["abstract"],
-        metadata={
-            "title": paper["title"],
-            "authors": paper["authors"],
-            "year": paper["year"],
-            "doi": paper["doi"]
-        }
-    )
-
-encoder.build_video("papers.mp4", "papers_index.json")
-```
-
-## Contributing
-
-Contributions welcome! See CONTRIBUTING.md for guidelines.
-
-## License
-
-MIT License - see LICENSE file for details.

benches/search_precision_benchmark.rs

@@ -0,0 +1,176 @@
+//! Search precision benchmarks for implicit AND operator change.
+//!
+//! This benchmark suite measures the performance impact of changing the implicit
+//! query operator from OR to AND. It verifies that the precision improvement
+//! (33% → 100%) comes with no query latency regression.
+//!
+//! # Benchmarks
+//!
+//! - `query_two_words`: Measures latency for simple two-word queries
+//! - `precision_calculation`: Measures precision metrics and filtering overhead
+//! - `result_count`: Measures result set size impact
+//!
+//! # Running
+//!
+//! ```bash
+//! cargo bench --bench search_precision_benchmark --features lex
+//! ```
+
+use criterion::{Criterion, criterion_group, criterion_main};
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+use std::time::Instant;
+
+/// Setup test corpus
+fn setup_corpus(size: usize) -> std::path::PathBuf {
+    let temp_file = std::env::temp_dir().join(format!("bench_{}.mv2", size));
+    let _ = std::fs::remove_file(&temp_file);
+
+    let mut mem = Memvid::create(&temp_file).unwrap();
+
+    let topics = [
+        "machine learning neural networks",
+        "python programming development",
+        "machine learning with python",
+        "rust systems programming",
+        "web development javascript",
+    ];
+
+    for i in 0..size {
+        let content = format!("Document {} about {}", i, topics[i % topics.len()]);
+        mem.put_bytes_with_options(
+            content.as_bytes(),
+            PutOptions::builder().title(format!("Doc {}", i)).build(),
+        )
+        .unwrap();
+
+        if (i + 1) % 100 == 0 {
+            mem.commit().unwrap();
+        }
+    }
+    mem.commit().unwrap();
+    temp_file
+}
+
+fn bench_query_latency(c: &mut Criterion) {
+    let corpus_path = setup_corpus(1000);
+
+    c.bench_function("query_two_words", |b| {
+        b.iter_custom(|iters| {
+            let mut total = std::time::Duration::ZERO;
+            for _ in 0..iters {
+                let mut mem = Memvid::open(&corpus_path).unwrap(); // FIX: mut
+                let start = Instant::now();
+                let _results = mem
+                    .search(SearchRequest {
+                        query: "machine learning".to_string(),
+                        top_k: 10,
+                        snippet_chars: 200,
+                        uri: None,
+                        scope: None,
+                        cursor: None,
+                        #[cfg(feature = "temporal_track")]
+                        temporal: None,
+                        as_of_frame: None,
+                        as_of_ts: None,
+                        no_sketch: false,
+                        acl_context: None,
+                        acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+                    })
+                    .unwrap();
+                total += start.elapsed();
+            }
+            total
+        });
+    });
+
+    std::fs::remove_file(&corpus_path).ok();
+}
+
+fn bench_precision(c: &mut Criterion) {
+    let corpus_path = setup_corpus(1000);
+
+    c.bench_function("precision_calculation", |b| {
+        b.iter_custom(|iters| {
+            let mut total = std::time::Duration::ZERO;
+            for _ in 0..iters {
+                let mut mem = Memvid::open(&corpus_path).unwrap(); // FIX: mut
+                let start = Instant::now();
+                let results = mem
+                    .search(SearchRequest {
+                        query: "machine python".to_string(),
+                        top_k: 100,
+                        snippet_chars: 200,
+                        uri: None,
+                        scope: None,
+                        cursor: None,
+                        #[cfg(feature = "temporal_track")]
+                        temporal: None,
+                        as_of_frame: None,
+                        as_of_ts: None,
+                        no_sketch: false,
+                        acl_context: None,
+                        acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+                    })
+                    .unwrap();
+
+                let _relevant = results
+                    .hits
+                    .iter()
+                    .filter(|hit| {
+                        let text = hit.text.to_lowercase();
+                        text.contains("machine") && text.contains("python")
+                    })
+                    .count();
+
+                total += start.elapsed();
+            }
+            total
+        });
+    });
+
+    std::fs::remove_file(&corpus_path).ok();
+}
+
+fn bench_result_count(c: &mut Criterion) {
+    let corpus_path = setup_corpus(1000);
+
+    c.bench_function("result_count", |b| {
+        b.iter_custom(|iters| {
+            let mut total = std::time::Duration::ZERO;
+            for _ in 0..iters {
+                let mut mem = Memvid::open(&corpus_path).unwrap(); // FIX: mut
+                let start = Instant::now();
+                let results = mem
+                    .search(SearchRequest {
+                        query: "machine learning".to_string(),
+                        top_k: 100,
+                        snippet_chars: 200,
+                        uri: None,
+                        scope: None,
+                        cursor: None,
+                        #[cfg(feature = "temporal_track")]
+                        temporal: None,
+                        as_of_frame: None,
+                        as_of_ts: None,
+                        no_sketch: false,
+                        acl_context: None,
+                        acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+                    })
+                    .unwrap();
+                let _count = results.hits.len();
+                total += start.elapsed();
+            }
+            total
+        });
+    });
+
+    std::fs::remove_file(&corpus_path).ok();
+}
+
+criterion_group!(
+    benches,
+    bench_query_latency,
+    bench_precision,
+    bench_result_count
+);
+criterion_main!(benches);

benches/vec_search_benchmark.rs

@@ -0,0 +1,145 @@
+use criterion::{Criterion, black_box, criterion_group, criterion_main};
+use memvid_core::types::FrameId;
+use memvid_core::vec::{VecDocument, VecIndex, VecIndexBuilder};
+
+fn generate_vectors(count: usize, dim: usize) -> Vec<Vec<f32>> {
+    let mut vectors = Vec::with_capacity(count);
+    for _ in 0..count {
+        let mut vec = Vec::with_capacity(dim);
+        for _ in 0..dim {
+            vec.push(fastrand::f32());
+        }
+        vectors.push(vec);
+    }
+    vectors
+}
+
+fn bench_search_10k(c: &mut Criterion) {
+    let count = 10_000;
+    let dim = 128; // Smaller dimension for faster setup in benchmarks
+    let vectors = generate_vectors(count, dim);
+    let query = generate_vectors(1, dim).pop().unwrap();
+
+    // Build HNSW Index (via Builder which triggers HNSW for > 1000)
+    let mut builder = VecIndexBuilder::new();
+    for (i, vec) in vectors.iter().enumerate() {
+        builder.add_document(i as FrameId, vec.clone());
+    }
+    let artifact = builder.finish().expect("finish hnsw");
+    let hnsw_index = VecIndex::decode(&artifact.bytes).expect("decode hnsw");
+
+    // Build Brute Force Index (Force Uncompressed)
+    let documents: Vec<VecDocument> = vectors
+        .iter()
+        .enumerate()
+        .map(|(i, vec)| VecDocument {
+            frame_id: i as FrameId,
+            embedding: vec.clone(),
+        })
+        .collect();
+    let brute_index = VecIndex::Uncompressed { documents };
+
+    let mut group = c.benchmark_group("search_10k");
+
+    group.bench_function("hnsw", |b| {
+        b.iter(|| {
+            hnsw_index.search(black_box(&query), black_box(10));
+        })
+    });
+
+    group.bench_function("brute_force", |b| {
+        b.iter(|| {
+            brute_index.search(black_box(&query), black_box(10));
+        })
+    });
+
+    group.finish();
+}
+
+fn bench_search_50k(c: &mut Criterion) {
+    let count = 50_000;
+    let dim = 128;
+    let vectors = generate_vectors(count, dim);
+    let query = generate_vectors(1, dim).pop().unwrap();
+
+    let mut builder = VecIndexBuilder::new();
+    for (i, vec) in vectors.iter().enumerate() {
+        builder.add_document(i as FrameId, vec.clone());
+    }
+    let artifact = builder.finish().expect("finish hnsw");
+    let hnsw_index = VecIndex::decode(&artifact.bytes).expect("decode hnsw");
+
+    let documents: Vec<VecDocument> = vectors
+        .iter()
+        .enumerate()
+        .map(|(i, vec)| VecDocument {
+            frame_id: i as FrameId,
+            embedding: vec.clone(),
+        })
+        .collect();
+    let brute_index = VecIndex::Uncompressed { documents };
+
+    let mut group = c.benchmark_group("search_50k");
+
+    group.bench_function("hnsw", |b| {
+        b.iter(|| {
+            hnsw_index.search(black_box(&query), black_box(10));
+        })
+    });
+
+    group.bench_function("brute_force", |b| {
+        b.iter(|| {
+            brute_index.search(black_box(&query), black_box(10));
+        })
+    });
+
+    group.finish();
+}
+
+fn bench_search_100k(c: &mut Criterion) {
+    let count = 100_000;
+    let dim = 128;
+    let vectors = generate_vectors(count, dim);
+    let query = generate_vectors(1, dim).pop().unwrap();
+
+    let mut builder = VecIndexBuilder::new();
+    for (i, vec) in vectors.iter().enumerate() {
+        builder.add_document(i as FrameId, vec.clone());
+    }
+    let artifact = builder.finish().expect("finish hnsw");
+    let hnsw_index = VecIndex::decode(&artifact.bytes).expect("decode hnsw");
+
+    let documents: Vec<VecDocument> = vectors
+        .iter()
+        .enumerate()
+        .map(|(i, vec)| VecDocument {
+            frame_id: i as FrameId,
+            embedding: vec.clone(),
+        })
+        .collect();
+    let brute_index = VecIndex::Uncompressed { documents };
+
+    let mut group = c.benchmark_group("search_100k");
+
+    group.bench_function("hnsw", |b| {
+        b.iter(|| {
+            let _ = hnsw_index.search(black_box(&query), black_box(10));
+        })
+    });
+
+    group.bench_function("brute_force", |b| {
+        b.iter(|| {
+            let _ = brute_index.search(black_box(&query), black_box(10));
+        })
+    });
+
+    group.finish();
+}
+
+criterion_group!(
+    benches,
+    bench_search_10k,
+    bench_search_50k,
+    bench_search_100k
+);
+criterion_main!(benches);

docker/README.md

@@ -0,0 +1,82 @@
+# Docker Images for Memvid
+
+This directory contains Docker configurations for Memvid components.
+
+## Available Images
+
+### Memvid Core (`core/`)
+
+The Memvid Core Docker images provide containerized Rust development, testing, and build environments for the `memvid-core` library.
+
+**Quick Start:**
+```bash
+# Development environment
+cd core
+docker-compose up -d dev
+docker-compose exec dev bash
+
+# Run tests
+docker-compose run --rm test
+
+# Build release
+docker-compose run --rm build
+```
+
+For detailed usage, see [core/README.md](core/README.md).
+
+### Memvid CLI (`cli/`)
+
+The Memvid CLI Docker image provides a containerized version of the `memvid-cli` tool, allowing you to run Memvid commands without installing Node.js or dealing with platform-specific binaries.
+
+**Quick Start:**
+
+```bash
+# Pull the image
+docker pull memvid/cli
+
+# Create a memory
+docker run --rm -v $(pwd):/data memvid/cli create my-memory.mv2
+
+# Add documents
+docker run --rm -v $(pwd):/data memvid/cli put my-memory.mv2 --input doc.pdf
+
+# Search
+docker run --rm -v $(pwd):/data memvid/cli find my-memory.mv2 --query "search"
+```
+
+For detailed usage instructions, examples, and Docker Compose configurations, see [cli/README.md](cli/README.md).
+
+## Building Images
+
+### Build CLI Image Locally
+
+```bash
+cd cli
+docker build -t memvid/cli:test .
+```
+
+## Publishing
+
+Docker images are automatically built and published to Docker Hub via GitHub Actions when tags are pushed. See `.github/workflows/docker-release.yml` for the CI/CD configuration.
+
+**Image Registry:**
+- Docker Hub: `memvid/cli`
+- Tags: `latest`, `2.0.129`, and version-specific tags
+
+## Architecture Support
+
+The CLI image supports multi-architecture builds:
+- `linux/amd64`
+- `linux/arm64`
+
+## Security
+
+The CLI image runs as a non-root user (`memvid`) for improved security. When mounting volumes, ensure your host directories have appropriate permissions.
+
+## Links
+
+- [Core Documentation](core/README.md)
+- [CLI Documentation](cli/README.md)
+- [CLI Testing Guide](cli/TESTING.md)
+- [Main Project README](../README.md)
+- [Memvid Documentation](https://docs.memvid.com)

docker/cli/.dockerignore

@@ -0,0 +1,7 @@
+node_modules
+npm-debug.log
+.git
+.gitignore
+*.md
+.env
+.DS_Store

docker/cli/Dockerfile

@@ -0,0 +1,33 @@
+# Memvid CLI Docker Image
+FROM ubuntu:24.04
+
+LABEL org.opencontainers.image.title="Memvid CLI"
+LABEL org.opencontainers.image.description="AI memory CLI with semantic search"
+LABEL org.opencontainers.image.url="https://memvid.com"
+LABEL org.opencontainers.image.source="https://github.com/memvid/memvid"
+LABEL org.opencontainers.image.vendor="Memvid"
+LABEL org.opencontainers.image.version="2.0.131"
+
+# Install system deps + Node.js 24
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    ca-certificates \
+    curl \
+    libssl3 \
+    && curl -fsSL https://deb.nodesource.com/setup_24.x | bash - \
+    && apt-get install -y --no-install-recommends nodejs \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install memvid CLI (pinned version)
+RUN npm install -g memvid-cli@latest \
+    && npm cache clean --force
+
+# Create non-root user (recommended)
+RUN useradd -m memvid
+USER memvid
+
+# Working directory for memory files
+WORKDIR /data
+
+# Entrypoint
+ENTRYPOINT ["memvid"]
+CMD ["--help"]

docker/cli/README.md

@@ -0,0 +1,101 @@
+# Memvid CLI Docker Image
+
+AI memory CLI with crash-safe, single-file storage and semantic search.
+
+## Quick Start
+
+```bash
+# Pull the image
+docker pull memvid/cli
+
+# Create a memory
+docker run --rm -v $(pwd):/data memvid/cli create my-memory.mv2
+
+# Add documents
+docker run --rm -v $(pwd):/data memvid/cli put my-memory.mv2 --input doc.pdf
+
+# Search
+docker run --rm -v $(pwd):/data memvid/cli find my-memory.mv2 --query "search"
+```
+
+## Basic Commands
+
+```bash
+# Show help
+docker run --rm memvid/cli
+
+# Show version
+docker run --rm memvid/cli --version
+
+# Create a memory file (mount local directory)
+docker run --rm -v $(pwd):/data memvid/cli create my-memory.mv2
+
+# Ingest a document
+docker run --rm -v $(pwd):/data memvid/cli put my-memory.mv2 --input document.pdf
+
+# Search the memory
+docker run --rm -v $(pwd):/data memvid/cli find my-memory.mv2 --query "search term"
+
+# Ask questions (requires API key for LLM)
+docker run --rm -v $(pwd):/data \
+  -e OPENAI_API_KEY="sk-..." \
+  memvid/cli ask my-memory.mv2 "What is this about?" -m openai
+
+# View stats
+docker run --rm -v $(pwd):/data memvid/cli stats my-memory.mv2
+```
+
+## With API Keys
+
+```bash
+# Pass Memvid API key for cloud features
+docker run --rm -v $(pwd):/data \
+  -e MEMVID_API_KEY="mv2_..." \
+  -e OPENAI_API_KEY="sk-..." \
+  memvid/cli ask my-memory.mv2 "your question"
+```
+
+## Shell Alias (Recommended)
+
+Add to `~/.bashrc` or `~/.zshrc`:
+
+```bash
+alias memvid='docker run --rm -v $(pwd):/data -e MEMVID_API_KEY -e OPENAI_API_KEY memvid/cli'
+```
+
+Then use normally:
+
+```bash
+memvid create my-memory.mv2
+memvid put my-memory.mv2 --input docs/
+memvid find my-memory.mv2 --query "hello"
+```
+
+## Docker Compose Example
+
+```yaml
+version: '3.8'
+services:
+  memvid:
+    image: memvid/cli:latest
+    volumes:
+      - ./data:/data
+    environment:
+      - MEMVID_API_KEY=${MEMVID_API_KEY}
+      - OPENAI_API_KEY=${OPENAI_API_KEY}
+    entrypoint: ["memvid"]
+    command: ["stats", "my-memory.mv2"]
+```
+
+## Features
+
+- Single-file `.mv2` storage
+- Semantic + lexical search
+- RAG question answering
+- PDF, DOCX, images, audio support
+
+## Links
+
+- Website: https://memvid.com
+- Docs: https://docs.memvid.com
+- GitHub: https://github.com/memvid/memvid

docker/cli/TESTING.md

@@ -0,0 +1,108 @@
+# Testing Memvid CLI Docker Image
+
+## Quick Test
+
+### 1. Build the Image
+
+```bash
+cd docker/cli
+docker build -t memvid/cli:test .
+```
+
+### 2. Test Basic Commands
+
+```bash
+# Test help command
+docker run --rm memvid/cli:test --help
+
+# Test version (if available)
+docker run --rm memvid/cli:test --version
+```
+
+### 3. Test with Volume Mount
+
+```bash
+# Create a test directory
+mkdir -p /tmp/memvid-test
+cd /tmp/memvid-test
+
+# Create a test document
+echo "This is a test document about AI and machine learning." > test.txt
+
+# Create a memory file
+docker run --rm \
+  -v $(pwd):/data \
+  memvid/cli:test create test-memory.mv2
+
+# Add the document
+docker run --rm \
+  -v $(pwd):/data \
+  memvid/cli:test put test-memory.mv2 --input test.txt
+
+# Search the memory
+docker run --rm \
+  -v $(pwd):/data \
+  memvid/cli:test find test-memory.mv2 --query "AI"
+
+# View stats
+docker run --rm \
+  -v $(pwd):/data \
+  memvid/cli:test stats test-memory.mv2
+```
+
+### 4. Test with API Keys (if needed)
+
+```bash
+docker run --rm \
+  -v $(pwd):/data \
+  -e OPENAI_API_KEY="sk-..." \
+  memvid/cli:test ask test-memory.mv2 "What is this about?" -m openai
+```
+
+## Automated Testing
+
+Run the test script:
+
+```bash
+cd docker/cli
+./test.sh
+```
+
+## Multi-Architecture Testing
+
+To test multi-arch builds locally (requires buildx):
+
+```bash
+# Create builder
+docker buildx create --name memvid-builder --use
+
+# Build for specific platform
+docker buildx build \
+  --platform linux/amd64 \
+  --tag memvid/cli:test-amd64 \
+  --load \
+  .
+
+# Test the amd64 image
+docker run --rm memvid/cli:test-amd64 --help
+```
+
+## Troubleshooting
+
+### Image not found
+Make sure you've built the image:
+```bash
+docker build -t memvid/cli:test docker/cli
+```
+
+### Permission errors
+Ensure Docker has permission to access the mounted directory:
+```bash
+chmod 755 /path/to/your/directory
+```
+
+### CLI not found
+Verify the npm package exists:
+```bash
+docker run --rm memvid/cli:test which memvid
+```

docker/cli/test.sh

@@ -0,0 +1,44 @@
+#!/bin/bash
+# Test script for Memvid CLI Docker image
+
+set -e
+
+IMAGE_NAME="memvid/cli:test"
+DOCKERFILE_DIR="$(cd "$(dirname "$0")" && pwd)"
+
+echo "🐳 Building Docker image..."
+docker build -t "$IMAGE_NAME" "$DOCKERFILE_DIR"
+
+echo ""
+echo "✅ Build complete! Testing basic commands..."
+echo ""
+
+echo "1️⃣ Testing help command..."
+docker run --rm "$IMAGE_NAME" --help | head -5
+
+echo ""
+echo "2️⃣ Testing version command..."
+docker run --rm "$IMAGE_NAME" --version || echo "Version command not available"
+
+echo ""
+echo "3️⃣ Testing with volume mount (create test memory)..."
+TEST_DIR=$(mktemp -d)
+cd "$TEST_DIR"
+
+# Create a test document
+echo "This is a test document about artificial intelligence and machine learning." > test.txt
+
+docker run --rm \
+  -v "$TEST_DIR":/data \
+  "$IMAGE_NAME" create test-memory.mv2 || echo "Create command may require additional setup"
+
+echo ""
+echo "🧹 Cleaning up test directory..."
+rm -rf "$TEST_DIR"
+
+echo ""
+echo "✅ Basic tests complete!"
+echo ""
+echo "To test manually, run:"
+echo "  docker run --rm $IMAGE_NAME --help"
+echo "  docker run --rm -v \$(pwd):/data $IMAGE_NAME create my-memory.mv2"

docker/core/DOCKER.md

@@ -0,0 +1,215 @@
+# Docker Setup for Memvid Core
+
+This document describes how to use Docker with the Memvid Core Rust library.
+
+## Quick Start
+
+### Development Environment
+
+```bash
+# From project root or docker/core directory
+cd docker/core
+docker-compose up -d dev
+
+# Enter the container
+docker-compose exec dev bash
+
+# Inside container, you can run:
+cargo build
+cargo test
+cargo run --example basic_usage
+```
+
+### Run Tests
+
+```bash
+# Run all tests
+cd docker/core
+docker-compose run --rm test
+
+# Or build test image manually (from project root)
+docker build -f docker/core/Dockerfile.test -t memvid-test .
+docker run --rm memvid-test
+```
+
+### Build Release
+
+```bash
+# Build release version
+cd docker/core
+docker-compose run --rm build
+
+# Or build manually (from project root)
+docker build -f docker/core/Dockerfile -t memvid-core:latest .
+```
+
+## Docker Images
+
+### 1. Development Image (`Dockerfile.dev`)
+
+Full development environment with all tools:
+
+```bash
+# From project root
+docker build -f docker/core/Dockerfile.dev -t memvid-dev .
+docker run -it --rm -v $(pwd):/app memvid-dev bash
+```
+
+**Features:**
+- Rust toolchain 1.92
+- All build dependencies
+- Cargo watch (optional)
+- Volume mounting for live development
+
+### 2. Test Image (`Dockerfile.test`)
+
+Optimized for running tests:
+
+```bash
+# From project root
+docker build -f docker/core/Dockerfile.test -t memvid-test .
+docker run --rm memvid-test
+```
+
+**Features:**
+- Rust toolchain 1.92
+- Test dependencies
+- Runs tests automatically
+
+### 3. Production Build (`Dockerfile`)
+
+Multi-stage build for optimized production image:
+
+```bash
+# From project root
+docker build -f docker/core/Dockerfile -t memvid-core:latest .
+```
+
+**Features:**
+- Multi-stage build (smaller final image)
+- Only runtime dependencies
+- Optimized release build
+
+## Docker Compose
+
+### Services
+
+- **`dev`** - Development environment with live code mounting
+- **`test`** - Test runner
+- **`build`** - Release builder
+
+### Usage
+
+```bash
+# Start development environment
+docker-compose up -d dev
+
+# Run tests
+docker-compose run --rm test
+
+# Build release
+docker-compose run --rm build
+
+# Stop all services
+docker-compose down
+```
+
+## Examples
+
+### Run Examples in Docker
+
+```bash
+# Development container
+docker-compose exec dev cargo run --example basic_usage
+
+# With features
+docker-compose exec dev cargo run --example pdf_ingestion --features lex,pdf_extract
+```
+
+### Memory-Constrained Testing
+
+Test OOM prevention with memory limits:
+
+```bash
+# Test with memory limit (for OOM testing)
+docker run --rm --memory=150m --memory-swap=150m \
+  -v $(pwd):/app \
+  memvid-test cargo test --features encryption --test encryption_capsule
+```
+
+### Build with Specific Features
+
+```bash
+# Build with all features
+docker-compose exec dev cargo build --release --all-features
+
+# Build with specific features
+docker-compose exec dev cargo build --release --features lex,vec,encryption
+```
+
+## Volume Mounts
+
+The docker-compose setup uses volumes for:
+- **Source code** - Live mounting for development
+- **Cargo cache** - Speeds up builds
+- **Target cache** - Preserves build artifacts
+
+## CI/CD Integration
+
+### GitHub Actions Example
+
+```yaml
+name: Docker Build
+
+on: [push, pull_request]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Build Docker image
+        run: docker build -t memvid-core:test .
+      - name: Run tests
+        run: docker run --rm memvid-core:test cargo test
+```
+
+## Troubleshooting
+
+### Build Fails
+
+```bash
+# Clean build
+docker-compose down -v
+docker-compose build --no-cache
+```
+
+### Permission Issues
+
+```bash
+# Fix permissions
+sudo chown -R $USER:$USER .
+```
+
+### Out of Memory
+
+```bash
+# Increase Docker memory limit in Docker Desktop settings
+# Or use memory limits in docker run:
+docker run --memory=2g --memory-swap=2g ...
+```
+
+## Best Practices
+
+1. **Use docker-compose** for development
+2. **Cache volumes** for faster builds
+3. **Multi-stage builds** for production
+4. **Test in containers** to match CI/CD environment
+5. **Use .dockerignore** to exclude unnecessary files
+
+## Related
+
+- [CLI Docker Setup](../cli/README.md)
+- [Docker Overview](../README.md)
+- [Main Project README](../../README.md)
+- [Contributing Guide](../../CONTRIBUTING.md)

docker/core/Dockerfile

@@ -0,0 +1,53 @@
+# Multi-stage Dockerfile for Memvid Core (Rust Library)
+# Build stage
+FROM rust:1.92-slim-trixie AS builder
+
+# Install build dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    pkg-config \
+    libssl-dev \
+    ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# Copy dependency files first for better caching
+COPY Cargo.toml Cargo.lock ./
+COPY rust-toolchain.toml ./
+
+# Create a dummy src/lib.rs to build dependencies
+RUN mkdir -p src && \
+    echo "// Dummy file for dependency building" > src/lib.rs && \
+    cargo build --release --features lex,pdf_extract && \
+    rm -rf src
+
+# Copy source code
+COPY src ./src
+COPY examples ./examples
+COPY tests ./tests
+
+# Build the library with default features
+RUN cargo build --release --features lex,pdf_extract
+
+# Runtime stage - minimal image for running examples/tests
+FROM debian:trixie-slim AS runtime
+
+# Install runtime dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    ca-certificates \
+    libssl3 \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# Copy the built artifacts from builder
+COPY --from=builder /app/target/release/examples ./examples
+COPY --from=builder /app/target/release/deps ./deps
+COPY --from=builder /app/target/release/libmemvid_core*.rlib ./lib/
+
+# Set environment variables
+ENV RUST_LOG=info
+ENV PATH="/app/examples:${PATH}"
+
+# Default command
+CMD ["/bin/bash"]

docker/core/Dockerfile.dev

@@ -0,0 +1,31 @@
+# Development Dockerfile for Memvid Core
+FROM rust:1.92-slim-trixie AS builder
+
+# Install development dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    pkg-config \
+    libssl-dev \
+    ca-certificates \
+    git \
+    curl \
+    build-essential \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# Copy dependency files
+COPY Cargo.toml Cargo.lock ./
+COPY rust-toolchain.toml ./
+
+# Install cargo-watch for development (optional)
+RUN cargo install cargo-watch --locked || true
+
+# Copy source code
+COPY . .
+
+# Set environment variables
+ENV RUST_LOG=debug
+ENV CARGO_TARGET_DIR=/app/target
+
+# Default command (can be overridden)
+CMD ["/bin/bash"]

docker/core/Dockerfile.test

@@ -0,0 +1,23 @@
+# Test Dockerfile for Memvid Core
+FROM rust:1.92-slim-trixie AS builder
+
+# Install test dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    pkg-config \
+    libssl-dev \
+    ca-certificates \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# Copy everything
+COPY . .
+
+# Set environment variables for testing
+ENV RUST_BACKTRACE=1
+ENV RUST_LOG=info
+ENV CARGO_TARGET_DIR=/app/target
+
+# Run tests with default features
+CMD ["cargo", "test", "--features", "lex,pdf_extract", "--", "--nocapture"]

docker/core/README.md

@@ -0,0 +1,215 @@
+# Docker Setup for Memvid Core
+
+This document describes how to use Docker with the Memvid Core Rust library.
+
+## Quick Start
+
+### Development Environment
+
+```bash
+# From project root or docker/core directory
+cd docker/core
+docker-compose up -d dev
+
+# Enter the container
+docker-compose exec dev bash
+
+# Inside container, you can run:
+cargo build
+cargo test
+cargo run --example basic_usage
+```
+
+### Run Tests
+
+```bash
+# Run all tests
+cd docker/core
+docker-compose run --rm test
+
+# Or build test image manually (from project root)
+docker build -f docker/core/Dockerfile.test -t memvid-test .
+docker run --rm memvid-test
+```
+
+### Build Release
+
+```bash
+# Build release version
+cd docker/core
+docker-compose run --rm build
+
+# Or build manually (from project root)
+docker build -f docker/core/Dockerfile -t memvid-core:latest .
+```
+
+## Docker Images
+
+### 1. Development Image (`Dockerfile.dev`)
+
+Full development environment with all tools:
+
+```bash
+# From project root
+docker build -f docker/core/Dockerfile.dev -t memvid-dev .
+docker run -it --rm -v $(pwd):/app memvid-dev bash
+```
+
+**Features:**
+- Rust toolchain 1.92
+- All build dependencies
+- Cargo watch (optional)
+- Volume mounting for live development
+
+### 2. Test Image (`Dockerfile.test`)
+
+Optimized for running tests:
+
+```bash
+# From project root
+docker build -f docker/core/Dockerfile.test -t memvid-test .
+docker run --rm memvid-test
+```
+
+**Features:**
+- Rust toolchain 1.92
+- Test dependencies
+- Runs tests automatically
+
+### 3. Production Build (`Dockerfile`)
+
+Multi-stage build for optimized production image:
+
+```bash
+# From project root
+docker build -f docker/core/Dockerfile -t memvid-core:latest .
+```
+
+**Features:**
+- Multi-stage build (smaller final image)
+- Only runtime dependencies
+- Optimized release build
+
+## Docker Compose
+
+### Services
+
+- **`dev`** - Development environment with live code mounting
+- **`test`** - Test runner
+- **`build`** - Release builder
+
+### Usage
+
+```bash
+# Start development environment
+docker-compose up -d dev
+
+# Run tests
+docker-compose run --rm test
+
+# Build release
+docker-compose run --rm build
+
+# Stop all services
+docker-compose down
+```
+
+## Examples
+
+### Run Examples in Docker
+
+```bash
+# Development container
+docker-compose exec dev cargo run --example basic_usage
+
+# With features
+docker-compose exec dev cargo run --example pdf_ingestion --features lex,pdf_extract
+```
+
+### Memory-Constrained Testing
+
+Test OOM prevention with memory limits:
+
+```bash
+# Test with memory limit (for OOM testing)
+docker run --rm --memory=150m --memory-swap=150m \
+  -v $(pwd):/app \
+  memvid-test cargo test --features encryption --test encryption_capsule
+```
+
+### Build with Specific Features
+
+```bash
+# Build with all features
+docker-compose exec dev cargo build --release --all-features
+
+# Build with specific features
+docker-compose exec dev cargo build --release --features lex,vec,encryption
+```
+
+## Volume Mounts
+
+The docker-compose setup uses volumes for:
+- **Source code** - Live mounting for development
+- **Cargo cache** - Speeds up builds
+- **Target cache** - Preserves build artifacts
+
+## CI/CD Integration
+
+### GitHub Actions Example
+
+```yaml
+name: Docker Build
+
+on: [push, pull_request]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Build Docker image
+        run: docker build -t memvid-core:test .
+      - name: Run tests
+        run: docker run --rm memvid-core:test cargo test
+```
+
+## Troubleshooting
+
+### Build Fails
+
+```bash
+# Clean build
+docker-compose down -v
+docker-compose build --no-cache
+```
+
+### Permission Issues
+
+```bash
+# Fix permissions
+sudo chown -R $USER:$USER .
+```
+
+### Out of Memory
+
+```bash
+# Increase Docker memory limit in Docker Desktop settings
+# Or use memory limits in docker run:
+docker run --memory=2g --memory-swap=2g ...
+```
+
+## Best Practices
+
+1. **Use docker-compose** for development
+2. **Cache volumes** for faster builds
+3. **Multi-stage builds** for production
+4. **Test in containers** to match CI/CD environment
+5. **Use .dockerignore** to exclude unnecessary files
+
+## Related
+
+- [CLI Docker Setup](../cli/README.md)
+- [Docker Overview](../README.md)
+- [Main Project README](../../README.md)
+- [Contributing Guide](../../CONTRIBUTING.md)

docker/core/docker-compose.yml

@@ -0,0 +1,46 @@
+services:
+  # Development environment
+  dev:
+    build:
+      context: ../..
+      dockerfile: docker/core/Dockerfile.dev
+    volumes:
+      - ../../:/app
+      - cargo-cache:/usr/local/cargo/registry
+      - target-cache:/app/target
+    environment:
+      - RUST_LOG=debug
+      - CARGO_TARGET_DIR=/app/target
+    stdin_open: true
+    tty: true
+    command: /bin/bash
+
+  # Test runner
+  test:
+    build:
+      context: ../..
+      dockerfile: docker/core/Dockerfile.test
+    volumes:
+      - ../../:/app
+      - cargo-cache:/usr/local/cargo/registry
+      - target-cache:/app/target
+    environment:
+      - RUST_BACKTRACE=1
+      - RUST_LOG=info
+    command: cargo test --features lex,pdf_extract -- --nocapture
+
+  # Build release
+  build:
+    build:
+      context: ../..
+      dockerfile: docker/core/Dockerfile
+      target: builder
+    volumes:
+      - ../../:/app
+      - cargo-cache:/usr/local/cargo/registry
+      - target-cache:/app/target
+    command: cargo build --release --features lex,pdf_extract
+
+volumes:
+  cargo-cache:
+  target-cache:

docs/i18n/CONTRIBUTING_TRANSLATIONS.md

@@ -0,0 +1,125 @@
+# Contributing Translations (i18n)
+
+Thank you for helping translate Memvid’s documentation and make the project accessible to a global audience.
+
+This guide explains how to contribute translations for the main `README.md`.
+
+---
+
+## What Can Be Translated
+
+- The main repository `README.md`
+- Translations are stored in `docs/i18n/`
+- Each language has a single translation file
+
+---
+
+## Translation Workflow
+
+### 1. Check Existing Issues
+
+Before starting:
+- Check open issues to see if your language is already in progress
+- If an issue exists, comment on it to indicate you want to work on it
+- Only one contributor should work on a language at a time
+
+---
+
+### 2. Create a Translation Issue (If Needed)
+
+If no issue exists for your language, open a new one using this format:
+
+**Title:**
+```
+README translation: <Language> (<code>)
+```
+
+**Labels to apply:**
+- `i18n`
+- `documentation`
+- `help wanted` or `good first issue`
+
+---
+
+### 3. Create the Translation File
+
+1. Copy the main `README.md` from the repository root
+2. Create a new file in `docs/i18n/` using this format:
+
+   ```
+   README.<language-code>.md
+   ```
+
+   **Examples:**
+- `README.es.md`
+- `README.zh-CN.md`
+- `README.pt-BR.md`
+
+Use standard ISO language codes (include region where applicable).
+
+---
+
+## Translation Guidelines
+
+- **Preserve structure**  
+  Keep headings, section order, and formatting identical to the original README
+
+- **Preserve links and badges**  
+  Do not modify URLs, badges, or shields
+
+- **Preserve code blocks**  
+  Keep code examples, commands, flags, and API names in English
+
+- **Maintain technical accuracy**  
+  Translate naturally, but do not change meaning or behavior
+
+- **Language quality matters**  
+  Native or fluent speakers are strongly preferred
+
+---
+
+## Submitting Your Translation
+
+1. **Create a new branch:**
+   ```bash
+   git checkout -b docs/i18n-<language-code>
+   ```
+
+2. **Add your translation file** under `docs/i18n/`
+
+3. **Commit your changes:**
+   ```bash
+   git commit -m "docs(i18n): add <Language> README translation"
+   ```
+
+4. **Push to your fork** and open a Pull Request
+
+5. **Reference the translation issue** in your PR description
+
+---
+
+## Review Process
+
+- Maintainers may request:
+  - Clarifications
+  - Formatting fixes
+  - Review by another native speaker
+- Once approved, the PR will be merged
+- The corresponding issue will be closed
+
+---
+
+## Keeping Translations Up to Date
+
+When the main `README.md` changes significantly:
+
+- Existing translations may need updates
+- Contributors are encouraged to help keep translations in sync
+
+---
+
+## Getting Help
+
+- Open an issue if you have questions about translations
+- Use GitHub Discussions for coordination
+- Contact: [contact@memvid.com](mailto:contact@memvid.com)

docs/i18n/README.ar.md

@@ -0,0 +1,160 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+# المساهمة في Memvid (الترجمة العربية)
+
+<p align="center">
+<strong>Memvid هي طبقة ذاكرة مكونة من ملف واحد لوكلاء الذكاء الاصطناعي (AI Agents)، توفر استرجاعاً فورياً وذاكرة طويلة المدى.</strong>
+
+ذاكرة دائمة، مؤرشفة، وقابلة للنقل، دون الحاجة إلى قواعد بيانات.
+
+</p>
+
+<h2 align="center">⭐️ اترك نجمة (STAR) لدعم المشروع ⭐️</h2>
+
+---
+
+## ما هو Memvid؟
+
+Memvid هو نظام ذاكرة محمول للذكاء الاصطناعي يقوم بتغليف بياناتك، والمتجهات (Embeddings)، وهيكل البحث، والبيانات الوصفية في **ملف واحد فقط**.
+
+بدلاً من تشغيل خطوط أنابيب RAG معقدة أو قواعد بيانات متجهة تعتمد على الخادم، يتيح Memvid استرجاعاً سريعاً للبيانات مباشرة من الملف.
+
+النتيجة هي طبقة ذاكرة مستقلة عن النموذج (Model-agnostic) ولا تحتاج إلى بنية تحتية، مما يمنح وكلاء الذكاء الاصطناعي ذاكرة دائمة وطويلة المدى يمكنهم حملها في أي مكان.
+
+---
+
+## لماذا "إطارات الفيديو" (Video Frames)؟
+
+يستلهم Memvid فكرته من ترميز الفيديو، ليس لتخزين الفيديو، بل **لتنظيم ذاكرة الذكاء الاصطناعي كمتسلسلة فائقة الكفاءة من "الإطارات الذكية" (Smart Frames) التي تُضاف باستمرار.**
+
+"الإطار الذكي" هو وحدة غير قابلة للتغيير تخزن المحتوى مع الطوابع الزمنية، والتحقق من البيانات (Checksums)، والبيانات الوصفية الأساسية. يتم تجميع هذه الإطارات بطريقة تسمح بضغط البيانات وفهرستها والقراءة المتوازية بكفاءة عالية.
+
+يسمح هذا التصميم بـ:
+
+-   **إضافة البيانات فقط:** الكتابة دون تعديل أو إفساد البيانات الموجودة.
+-   **الاستعلام عبر الزمن:** البحث في حالات الذاكرة السابقة.
+-   **جدول زمني للمعرفة:** فحص كيفية تطور المعرفة بمرور الوقت.
+-   **سلامة البيانات:** ضمان عدم فقدان البيانات عند التعطل بفضل الإطارات الثابتة.
+
+---
+
+## المفاهيم الأساسية
+
+-   **محرك الذاكرة الحية:** إضافة وتطوير الذاكرة باستمرار عبر الجلسات.
+-   **كبسولة السياق (`.mv2`):** كبسولات ذاكرة ذاتية الاحتواء وقابلة للمشاركة مع قواعد وصلاحية محددة.
+-   **تصحيح السفر عبر الزمن:** إرجاع أو إعادة تشغيل أو تفريع أي حالة من حالات الذاكرة.
+-   **الاستدعاء الذكي:** وصول محلي للذاكرة في أقل من 5 ملي ثانية مع ذاكرة تخزين مؤقت تنبؤية.
+-   **ذكاء الترميز:** يختار ويحدث تقنيات الضغط تلقائياً بمرور الوقت.
+
+---
+
+## حالات الاستخدام
+
+نظرًا لأن Memvid يعمل دون اتصال بالإنترنت ومستقل عن النماذج، فإنه يُستخدم في:
+
+-   وكلاء الذكاء الاصطناعي طويلي الأمد.
+-   قواعد المعرفة للمؤسسات.
+-   أنظمة الذكاء الاصطناعي التي تعمل "بدون إنترنت أولاً".
+-   فهم الأكواد البرمجية (Codebases).
+-   المساعدين الشخصيين والأنظمة الطبية والقانونية والمالية.
+
+---
+
+## أدوات المطورين (SDKs)
+
+| الحزمة                        | طريقة التثبيت               |
+| ----------------------------- | --------------------------- |
+| **واجهة السطر البرمجي (CLI)** | `npm install -g memvid-cli` |
+| **Node.js SDK**               | `npm install @memvid/sdk`   |
+| **Python SDK**                | `pip install memvid-sdk`    |
+| **Rust**                      | `cargo add memvid-core`     |
+
+---
+
+## هيكل الملف
+
+كل شيء يعيش داخل ملف واحد بصيغة `.mv2`:
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  العنوان: النسخة والقدرة الاستيعابية
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  سجل العمليات للتعافي من الأعطال
+├────────────────────────────┤
+│ Data Segments              │  أجزاء البيانات: الإطارات المضغوطة
+├────────────────────────────┤
+│ Lex Index                  │  الفهرس اللغوي: البحث النصي الكامل
+├────────────────────────────┤
+│ Vec Index                  │  الفهرس المتجهي: البحث بالمتجهات (HNSW)
+├────────────────────────────┤
+│ Time Index                 │  الفهرس الزمني: الترتيب الزمني
+├────────────────────────────┤
+│ TOC (Footer)               │  جدول المحتويات: مواقع الأجزاء
+└────────────────────────────┘
+
+```
+
+---
+
+## الدعم
+
+هل لديك أسئلة؟
+البريد الإلكتروني: contact@memvid.com
+
+**لا تنسَ ترك ⭐ لدعم المشروع!**
+
+---
+
+## الترخيص
+
+رخصة Apache 2.0 — راجع ملف [LICENSE](../../LICENSE) لمزيد من التفاصيل.

docs/i18n/README.bn.md

@@ -0,0 +1,379 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+
+# মেমভিড-এ অবদান (বাংলা অনুবাদ)
+
+<p align="center">
+  <strong>মেমভিড হল এআই এজেন্টদের জন্য একটি একক-ফাইল মেমরি স্তর যার তাৎক্ষণিক পুনরুদ্ধার এবং দীর্ঘমেয়াদী মেমরি রয়েছে।</strong><br/>
+  ডাটাবেস ছাড়াই স্থায়ী, সংস্করণযুক্ত এবং পোর্টেবল মেমরি।
+</p>
+
+<p align="center">
+  <a href="https://www.memvid.com">ওয়েবসাইট</a>
+  ·
+  <a href="https://sandbox.memvid.com">স্যান্ডবক্সি চেষ্টা করে দেখুন</a>
+  ·
+  <a href="https://docs.memvid.com">ডক্স</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">আলোচনা</a>
+</p>
+
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/</a>
+</p>
+
+<h2 align="center">⭐️ প্রকল্পটি সমর্থন করার জন্য একটি তারকা দিন। ⭐️</h2>
+</p>
+
+## মেমভিড কী?
+
+মেমভিড হল একটি পোর্টেবল এআই মেমরি সিস্টেম যা আপনার ডেটা, এম্বেডিং, অনুসন্ধান কাঠামো এবং মেটাডেটা একটি একক ফাইলে প্যাক করে।
+
+জটিল RAG পাইপলাইন বা সার্ভার-ভিত্তিক ভেক্টর ডাটাবেস চালানোর পরিবর্তে, Memvid ফাইল থেকে সরাসরি দ্রুত ডেটা পুনরুদ্ধারে সহায়তা করে।
+
+ফলাফল হল একটি মডেল-অজ্ঞেয়বাদী, অবকাঠামো-মুক্ত মেমোরি স্তর যা AI এজেন্টদের স্থায়ী, দীর্ঘমেয়াদী মেমোরি দেয় যা তারা যেকোনো জায়গায় নিতে পারে।
+---
+
+## ভিডিও ফ্রেম কেন?
+
+মেমভিড ভিডিও এনকোডিং থেকে অনুপ্রেরণা নেয়, ভিডিও সংরক্ষণের জন্য নয়, বরং এআই মেমোরিকে অ্যাপেন্ড-ওনলি, স্মার্ট ফ্রেমের অতি-দক্ষ সিকোয়েন্স হিসেবে সংগঠিত করার জন্য।
+
+একটি স্মার্ট ফ্রেম হল একটি অ-পরিবর্তনযোগ্য ইউনিট যা টাইমস্ট্যাম্প, চেকসাম এবং মৌলিক মেটাডেটা সহ বিষয়বস্তু সংরক্ষণ করে।
+ফ্রেমগুলিকে এমনভাবে গোষ্ঠীভুক্ত করা হয় যা দক্ষ কম্প্রেশন, ইনডেক্সিং এবং সমান্তরাল পঠনের সুযোগ করে দেয়।
+
+এই ফ্রেম-ভিত্তিক নকশাটি সক্ষম করে:
+
+- বিদ্যমান ডেটা পরিবর্তন বা দূষিত না করে কেবল ডেটা যোগ করা
+- পূর্ববর্তী মেমরি অবস্থা অনুসন্ধান করা
+- জ্ঞান কীভাবে বিকশিত হয় তার টাইমলাইন-স্টাইল তদন্ত
+- প্রতিশ্রুতিবদ্ধ, অপরিবর্তনীয় ফ্রেমের মাধ্যমে ক্র্যাশ সুরক্ষা
+- ভিডিও এনকোডিং থেকে গৃহীত কৌশল ব্যবহার করে দক্ষ কম্প্রেশন।
+
+ফলাফল হল একটি একক ফাইল যা AI সিস্টেমের জন্য একটি রিওয়াইন্ডেবল মেমরি টাইমলাইন হিসাবে কাজ করে।
+
+---
+
+## মূল ধারণা
+
+- **লিভিং মেমোরি ইঞ্জিন**
+
+একটি সেশনের সময় স্থায়ী মেমোরি যোগ করুন, শাখা করুন এবং বিকশিত করুন।
+
+- **ক্যাপসুল রেফারেন্স (`.mv2`)**
+
+নিয়ম এবং মেয়াদোত্তীর্ণতা সহ স্বয়ংসম্পূর্ণ, শেয়ারযোগ্য মেমোরি ক্যাপসুল।
+
+- **টাইম-ট্রাভেল ডিবাগিং**
+
+যেকোন মেমোরি স্টেট রিওয়াইন্ড, রিপ্লে বা শাখা করুন।
+
+- **স্মার্ট রিকল**
+
+প্রেডিক্টিভ ক্যাশিং সহ সাব-5ms স্থানীয় মেমোরি অ্যাক্সেস।
+
+- **কোডেক ইন্টেলিজেন্স**
+
+সময়ের সাথে সাথে স্বয়ংক্রিয়ভাবে কম্প্রেশন নির্বাচন এবং আপগ্রেড করে।
+
+---
+
+## ব্যবহারের ক্ষেত্রে
+
+মেমভিড হল একটি পোর্টেবল, সার্ভারলেস মেমরি স্তর যা এআই এজেন্টদের স্থায়ী মেমরি এবং দ্রুত রিকল দেয়। যেহেতু এটি মডেল-অ্যাগনস্টিক, মাল্টি-মডেল এবং সম্পূর্ণ অফলাইনে কাজ করে, তাই ডেভেলপাররা বিভিন্ন বাস্তব-বিশ্ব অ্যাপ্লিকেশনে মেমভিড ব্যবহার করছে।
+
+- দীর্ঘমেয়াদী এআই এজেন্ট
+- এন্টারপ্রাইজ জ্ঞান ভিত্তি
+- অফলাইন-প্রথম এআই সিস্টেম
+- কোডবেস বোঝা
+- গ্রাহক সহায়তা এজেন্ট
+- ওয়ার্কফ্লো অটোমেশন
+- বিক্রয় এবং বিপণন সহ-পাইলট
+- ব্যক্তিগত জ্ঞান সহকারী
+- চিকিৎসা, আইনি এবং আর্থিক এজেন্ট
+- নিরীক্ষণযোগ্য এবং ডিবাগযোগ্য এআই কর্মপ্রবাহ
+- ​​কাস্টম অ্যাপ্লিকেশন
+
+---
+
+## SDKs & CLI
+
+আপনার পছন্দের ভাষায় Memvid ব্যবহার করুন:
+
+| Package         | Install                     | Links                                                                                                               |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## Installation (Rust)
+
+### আবশ্যকতা
+
+-   **Rust 1.85.0+** — Install from [rustup.rs](https://rustup.rs)
+
+### আপনার প্রকল্পে যোগ করুন
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Feature Flags
+
+| Feature             | Description                                    |
+| ------------------- | ---------------------------------------------- |
+| `lex`               | Full-text search with BM25 ranking (Tantivy)   |
+| `pdf_extract`       | Pure Rust PDF text extraction                  |
+| `vec`               | Vector similarity search (HNSW + ONNX)         |
+| `clip`              | CLIP visual embeddings for image search        |
+| `whisper`           | Audio transcription with Whisper               |
+| `temporal_track`    | Natural language date parsing ("last Tuesday") |
+| `parallel_segments` | Multi-threaded ingestion                       |
+| `encryption`        | Password-based encryption capsules (.mv2e)     |
+
+Enable features as needed:
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## দ্রুত শুরু
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Create a new memory file
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Add documents with metadata
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // Search
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## তৈরি করুন
+
+রিপোজিটরিটি ক্লোন করুন:
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+ডিবাগ মোডে তৈরি করুন:
+
+```bash
+cargo build
+```
+
+রিলিজ মোডে তৈরি করুন (অপ্টিমাইজ করা):
+
+```bash
+cargo build --release
+```
+
+স্বতন্ত্র বৈশিষ্ট্য সহ তৈরি করুন:
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+## পরীক্ষাগুলি চালান
+
+সকল পরীক্ষা চালান:
+
+```bash
+cargo test
+```
+
+নিম্নলিখিত আউটপুট দিয়ে পরীক্ষাটি চালান:
+
+```bash
+cargo test -- --nocapture
+```
+
+একটি নির্দিষ্ট পরীক্ষা চালান:
+```bash
+cargo test test_name
+```
+শুধুমাত্র ইন্টিগ্রেশন পরীক্ষা চালান:
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## উদাহরণ
+
+The `examples/` কার্যকরী ডিরেক্টরিগুলির উদাহরণ হল:
+
+### মৌলিক ব্যবহার
+
+এটি তৈরি, পুট, অনুসন্ধান এবং টাইমলাইন ক্রিয়াকলাপগুলি দেখায়:
+
+```bash
+cargo run --example basic_usage
+```
+
+### পিডিএফ ইনজেকশন
+
+পিডিএফ ডকুমেন্টগুলি গ্রহণ করুন এবং অনুসন্ধান করুন ("Attention Is All You Need" পেপার ব্যবহার করে):
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### CLIP ভিজ্যুয়াল সার্চ
+
+CLIP এম্বেডিং ব্যবহার করে ছবি সার্চ (`ক্লিপ` বৈশিষ্ট্য প্রয়োজন):
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### হুইস্পার ট্রান্সক্রিপশন
+
+অডিও ট্রান্সক্রিপশন (`হুইস্পার` বৈশিষ্ট্য প্রয়োজন):
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## ফাইল ফরম্যাট
+
+সবকিছু একটি একক `.mv2` ফাইলের মধ্যে রয়েছে:
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  Magic, version, capacity
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  Crash recovery
+├────────────────────────────┤
+│ Data Segments              │  Compressed frames
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy full-text
+├────────────────────────────┤
+│ Vec Index                  │  HNSW vectors
+├────────────────────────────┤
+│ Time Index                 │  Chronological ordering
+├────────────────────────────┤
+│ TOC (Footer)               │  Segment offsets
+└────────────────────────────┘
+```
+
+কোন `.wal`, `.lock`, `.shm`, অথবা সাইডকার ফাইল নেই। কখনও না।
+
+সম্পূর্ণ ফাইল ফর্ম্যাট স্পেসিফিকেশনের জন্য [MV2_SPEC.md](MV2_SPEC.md) দেখুন।
+
+---
+
+## সহায়তা
+
+আপনার কি কোন প্রশ্ন বা প্রতিক্রিয়া আছে?
+
+ইমেল: contact@memvid.com
+
+**সমর্থন দেখানোর জন্য ⭐ দিন**
+
+---
+
+## লাইসেন্স
+
+Apache License 2.0 — আরও তথ্যের জন্য [LICENSE](LICENSE) ফাইলটি দেখুন।

docs/i18n/README.cs.md

@@ -0,0 +1,424 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+  <strong>Memvid je jednosouborová paměťová vrstva pro AI agenty s okamžitým vyhledáváním a dlouhodobou pamětí.</strong><br/>
+  Trvalá, verzovaná a přenosná paměť, bez databází.
+</p>
+
+<h2 align="center">⭐️ Zanechte hvězdičku na podporu projektu ⭐️</h2>
+</p>
+
+## Co je Memvid?
+
+Memvid je systém pro tvorbu AI pamětí, který balí vaše data, embeddingy, strukturu vyhledávání a metadata do jediného souboru.
+
+Místo spouštění složitých RAG řešení nebo serverových vektorových databází umožňuje Memvid rychlé vyhledávání přímo ze souboru.
+
+Výsledkem je modelově nezávislá paměťová vrstva bez infrastruktury, která poskytuje agentům AI trvalou, dlouhodobou paměť, kterou lze přenášet kamkoli.
+
+---
+
+## Co jsou inteligentní rámce?
+
+Memvid čerpá inspiraci z enkódování videa, nikoli za účelem ukládání videa, ale za účelem **organizace paměti AI jako ultraefektivní sekvence inteligentních rámců, do kterých lze data  pouze přidávat.**
+
+Inteligentní rámec je neměnná jednotka, která ukládá obsah spolu s časovými značkami, kontrolními součty a základními metadaty.
+Rámce jsou seskupeny tak, aby umožňovaly efektivní kompresi, indexování a paralelní čtení.
+
+Tento design založený na rámcích umožňuje:
+
+-  Pouze zápisy bez úpravy nebo poškození existujících dat
+-  Dotazy na minulé stavy paměti
+-  Kontrolu vývoje znalostí ve stylu časové osy
+-  Bezpečnost proti selhání díky závazným, neměnným rámcům
+-  Efektivní kompresi pomocí technik převzatých z kódování videa
+
+Výsledkem je jediný soubor, který se chová jako časová osa paměti pro systémy AI, ve které lze snadno hledat.
+
+---
+
+## Základní koncepty
+
+-   **Living Memory Engine**
+    Kontinuální přidávání, rozvětvování a vývoj paměti, napříč relacemi.
+
+-   **Capsule Context (`.mv2`)**
+    Samostatné, sdílené paměťové kapsle s pravidly a dobou platnosti.
+
+-   **Time-Travel Debugging**
+    Převíjení, přehrávání nebo rozvětvování libovolného stavu paměti.
+
+-   **Smart Recall**
+    Přístup k lokální paměti za méně než 5 ms s prediktivním ukládáním do mezipaměti.
+
+-   **Codec Intelligence**
+    Automaticky vybírá a vylepšuje kompresi v průběhu času.
+
+---
+
+## Případy použití
+
+Memvid je přenosná paměťová vrstva bez serveru, která poskytuje agentům AI trvalou paměť a rychlé vyvolání. Protože je modelově nezávislá, multimodální a funguje zcela offline, vývojáři používají Memvid v široké škále reálných aplikací.
+
+- Dlouhodobě běžící AI agenti
+- Firemní znalostní báze
+- Offline-First AI systémy
+- Porozumění kódu
+- Agenti zákaznické podpory
+- Automatizace pracovních postupů
+- Asistenti prodeje a marketingu
+- Osobní znalostní asistenti
+- Lékařští, právní a finanční agenti
+- Auditovatelné a laditelné AI pracovní postupy
+- Vlastní aplikace
+
+---
+
+## SDK a CLI
+
+Používejte Memvid ve svém preferovaném jazyce:
+
+| Balíček         | Instalace                   | Odkazy                                                                                                              |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## Instalace (Rust)
+
+### Požadavky
+
+-   **Rust 1.85.0+** — Instalace z [rustup.rs](https://rustup.rs)
+
+### Přidejte do svého projektu
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Funkční příznaky
+
+| Funkce              | Popis                                                      |
+| ------------------- | ---------------------------------------------------------- |
+| `lex`               | Fulltextové vyhledávání s hodnocením BM25 (Tantivy)        |
+| `pdf_extract`       | Čistá extrakce textu z PDF v Rustu                         |
+| `vec`               | Vektorové vyhledávání podobnosti (HNSW + lokální vkládání textu přes ONNX) |
+| `clip`              | Vizuální vkládání CLIP pro vyhledávání obrázků             |
+| `whisper`           | Přepis zvuku pomocí Whisper                                |
+| `temporal_track`    | Analýza datumu v přirozeném jazyce ("minulé úterý")        |
+| `parallel_segments` | Vícevláknové načítání                                      |
+| `encryption`        | Kapsle šifrované pomocí hesla (.mv2e)                 |
+
+Povolte funkce podle potřeby:
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## Rychlý start
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Vytvoř nový paměťový soubor
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Přidej dokumenty s metadaty
+    let opts = PutOptions::builder()
+        .title("Zápis jednání")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 plánované diskuze...", opts)?;
+    mem.commit()?;
+
+    // Vyhledávání
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## Sestavení
+
+Klonujte repozitář:
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+Sestavení v režimu developkment:
+
+```bash
+cargo build
+```
+
+Sestavení v režimu production (optimalizované):
+
+```bash
+cargo build --release
+```
+
+Sestavení s konkrétními funkcemi:
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## Spuštění testů
+
+Spuštění všech testů:
+
+```bash
+cargo test
+```
+
+Spuštění testů s výstupem:
+
+```bash
+cargo test -- --nocapture
+```
+
+Spuštění konkrétního testu:
+
+```bash
+cargo test test_name
+```
+
+Spuštění pouze integračních testů:
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## Příklady
+
+Adresář `examples/` obsahuje funkční příklady:
+
+### Základní použití
+
+Demonstruje operace vytváření, vkládání, vyhledávání a práci s časovou osou:
+
+```bash
+cargo run --example basic_usage
+```
+
+### Načítání PDF
+
+Načítání a vyhledávání v dokumentech PDF (demo používá dokument "Attention Is All You Need"):
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### Vizuální vyhledávání CLIP
+
+Vyhledávání obrázků pomocí vložení CLIP (vyžaduje funkci `clip`):
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Přepis Whisper
+
+Přepis zvuku (vyžaduje funkci `whisper`):
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## Modely vkládání textu
+
+Funkce `vec` zahrnuje podporu lokálního vkládání textu pomocí modelů ONNX. Před použitím lokálního vkládání textu je nutné ručně stáhnout soubory modelů.
+
+### Rychlý start: BGE-small (doporučeno)
+
+Stáhněte si výchozí model BGE-small (384 dimenzí, rychlý a efektivní):
+
+```bash
+mkdir -p ~/.cache/memvid/text-models
+
+# Download ONNX model
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5.onnx
+
+# Download tokenizer
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5_tokenizer.json
+```
+
+### Dostupné modely
+
+| Model                   | Rozměry    | Velikost | Nejvhodnější pro     |
+| ----------------------- | ---------- | ------- | --------------------- |
+| `bge-small-en-v1.5`     | 384        | ~120 MB | Výchozí, rychlý       |
+| `bge-base-en-v1.5`      | 768        | ~420 MB | Lepší kvalita         |
+| `nomic-embed-text-v1.5` | 768        | ~530 MB | Univerzální úkoly     |
+| `gte-large`             | 1024       | ~1,3 GB | Nejvyšší kvalita      |
+
+### Další modely
+
+**BGE-base** (768 dimensions):
+```bash
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5.onnx
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5_tokenizer.json
+```
+
+**Nomic** (768 dimensions):
+```bash
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5.onnx
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5_tokenizer.json
+```
+
+**GTE-large** (1024 dimensions):
+```bash
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/gte-large.onnx
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/gte-large_tokenizer.json
+```
+
+### Použití v kódu
+
+```rust
+use memvid_core::text_embed::{LocalTextEmbedder, TextEmbedConfig};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// Použít výchozí model (BGE-small)
+let config = TextEmbedConfig::default();
+let embedder = LocalTextEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 384);
+
+// Použijte jiný model
+let config = TextEmbedConfig::bge_base();
+let embedder = LocalTextEmbedder::new(config)?;
+```
+
+Kompletní příklad s výpočtem podobnosti a hodnocením vyhledávání najdete v souboru `examples/text_embedding.rs`.
+
+---
+
+## Formát souboru
+
+Vše je uloženo v jediném souboru `.mv2`:
+
+```
+┌────────────────────────────┐
+│ Záhlaví (4 KB)             │  Magie, verze, kapacita
+├────────────────────────────┤
+│ Embedded WAL (1-64 MB)     │  Obnova po selhání
+├────────────────────────────┤
+│ Datové segmenty            │  Komprimované rámce
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy full-text
+├────────────────────────────┤
+│ Vec Index                  │  Vektory HNSW
+├────────────────────────────┤
+│ Time Index                 │  Chronologické řazení
+├────────────────────────────┤
+│ TOC (zápatí)               │  Segmentové posuny
+└────────────────────────────┘
+```
+
+Žádné další soubory, jako je `.wal`, `.lock`, `.shm` nejsou potřeba. Nikdy.
+
+Kompletní specifikace formátu souboru najdete v [MV2_SPEC.md](MV2_SPEC.md).
+
+---
+
+## Podpora
+
+Máte dotazy nebo připomínky?
+E-mail: contact@memvid.com
+
+**Dejte ⭐ a projevte svou podporu**
+
+---
+
+## Licence
+
+Apache License 2.0 — podrobnosti najdete v souboru [LICENSE](LICENSE).

docs/i18n/README.es.md

@@ -0,0 +1,500 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+  <strong>Memvid es una capa de memoria de un solo archivo para agentes de IA, con recuperación instantánea y memoria a largo plazo.</strong><br/>
+  Memoria persistente, versionada y portable, sin bases de datos.
+</p>
+
+<h2 align="center">⭐️ Deja una STAR para apoyar el proyecto ⭐️</h2>
+</p>
+
+## Lo más destacado de los benchmarks
+
+**🚀 Mayor precisión que cualquier otro sistema de memoria:** +35% SOTA en LoCoMo, con recall y razonamiento conversacional de largo horizonte de primer nivel.
+
+**🧠 Mejor razonamiento multi-hop y temporal:** +76% en multi-hop y +56% en temporal frente al promedio de la industria.
+
+**⚡ Latencia ultra baja a escala:** 0.025 ms P50 y 0.075 ms P99, con 1,372× más throughput que los enfoques estándar.
+
+**🔬 Benchmarks totalmente reproducibles:** LoCoMo (10 conversaciones de ~26K tokens), evaluación open source y LLM-as-Judge.
+
+---
+
+## ¿Qué es Memvid?
+
+Memvid es un sistema de memoria portable para IA que empaqueta tus datos, embeddings, estructura de búsqueda y metadatos en un solo archivo.
+
+En lugar de ejecutar pipelines RAG complejos o bases de datos vectoriales basadas en servidor, Memvid permite una recuperación rápida directamente desde el archivo.
+
+El resultado es una capa de memoria agnóstica al modelo, sin infraestructura, que da a los agentes de IA una memoria persistente y a largo plazo que pueden llevar a cualquier parte.
+
+---
+
+## ¿Por qué fotogramas de vídeo?
+
+Memvid se inspira en la codificación de vídeo, no para almacenar vídeo, sino para **organizar la memoria de IA como una secuencia de Smart Frames ultrarrápida y append-only.**
+
+Un Smart Frame es una unidad inmutable que almacena contenido junto con marcas de tiempo (timestamps), checksums y metadatos básicos.
+Los frames se agrupan de una forma que permite una compresión, indexación y lecturas paralelas eficientes.
+
+Este diseño basado en frames permite:
+
+-   Escrituras append-only sin modificar ni corromper los datos existentes
+-   Consultas sobre estados pasados de la memoria
+-   Inspección estilo línea temporal (timeline) de cómo evoluciona el conocimiento
+-   Seguridad ante fallos (crash safety) mediante frames confirmados e inmutables
+-   Compresión eficiente usando técnicas adaptadas de la codificación de vídeo
+
+El resultado es un único archivo que se comporta como una línea temporal de memoria “rebobinable” para sistemas de IA.
+
+---
+
+## Conceptos principales
+
+-   **Living Memory Engine**
+    Añade, ramifica (branch) y evoluciona la memoria de forma continua entre sesiones.
+
+-   **Capsule Context (`.mv2`)**
+    Cápsulas de memoria autocontenidas y compartibles, con reglas y caducidad.
+
+-   **Time-Travel Debugging**
+    Rebobina, reproduce (replay) o ramifica cualquier estado de memoria.
+
+-   **Smart Recall**
+    Acceso local a memoria en menos de 5ms con caché predictiva.
+
+-   **Codec Intelligence**
+    Selecciona y actualiza la compresión automáticamente con el tiempo.
+
+---
+
+## Casos de uso
+
+Memvid es una capa de memoria portable y serverless que da a los agentes de IA memoria persistente y recuerdo rápido. Como es agnóstica al modelo, multi-modal y funciona totalmente offline, los desarrolladores están usando Memvid en una amplia gama de aplicaciones reales.
+
+-   Agentes de IA de larga duración
+-   Bases de conocimiento empresariales
+-   Sistemas de IA offline-first
+-   Comprensión de codebases
+-   Agentes de soporte al cliente
+-   Automatización de flujos de trabajo
+-   Copilotos de ventas y marketing
+-   Asistentes de conocimiento personal
+-   Agentes médicos, legales y financieros
+-   Flujos de trabajo de IA auditables y depurables
+-   Aplicaciones personalizadas
+
+---
+
+## SDKs & CLI
+
+Usa Memvid en tu lenguaje preferido:
+
+| Package         | Install                     | Links                                                                                                               |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## Instalación (Rust)
+
+### Requisitos
+
+-   **Rust 1.85.0+** — Instálalo desde [rustup.rs](https://rustup.rs)
+
+### Añadir a tu proyecto
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Feature Flags
+
+| Feature             | Descripción                                                        |
+| ------------------- | ------------------------------------------------------------------ |
+| `lex`               | Búsqueda full-text con ranking BM25 (Tantivy)                      |
+| `pdf_extract`       | Extracción de texto PDF 100% en Rust                               |
+| `vec`               | Búsqueda por similitud vectorial (HNSW + embeddings locales vía ONNX) |
+| `clip`              | Embeddings visuales CLIP para búsqueda de imágenes                 |
+| `whisper`           | Transcripción de audio con Whisper                                 |
+| `api_embed`         | Embeddings en la nube mediante API (OpenAI)                        |
+| `temporal_track`    | Interpretación de fechas en lenguaje natural ("el martes pasado")  |
+| `parallel_segments` | Ingesta multi-hilo                                                 |
+| `encryption`        | Cápsulas cifradas con contraseña (.mv2e)                           |
+| `symspell_cleanup`  | Reparación robusta de texto PDF (corrige "emp lo yee" -> "employee") |
+
+Activa las features según lo necesites:
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## Inicio rápido
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Create a new memory file
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Add documents with metadata
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // Search
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## Build
+
+Clona el repositorio:
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+Compila en modo debug:
+
+```bash
+cargo build
+```
+
+Compila en modo release (optimizado):
+
+```bash
+cargo build --release
+```
+
+Compila con features específicas:
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## Ejecutar tests
+
+Ejecuta todos los tests:
+
+```bash
+cargo test
+```
+
+Ejecuta los tests con salida:
+
+```bash
+cargo test -- --nocapture
+```
+
+Ejecuta un test específico:
+
+```bash
+cargo test test_name
+```
+
+Ejecuta solo tests de integración:
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## Ejemplos
+
+El directorio `examples/` contiene ejemplos funcionales:
+
+### Uso básico
+
+Demuestra operaciones de create, put, search y timeline:
+
+```bash
+cargo run --example basic_usage
+```
+
+### Ingesta de PDF
+
+Ingiere y busca documentos PDF (usa el paper “Attention Is All You Need”):
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### Búsqueda visual con CLIP
+
+Búsqueda de imágenes usando embeddings de CLIP (requiere la feature `clip`):
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Transcripción con Whisper
+
+Transcripción de audio (requiere la feature `whisper`):
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## Modelos de embeddings de texto
+
+La feature `vec` incluye soporte para embeddings de texto locales usando modelos ONNX. Antes de usar embeddings locales, necesitas descargar manualmente los archivos del modelo.
+
+### Inicio rápido: BGE-small (recomendado)
+
+Descarga el modelo BGE-small por defecto (384 dimensiones, rápido y eficiente):
+
+```bash
+mkdir -p ~/.cache/memvid/text-models
+
+# Descargar modelo ONNX
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5.onnx
+
+# Descargar tokenizer
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5_tokenizer.json
+```
+
+### Modelos disponibles
+
+| Modelo                  | Dimensiones | Tamaño | Mejor para            |
+| ----------------------- | ----------- | ------ | --------------------- |
+| `bge-small-en-v1.5`     | 384         | ~120MB | Opción por defecto, rápido |
+| `bge-base-en-v1.5`      | 768         | ~420MB | Mejor calidad         |
+| `nomic-embed-text-v1.5` | 768         | ~530MB | Tareas versátiles     |
+| `gte-large`             | 1024        | ~1.3GB | Máxima calidad        |
+
+### Otros modelos
+
+**BGE-base** (768 dimensiones):
+```bash
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5.onnx
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5_tokenizer.json
+```
+
+**Nomic** (768 dimensiones):
+```bash
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5.onnx
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5_tokenizer.json
+```
+
+**GTE-large** (1024 dimensiones):
+```bash
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/gte-large.onnx
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/gte-large_tokenizer.json
+```
+
+### Uso en código
+
+```rust
+use memvid_core::text_embed::{LocalTextEmbedder, TextEmbedConfig};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// Usar el modelo por defecto (BGE-small)
+let config = TextEmbedConfig::default();
+let embedder = LocalTextEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 384);
+
+// Usar un modelo distinto
+let config = TextEmbedConfig::bge_base();
+let embedder = LocalTextEmbedder::new(config)?;
+```
+
+Consulta `examples/text_embedding.rs` para ver un ejemplo completo con cálculo de similitud y ranking de búsqueda.
+
+### Consistencia del modelo
+
+Para evitar mezclar modelos por accidente, por ejemplo consultar un índice BGE-small con embeddings de OpenAI, puedes asociar explícitamente tu instancia de Memvid a un nombre de modelo:
+
+```rust
+// Vincula el índice a un modelo concreto.
+// Si el índice ya fue creado con otro modelo, devolverá un error.
+mem.set_vec_model("bge-small-en-v1.5")?;
+```
+
+Esta vinculación es persistente. Una vez definida, cualquier intento futuro de usar otro nombre de modelo fallará de inmediato con un error `ModelMismatch`.
+
+---
+
+## Embeddings por API (OpenAI)
+
+La feature `api_embed` habilita la generación de embeddings en la nube usando la API de OpenAI.
+
+### Configuración
+
+Define tu clave de API de OpenAI:
+
+```bash
+export OPENAI_API_KEY="sk-..."
+```
+
+### Uso
+
+```rust
+use memvid_core::api_embed::{OpenAIConfig, OpenAIEmbedder};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// Usar el modelo por defecto (text-embedding-3-small)
+let config = OpenAIConfig::default();
+let embedder = OpenAIEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 1536);
+
+// Usar un modelo de mayor calidad
+let config = OpenAIConfig::large();  // text-embedding-3-large (3072 dims)
+let embedder = OpenAIEmbedder::new(config)?;
+```
+
+### Modelos disponibles
+
+| Modelo                   | Dimensiones | Mejor para                       |
+| ------------------------ | ----------- | -------------------------------- |
+| `text-embedding-3-small` | 1536        | Por defecto, más rápido y económico |
+| `text-embedding-3-large` | 3072        | Máxima calidad                   |
+| `text-embedding-ada-002` | 1536        | Modelo heredado                  |
+
+Consulta `examples/openai_embedding.rs` para ver un ejemplo completo.
+
+---
+
+## Formato de archivo
+
+Todo vive en un único archivo `.mv2`:
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  Magic, version, capacity
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  Crash recovery
+├────────────────────────────┤
+│ Data Segments              │  Compressed frames
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy full-text
+├────────────────────────────┤
+│ Vec Index                  │  HNSW vectors
+├────────────────────────────┤
+│ Time Index                 │  Chronological ordering
+├────────────────────────────┤
+│ TOC (Footer)               │  Segment offsets
+└────────────────────────────┘
+```
+
+Sin archivos `.wal`, `.lock`, `.shm` ni sidecars. Nunca.
+
+Consulta [MV2_SPEC.md](MV2_SPEC.md) para la especificación completa del formato de archivo.
+
+---
+
+## Soporte
+
+¿Tienes preguntas o feedback?
+Email: contact@memvid.com
+
+**Deja una ⭐ para mostrar apoyo**
+
+---
+
+> **Memvid v1 (memoria basada en QR) está obsoleto**
+>
+> Si estás viendo referencias a códigos QR, estás usando información desactualizada.
+>
+> Consulta: https://docs.memvid.com/memvid-v1-deprecation
+
+---
+
+## Licencia
+
+Apache License 2.0 — consulta el archivo [LICENSE](LICENSE) para más detalles.

docs/i18n/README.fr.md

@@ -0,0 +1,346 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+  <strong>Memvid est une couche mémoire à fichier unique pour agents IA, avec récupération instantanée et mémoire long terme.</strong><br/>
+  Mémoire persistante, versionnée et portable, sans bases de données.
+</p>
+
+<h2 align="center">⭐️ Laissez une STAR pour soutenir le projet ⭐️</h2>
+</p>
+
+## Qu'est-ce que Memvid ?
+
+Memvid est un système de mémoire IA portable qui regroupe vos données, embeddings, structure de recherche et métadonnées dans un seul fichier.
+
+Au lieu d'exécuter des pipelines RAG complexes ou des bases de données vectorielles côté serveur, Memvid permet une récupération rapide directement depuis le fichier.
+
+Le résultat est une couche mémoire agnostique au modèle, sans infrastructure, qui donne aux agents IA une mémoire persistante et longue durée qu'ils peuvent emporter partout.
+
+---
+
+## Pourquoi des images vidéo ?
+
+Memvid s'inspire de l'encodage vidéo, non pas pour stocker de la vidéo, mais pour **organiser la mémoire IA en une séquence append-only ultra-efficace de Smart Frames.**
+
+Une Smart Frame est une unité immuable qui stocke le contenu avec des horodatages, des checksums et des métadonnées de base.
+Les frames sont regroupées d'une manière qui permet une compression, une indexation et des lectures parallèles efficaces.
+
+Ce design basé sur les frames permet :
+
+- Écritures append-only sans modifier ni corrompre les données existantes
+- Requêtes sur des états mémoire passés
+- Inspection type timeline de l'évolution des connaissances
+- Sécurité en cas de crash via des frames immuables et validées
+- Compression efficace grâce à des techniques adaptées de l'encodage vidéo
+
+Le résultat est un fichier unique qui se comporte comme une timeline mémoire rembobinable pour les systèmes IA.
+
+---
+
+## Concepts de base
+
+- **Moteur de mémoire vivant**
+  Ajoutez, branchez et faites évoluer la mémoire en continu entre les sessions.
+
+- **Capsule de Contexte  (`.mv2`)**
+  Capsules mémoire autonomes et partageables avec règles et expiration.
+
+- **Débogage par 'voyage temporel'**
+  Rembobinez, rejouez ou branchez n'importe quel état mémoire.
+
+- **Rappel intelligent**
+  Accès mémoire local en moins de 5 ms avec cache prédictif.
+
+- **Intelligence du codec**
+  Sélection et mise à niveau automatiques de la compression au fil du temps.
+
+---
+
+## Cas d'usage
+Memvid est une couche mémoire portable et sans serveur qui donne aux agents IA une mémoire persistante et un rappel rapide. Parce qu'il est agnostique au modèle, multimodal et fonctionne entièrement hors ligne, les développeurs utilisent Memvid pour un large éventail d'applications réelles.
+
+- Agents IA longue durée
+- Bases de connaissances d'entreprise
+- Systèmes IA offline-first
+- Compréhension de codebase
+- Agents de support client
+- Automatisation des workflows
+- Copilotes ventes et marketing
+- Assistants de connaissance personnels
+- Agents médicaux, juridiques et financiers
+- Workflows IA auditables et débogables
+- Applications sur mesure
+
+---
+
+## SDKs & CLI
+
+Utilisez Memvid dans votre langage préféré :
+
+| Package | Installation | Liens |
+|---------|---------|-------|
+| **CLI** | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli) |
+| **Node.js SDK** | `npm install @memvid/sdk` | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk) |
+| **Python SDK** | `pip install memvid-sdk` | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/) |
+| **Rust** | `cargo add memvid-core` | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## Installation (Rust)
+
+### Prérequis
+
+- **Rust 1.85.0+** — Installer depuis [rustup.rs](https://rustup.rs)
+
+### Ajouter à votre projet
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Feature Flags
+
+| Feature | Description |
+|---------|-------------|
+| `lex` | Full-text search with BM25 ranking (Tantivy) |
+| `pdf_extract` | Pure Rust PDF text extraction |
+| `vec` | Vector similarity search (HNSW + ONNX) |
+| `clip` | CLIP visual embeddings for image search |
+| `whisper` | Audio transcription with Whisper |
+| `temporal_track` | Natural language date parsing ("last Tuesday") |
+| `parallel_segments` | Multi-threaded ingestion |
+| `encryption` | Password-based encryption capsules (.mv2e) |
+
+Activez les features selon vos besoins :
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## Démarrage rapide
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Créer un nouveau fichier de mémoire
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Ajouter des documents avec des métadonnées
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // Rechercher
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## Compiler
+
+Cloner le repository :
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+Compiler en mode debug :
+
+```bash
+cargo build
+```
+
+Compiler en mode release (optimisé) :
+
+```bash
+cargo build --release
+```
+
+Compiler avec des features spécifiques :
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## Exécuter les tests
+
+Exécuter tous les tests :
+
+```bash
+cargo test
+```
+
+Exécuter les tests avec sortie :
+
+```bash
+cargo test -- --nocapture
+```
+
+Exécuter un test spécifique :
+
+```bash
+cargo test test_name
+```
+
+Exécuter uniquement les tests d'intégration :
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## Exemples
+
+Le répertoire `examples/` contient des exemples fonctionnels :
+
+### Utilisation de base
+
+Démontre create, put, search et les opérations de timeline :
+
+```bash
+cargo run --example basic_usage
+```
+
+### Ingestion de PDF
+
+Ingérer et rechercher des documents PDF (utilise l'article "Attention Is All You Need") :
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### Recherche visuelle CLIP
+
+Recherche d'images à l'aide d'embeddings CLIP (nécessite la feature `clip`) :
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Transcription Whisper
+
+Transcription audio (nécessite la feature `whisper`) :
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## Format de fichier
+
+Tout est dans un seul fichier `.mv2` :
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  Magic, version, capacity
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  Crash recovery
+├────────────────────────────┤
+│ Data Segments              │  Compressed frames
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy full-text
+├────────────────────────────┤
+│ Vec Index                  │  HNSW vectors
+├────────────────────────────┤
+│ Time Index                 │  Chronological ordering
+├────────────────────────────┤
+│ TOC (Footer)               │  Segment offsets
+└────────────────────────────┘
+```
+
+Pas de `.wal`, `.lock`, `.shm` ou fichiers auxiliaires. Jamais.
+
+Voir [MV2_SPEC.md](MV2_SPEC.md) pour la spécification complète du format de fichier.
+
+---
+
+## Support
+
+Vous avez des questions ou des retours ?
+Email : contact@memvid.com
+
+**Laissez une ⭐ pour montrer votre soutien**
+
+---
+
+## Licence
+
+Apache License 2.0 — voir le fichier [LICENSE](LICENSE) pour plus de détails.

docs/i18n/README.hi.md

@@ -0,0 +1,347 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+  <strong>मेमविड AI एजेंट के लिए एक सिंगल-फाइल मेमोरी लेयर है जिसमें तुरंत रिट्रीवल और लॉन्ग-टर्म मेमोरी होती है।</strong><br/>
+  बिना डेटाबेस के, स्थायी, वर्शन वाली और पोर्टेबल मेमोरी।
+</p>
+
+<h2 align="center">⭐️ प्रोजेक्ट को सपोर्ट करने के लिए एक स्टार दें। ⭐️</h2>
+</p>
+
+## What is Memvid?
+
+मेमविड एक पोर्टेबल AI मेमोरी सिस्टम है जो आपके डेटा, एम्बेडिंग, सर्च स्ट्रक्चर और मेटाडेटा को एक ही फ़ाइल में पैक करता है।
+
+जटिल RAG पाइपलाइन या सर्वर-आधारित वेक्टर डेटाबेस चलाने के बजाय, Memvid सीधे फ़ाइल से तेज़ी से डेटा रिट्रीव करने में मदद करता है।
+
+इसका नतीजा एक मॉडल-एग्नोस्टिक, इंफ्रास्ट्रक्चर-फ्री मेमोरी लेयर है जो AI एजेंट को परमानेंट, लॉन्ग-टर्म मेमोरी देती है जिसे वे कहीं भी ले जा सकते हैं।
+
+---
+
+## वीडियो फ़्रेम क्यों?
+
+मेमविड वीडियो एन्कोडिंग से प्रेरणा लेता है, वीडियो स्टोर करने के लिए नहीं, बल्कि**AI मेमोरी को स्मार्ट फ्रेम्स के अपेंड-ओनली, अल्ट्रा-एफ़िशिएंट सीक्वेंस के तौर पर ऑर्गनाइज़ करें।**
+
+स्मार्ट फ्रेम एक ऐसा यूनिट है जिसे बदला नहीं जा सकता, जो टाइमस्टैम्प, चेकसम और बेसिक मेटाडेटा के साथ कंटेंट स्टोर करता है।
+फ्रेम को इस तरह से ग्रुप किया जाता है जिससे कुशल कम्प्रेशन, इंडेक्सिंग और पैरेलल रीड संभव हो सके।
+
+यह फ्रेम-आधारित डिज़ाइन इन चीज़ों को संभव बनाता है:
+
+-   मौजूदा डेटा को संशोधित या दूषित किए बिना केवल डेटा जोड़ना
+-   पिछली मेमोरी स्थितियों पर प्रश्न
+-   ज्ञान कैसे विकसित होता है, इसकी टाइमलाइन-शैली में जांच
+-   प्रतिबद्ध, अपरिवर्तनीय फ्रेम के माध्यम से क्रैश सुरक्षा
+-   वीडियो एन्कोडिंग से अपनाई गई तकनीकों का उपयोग करके कुशल कम्प्रेशन।
+
+इसका नतीजा एक सिंगल फ़ाइल होती है जो AI सिस्टम के लिए रिवाइंड करने लायक मेमोरी टाइमलाइन की तरह काम करती है।
+
+---
+
+## मुख्य अवधारणाएँ
+
+-   **लिविंग मेमोरी इंजन**
+    सेशन के दौरान लगातार मेमोरी को जोड़ें, ब्रांच करें और विकसित करें।
+
+-   **कैप्सूल संदर्भ (`.mv2`)**
+    नियमों और एक्सपायरी के साथ सेल्फ-कंटेन्ड, शेयर करने लायक मेमोरी कैप्सूल।
+
+-   **टाइम-ट्रैवल डिबगिंग**
+    किसी भी मेमोरी स्टेट को रिवाइंड, रिप्ले या ब्रांच करें।
+
+-   **स्मार्ट रिकॉल**
+    प्रेडिक्टिव कैशिंग के साथ सब-5ms लोकल मेमोरी एक्सेस।
+
+-   **कोडेक इंटेलिजेंस**
+    यह समय के साथ कम्प्रेशन को ऑटो-सेलेक्ट और अपग्रेड करता है।
+
+---
+
+## उपयोग के मामले
+
+मेमविड एक पोर्टेबल, सर्वरलेस मेमोरी लेयर है जो AI एजेंट को परमानेंट मेमोरी और तेज़ रिकॉल देता है। क्योंकि यह मॉडल-एग्नोस्टिक, मल्टी-मॉडल है, और पूरी तरह से ऑफ़लाइन काम करता है, इसलिए डेवलपर्स मेमविड का इस्तेमाल कई तरह के रियल-वर्ल्ड एप्लीकेशन में कर रहे हैं।
+
+- लंबे समय तक चलने वाले AI एजेंट
+- एंटरप्राइज़ नॉलेज बेस
+- ऑफ़लाइन-फ़र्स्ट AI सिस्टम
+- कोडबेस को समझना
+- कस्टमर सपोर्ट एजेंट
+- वर्कफ़्लो ऑटोमेशन
+- सेल्स और मार्केटिंग कोपायलट
+- पर्सनल नॉलेज असिस्टेंट
+- मेडिकल, लीगल और फाइनेंशियल एजेंट
+- ऑडिट करने योग्य और डीबग करने योग्य AI वर्कफ़्लो
+- कस्टम एप्लीकेशन
+
+---
+
+## SDKs & CLI
+
+मेमविड को अपनी पसंदीदा भाषा में इस्तेमाल करें:
+
+| Package         | Install                     | Links                                                                                                               |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## Installation (Rust)
+
+### आवश्यकताएं
+
+-   **Rust 1.85.0+** — Install from [rustup.rs](https://rustup.rs)
+
+### अपने प्रोजेक्ट में जोड़ें
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Feature Flags
+
+| Feature             | Description                                    |
+| ------------------- | ---------------------------------------------- |
+| `lex`               | Full-text search with BM25 ranking (Tantivy)   |
+| `pdf_extract`       | Pure Rust PDF text extraction                  |
+| `vec`               | Vector similarity search (HNSW + ONNX)         |
+| `clip`              | CLIP visual embeddings for image search        |
+| `whisper`           | Audio transcription with Whisper               |
+| `temporal_track`    | Natural language date parsing ("last Tuesday") |
+| `parallel_segments` | Multi-threaded ingestion                       |
+| `encryption`        | Password-based encryption capsules (.mv2e)     |
+
+Enable features as needed:
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## त्वरित प्रारंभ
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Create a new memory file
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Add documents with metadata
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // Search
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## निर्माण
+
+रिपॉजिटरी को क्लोन करें:
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+डीबग मोड में बिल्ड करें:
+
+```bash
+cargo build
+```
+
+रिलीज़ मोड में बिल्ड करें (ऑप्टिमाइज़्ड):
+
+```bash
+cargo build --release
+```
+
+विशिष्ट विशेषताओं के साथ बनाएं:
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## टेस्ट चलाएँ
+
+सभी टेस्ट चलाएँ:
+
+```bash
+cargo test
+```
+
+आउटपुट के साथ टेस्ट चलाएँ:
+
+```bash
+cargo test -- --nocapture
+```
+
+एक विशिष्ट टेस्ट चलाएँ:
+
+```bash
+cargo test test_name
+```
+
+केवल इंटीग्रेशन टेस्ट चलाएँ:
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## उदाहरण
+
+The `examples/` डायरेक्टरी में काम करने वाले उदाहरण हैं:
+
+### बेसिक उपयोग
+
+यह क्रिएट, पुट, सर्च और टाइमलाइन ऑपरेशन दिखाता है:
+
+```bash
+cargo run --example basic_usage
+```
+
+### PDF इन्जेक्शन
+
+PDF डॉक्यूमेंट्स को इन्जेस्ट करें और सर्च करें ("अटेंशन इज़ ऑल यू नीड" पेपर का इस्तेमाल करता है):
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### CLIP विज़ुअल सर्च
+
+CLIP एम्बेडिंग का इस्तेमाल करके इमेज सर्च (इसके लिए `clip` फ़ीचर ज़रूरी है):
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### व्हिस्पर ट्रांसक्रिप्शन
+
+ऑडियो ट्रांसक्रिप्शन (`whisper` फीचर ज़रूरी है):
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## फ़ाइल फ़ॉर्मेट
+
+सब कुछ एक ही `.mv2` फ़ाइल में होता है:
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  Magic, version, capacity
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  Crash recovery
+├────────────────────────────┤
+│ Data Segments              │  Compressed frames
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy full-text
+├────────────────────────────┤
+│ Vec Index                  │  HNSW vectors
+├────────────────────────────┤
+│ Time Index                 │  Chronological ordering
+├────────────────────────────┤
+│ TOC (Footer)               │  Segment offsets
+└────────────────────────────┘
+```
+
+कोई `.wal`, `.lock`, `.shm`, या साइडकार फ़ाइल नहीं। कभी नहीं।
+
+पूरे फ़ाइल फ़ॉर्मेट स्पेसिफ़िकेशन के लिए [MV2_SPEC.md](MV2_SPEC.md) देखें।
+
+---
+
+## सपोर्ट
+
+क्या आपके कोई सवाल या फीडबैक हैं?
+Email: contact@memvid.com
+
+**सपोर्ट दिखाने के लिए ⭐ दें**
+
+---
+
+## लाइसेंस
+
+Apache License 2.0 — ज़्यादा जानकारी के लिए [LICENSE](LICENSE) फ़ाइल देखें।

docs/i18n/README.ja.md

@@ -0,0 +1,429 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+<strong>Memvidは、AIエージェントのための即時検索と長期記憶を備えた単一ファイル型メモリレイヤーです。</strong>
+</br>
+データベースを必要とせず、永続化、バージョン管理、ポータブル性を備えたメモリを実現します。
+</p>
+
+<h2 align="center">⭐️ スターで応援お願いします ⭐️</h2>
+</p>
+
+## Memvidとは？
+
+Memvidは、データ、埋め込み、検索構造、メタデータを１つのファイルにパッケージ化するポータブルAIメモリシステムです。
+
+複雑なRAGパイプラインやサーバーベースのベクトルデータベースを運用する代わりに、Memvidを使用することで直接ファイルから高速な検索が可能になります。
+
+その結果、モデルに依存せずインフラ不要のメモリレイヤーが実現し、AIエージェントはどこでも使える永続的な長期記憶を持つことができます。
+
+---
+
+## スマートフレーム (Smart Frames) とは？
+
+Memvidは、（ビデオを保存するためではなく）**追記に特化した効率的なスマートフレームのシーケンスとしてAIメモリを整理するため**に、ビデオエンコーディング技術から着想を得ています。
+
+スマートフレームは、コンテンツをタイムスタンプ、チェックサム、基本メタデータとともに保存する不変（イミュータブル）な単位です。フレームは効率的な圧縮、インデックス作成、並列読み取りができるようグループ化されています。
+
+このフレームベースの設計により、以下が可能になります。
+
+- 既存のデータを変更したり破損したりすることなくデータを追加
+- 過去のメモリ状態に対するクエリ
+- 知識がどのように進化するかをタイムライン形式で検査
+- コミットされた不変フレームによるクラッシュ耐性
+- ビデオエンコーディング技術を応用した効率的な圧縮
+
+その結果、AIシステムの「巻き戻し可能なメモリタイムライン」のように機能する単一のファイルが生成されます。
+
+---
+
+## コアコンセプト
+
+- **成長するメモリエンジン (Living Memory Engine)**
+  セッションをまたいでメモリを継続的に追加、分岐、進化させます。
+
+- **カプセル・コンテキスト (`.mv2`)**
+  ルールや有効期限を設定できる、自己完結型で共有可能なメモリカプセル。
+
+- **タイムトラベル・デバッグ**
+  任意のメモリ状態を巻き戻し、再現、または分岐させることができます。
+
+- **スマート・リコール**
+  予測キャッシングによる5ミリ秒未満のローカルメモリーアクセス。
+
+- **コーデック・インテリジェンス**
+  圧縮方式を自動選択し、時間の経過とともにアップグレードします。
+
+---
+
+## ユースケース
+
+Memvidは、AIエージェントに永続的な記憶と高速な呼び出し機能を提供するポータブルでサーバーレスなメモリレイヤーです。モデルに依存せず、マルチモーダルに対応し、完全にオフラインで動作するため、実用的なアプリケーションで幅広く利用されています。
+
+- 長期稼働AIエージェント
+- エンタープライズ向けナレッジベース
+- オフラインファーストAIシステム
+- コードベースの理解
+- カスタマーサポートエージェント
+- ワークフロー自動化
+- セールス・マーケティング支援
+- パーソナル・ナレッジ・アシスタント
+- 医療・法律・金融特化型エージェント
+- 監査・デバッグ可能なAIワークフロー
+- カスタムアプリケーション
+
+---
+
+## SDK と CLI
+
+お好みの言語でMemvidを利用できます。
+
+| パッケージ      | インストール                | リンク                                                                                                              |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## インストール (Rust)
+
+### 要件
+
+- **Rust 1.85.0+** - [rustup.rs](https://rustup.rs) からインストールしてください。
+
+### プロジェクトへの追加
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+
+```
+
+### 機能フラグ (Feature Flags)
+
+| 機能                | 説明                                                           |
+| ------------------- | -------------------------------------------------------------- |
+| `lex`               | BM25ランキングによる全文検索 (Tantivy)                         |
+| `pdf_extract`       | Pure RustによるPDFテキスト抽出                                 |
+| `vec`               | ベクトル類似性検索 (HNSW + ONNXによるローカルテキスト埋め込み) |
+| `clip`              | 画像検索用のCLIPビジュアル埋め込み                             |
+| `whisper`           | Whisperによる音声文字起こし                                    |
+| `temporal_track`    | 自然言語による日付解析 (例: "last Tuesday")                    |
+| `parallel_segments` | マルチスレッドによるデータ取り込み                             |
+| `encryption`        | パスワードベースの暗号化カプセル (.mv2e)                       |
+
+以下のように、必要に応じて有効化してください。
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## クイックスタート
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // 新しいメモリファイルを作成
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // メタデータ付きでドキュメントを追加
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // 検索の実行
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## ビルド
+
+リポジトリをクローン：
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+デバッグモードでビルド：
+
+```bash
+cargo build
+```
+
+リリースモードでビルド（最適化）：
+
+```bash
+cargo build --release
+```
+
+特定の機能フラグ付きでビルド：
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## テストの実行
+
+すべてのテストを実行：
+
+```bash
+cargo test
+```
+
+標準出力でテストを実行：
+
+```bash
+cargo test -- --nocapture
+```
+
+特定のテストを実行：
+
+```bash
+cargo test test_name
+```
+
+統合テストのみを実行：
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## サンプル (Examples)
+
+`examples/` ディレクトリには、実際に動作するサンプルコードが用意されています。
+
+### 基本的な使い方 (Basic Usage)
+
+作成 (create)、追加 (put)、検索 (search)、およびタイムライン操作のデモです。
+
+```bash
+cargo run --example basic_usage
+```
+
+### PDFの取り込み (PDF Ingestion)
+
+PDFドキュメントの取り込みと検索のサンプルです。（論文「Attention Is All You Need」を使用）
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### CLIPによる画像検索 (CLIP Visual Search)
+
+CLIP埋め込みを使用した画像検索のサンプルです。
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Whisperによる文字起こし (Whisper Transcription)
+
+音声文字起こしのサンプルです。
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## テキスト埋め込みモデル
+
+`vec` 機能は、ONNXモデルを使用したローカルでのテキスト埋め込みをサポートしています。利用前にモデルファイルを手動でダウンロードする必要があります。
+
+### 推奨：BGE-small (デフォルト)
+
+高速で効率的なBGE-smallモデル（384次元）をダウンロードします。
+
+```bash
+mkdir -p ~/.cache/memvid/text-models
+
+# ONNXモデルのダウンロード
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5.onnx
+
+# トークナイザーのダウンロード
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5_tokenizer.json
+```
+
+### モデル一覧
+
+| モデル                  | 次元数 | サイズ | 最適な用途               |
+| ----------------------- | ------ | ------ | ------------------------ |
+| `bge-small-en-v1.5`     | 384    | ~120MB | デフォルト（高速・軽量） |
+| `bge-base-en-v1.5`      | 768    | ~420MB | より高い精度が必要な場合 |
+| `nomic-embed-text-v1.5` | 768    | ~530MB | 多目的なタスク           |
+| `gte-large`             | 1024   | ~1.3GB | 最高精度                 |
+
+### 他のモデル
+
+**BGE-base** (768次元):
+
+```bash
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5.onnx
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5_tokenizer.json
+```
+
+**Nomic** (768次元):
+
+```bash
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5.onnx
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5_tokenizer.json
+```
+
+**GTE-large** (1024次元):
+
+```bash
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/gte-large.onnx
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/gte-large_tokenizer.json
+```
+
+### 使用例
+
+```rust
+use memvid_core::text_embed::{LocalTextEmbedder, TextEmbedConfig};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// デフォルトモデルを使用する場合 (BGE-small)
+let config = TextEmbedConfig::default();
+let embedder = LocalTextEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 384);
+
+// モデルを変更する場合
+let config = TextEmbedConfig::bge_base();
+let embedder = LocalTextEmbedder::new(config)?;
+```
+
+類似性の計算と検索ランキングを含む完全な例については、`examples/text_embedding.rs` を参照してください。
+
+---
+
+## ファイル構成
+
+すべてが単一の `.mv2` ファイルに収められます。
+
+```
+┌────────────────────────────┐
+│ ヘッダー (4KB)              │  マジックナンバー、バージョン、容量
+├────────────────────────────┤
+│ 組み込みWAL (1-64MB)       │  クラッシュリカバリ用
+├────────────────────────────┤
+│ データセグメント            │  圧縮されたフレーム
+├────────────────────────────┤
+│ 全文検索インデックス (Lex)  │  Tantivy全文検索
+├────────────────────────────┤
+│ ベクトルインデックス (Vec)  │  HNSWベクトル
+├────────────────────────────┤
+│ タイムインデックス          │  時系列順序
+├────────────────────────────┤
+│ TOC (フッター)             │  セグメントオフセット
+└────────────────────────────┘
+
+```
+
+`.wal`、`.lock`、`.shm` などのサイドカーファイルは一切生成されません。
+
+フォーマット仕様の詳細は [MV2_SPEC.md](MV2_SPEC.md) を参照してください。
+
+---
+
+## サポート
+
+ご質問やフィードバックはこちらまでご連絡ください。
+メール: contact@memvid.com
+
+**⭐でプロジェクトをサポートしてください。**
+
+---
+
+## ライセンス
+
+Apache License 2.0 - 詳細は [LICENSE](LICENSE) ファイルをご覧ください。

docs/i18n/README.ko.md

@@ -0,0 +1,424 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+  <strong>Memvid는 AI 에이전트를 위한 단일 파일 메모리 레이어로, 인스턴스 검색 및 장기 메모리 기능을 제공합니다.</strong><br/>
+  데이터 베이스 없이 지속적이고, 버전 관리가 용이하며 여러 어플리케이션에 자유로운 적용이 가능합니다.
+</p>
+
+<h2 align="center">⭐️ STAR로 이 프로젝트를 지원해주세요 ⭐️</h2>
+</p>
+
+## Memvid란?
+
+Memvid는 데이터, 임베딩, 검색 구조 및 메타데이터를 단일 파일로 패키징하는 이식 가능한 AI 메모리 시스템입니다.
+​
+복잡한 RAG 파이프라인이나 서버 기반 벡터 데이터베이스를 실행하는 대신, Memvid는 파일에서 직접 빠른 검색을 가능하게 합니다.
+
+결과적으로 모델에 독립적이며 인프라 구조와는 독립적인 메모리 레이어로, AI 에이전트가 어디서나 휴대할 수 있는 지속적 장기 메모리를 제공합니다
+
+---
+
+## Smart Frames란?
+
+Memvid는 **AI 메모리를 추가 전용(append-only)의 초고효율 Smart Frame 시퀀스로 구성하기 위해** 비디오 인코딩에서 영감을 받았습니다.
+
+Smart Frame은 타임스탬프, 체크섬 및 기본 메타데이터와 함께 콘텐츠를 저장하는 불변 단위입니다.
+프레임은 효율적인 압축, 인덱싱 및 병렬 읽기를 허용하는 방식으로 그룹화됩니다.
+
+이러한 프레임 기반 설계는 다음을 가능하게 합니다:
+
+-   기존 데이터를 수정하거나 손상시키지 않는 추가 전용(append-only) 쓰기
+-   과거 메모리 상태에 대한 쿼리
+-   지식이 어떻게 변화하는지에 대한 타임라인 스타일 검사
+-   불변 프레임워크를 통한 크래시 안전성
+-   비디오 인코딩에서 차용한 기술을 사용한 효율적인 압축
+
+이를 위한 결과물은 AI 시스템을 위한 되감기 가능한 메모리 타임라인처럼 동작하는 단일 파일입니다.
+
+---
+
+## 주요 개념
+
+-   **실시간 변화하는 메모리 엔진**
+    세션 간에 메모리를 지속적으로 추가, 분기 및 변화시킵니다.
+
+-   **문맥 캡슐화 (`.mv2`)**
+    규칙과 만료 시간이 포함된 자립형 공유 가능 형대의 메모리 캡슐입니다.
+
+-   **시간 기반 디버깅**
+    임의의 메모리 상태로 되감기, 재생 또는 분기합니다.
+
+-   **예측 기반 호출**
+    예측 캐싱을 사용한 5ms 미만 로컬 메모리 액세스를 제공합니다.
+
+-   **코덱 인텔리전스**
+    시간 경과에 따라 압축을 자동 선택 및 업그레이드합니다.
+
+---
+
+## 이용 사례
+
+Memvid 이동 가능한 서버리스 메모리 레이어로 AI 에이전트에 지속적인 메모리와 빠른 호출을 제공합니다. 이는 모델과 독립적이고, 멀티모달을 지원하며, 인터넷을 사용하지 않으므로, 개발자들은 다양한 실제 어플리케이션에서 Memvid를 활용하고 있습니다.
+
+- 장기 실행 AI 에이전트
+- 기업 내의 지식 베이스
+- 오프라인 우선의 AI 시스템
+- 코드베이스 이해
+- 고객 지원 에이전트
+- 워크플로 자동화
+- 판매 및 마케팅 코파일럿
+- 개인 지식 어시스턴트
+- 의료, 법률 및 금융 에이전트
+- 모니터링 및 디버깅 가능한 AI 워크플로
+- 그 외의 여러 애플리케이션
+
+---
+
+## SDKs & CLI
+
+원하는 언어로 Memvid를 사용하세요:
+
+| 패키지         | 설치 커맨드                     | 링크                                                                                                               |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## 설치 (Rust)
+
+### 요구 사항
+
+-   **Rust 1.85.0+** — [rustup.rs](https://rustup.rs)에서 설치 가능합니다.
+
+### 프로젝트에 추가
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Feature Flags
+
+| Feature             | Description                                         |
+| ------------------- | --------------------------------------------------- |
+| `lex`               | BM25 랭킹 기반 전체 텍스트 검색 (Tantivy)              |
+| `pdf_extract`       | Rust 기반 PDF 텍스트 추출                       |
+| `vec`               | 벡터 유사도 검색 (HNSW + ONNX) |
+| `clip`              | 이미지 검색을 위한 CLIP 임베딩             |
+| `whisper`           | Whisper 기반 오디오 전사                    |
+| `temporal_track`    | 자연어 날짜 추출 ("지난 화요일")      |
+| `parallel_segments` | 멀티-스레딩 처리                            |
+| `encryption`        | Password 기반 암호화 (.mv2e)          |
+
+필요한 기능을 아래 방식으로 활성화하세요:
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## Quick Start
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Create a new memory file
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Add documents with metadata
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // Search
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## 빌드
+
+이 레포지토리 클론:
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+디버그 모드로 빌드:
+
+```bash
+cargo build
+```
+
+배포 모드로 빌드 (optimized):
+
+```bash
+cargo build --release
+```
+
+특수 기능을 포함하도록 빌드:
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## 테스트
+
+전체 테스트 실행:
+
+```bash
+cargo test
+```
+
+테스트 실행 및 결과 출력:
+
+```bash
+cargo test -- --nocapture
+```
+
+특정 테스트 실행:
+
+```bash
+cargo test test_name
+```
+
+인테그레이션 테스트만 실행:
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## 예시
+
+`examples/` 디렉토리에 예제가 있습니다:
+
+### 기본 사용법
+
+생성, 추가, 검색 및 타임라인 작업을 보여줍니다:
+
+```bash
+cargo run --example basic_usage
+```
+
+### PDF 수집
+
+PDF 문서 수집 및 검색 ("Attention Is All You Need" 논문 사용):
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### CLIP 이미지 검색
+
+CLIP 임베딩을 사용한 이미지 검색 (`clip` 기능 필요):
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Whisper 전사
+
+오디오 전사 (`whisper` 기능 필요):
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## Text Embedding 모델
+
+`vec` 기능은 ONNX 모델을 사용한 로컬 텍스트 임베딩을 포함합니다. 로컬 텍스트 임베딩을 사용하기 전에 모델 파일을 수동으로 다운로드해야 합니다.
+
+### Quick Start: BGE-small (추천함)
+
+기본 BGE-small 모델(384 차원, 빠르고 효율적) 다운로드:
+
+```bash
+mkdir -p ~/.cache/memvid/text-models
+
+# Download ONNX model
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5.onnx
+
+# Download tokenizer
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5_tokenizer.json
+```
+
+### 지원 모델
+
+| 모델명                   | 차원 수     | 크기   | 권장 용도               |
+| ----------------------- | ---------- | -----  | --------------------- |
+| `bge-small-en-v1.5`     | 384        | ~120MB | 기본 설정, 가장 빠름     |
+| `bge-base-en-v1.5`      | 768        | ~420MB | 꽤 좋은 성능            |
+| `nomic-embed-text-v1.5` | 768        | ~530MB | 다양한 업무 가능       |
+| `gte-large`             | 1024       | ~1.3GB | 가장 좋은 성능      |
+
+### 타 모델
+
+**BGE-base** (768 dimensions):
+```bash
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5.onnx
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5_tokenizer.json
+```
+
+**Nomic** (768 dimensions):
+```bash
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5.onnx
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5_tokenizer.json
+```
+
+**GTE-large** (1024 dimensions):
+```bash
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/gte-large.onnx
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/gte-large_tokenizer.json
+```
+
+### 코드 내 사용법
+
+```rust
+use memvid_core::text_embed::{LocalTextEmbedder, TextEmbedConfig};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// Use default model (BGE-small)
+let config = TextEmbedConfig::default();
+let embedder = LocalTextEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 384);
+
+// Use different model
+let config = TextEmbedConfig::bge_base();
+let embedder = LocalTextEmbedder::new(config)?;
+```
+
+유사도 계산 및 검색 랭킹이 포함된 전체 예제는 `examples/text_embedding.rs`를 참조하세요.
+
+---
+
+## 파일 구조
+
+모든 구성 요소는 단일 `.mv2` 파일 내에 구성됩니다:
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  Magic, version, capacity
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  Crash recovery
+├────────────────────────────┤
+│ Data Segments              │  Compressed frames
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy full-text
+├────────────────────────────┤
+│ Vec Index                  │  HNSW vectors
+├────────────────────────────┤
+│ Time Index                 │  Chronological ordering
+├────────────────────────────┤
+│ TOC (Footer)               │  Segment offsets
+└────────────────────────────┘
+```
+
+`.wal`, `.lock`, `.shm`, 혹은 그 외의 별도 구성 요소는 없습니다.
+
+[MV2_SPEC.md](MV2_SPEC.md)에서 파일 세부 형식을 확인할 수 있습니다.
+
+---
+
+## Support
+
+문의 사항은 아래 이메일로 부탁드립니다.
+Email: contact@memvid.com
+
+**⭐를 눌러 이 프로젝트를 지원해주세요**
+
+---
+
+## License
+
+Apache License 2.0 — [LICENSE](LICENSE) 파일 참고.

docs/i18n/README.md

@@ -0,0 +1,90 @@
+# Internationalization (i18n) - README Translations
+
+This folder contains translated versions of the main [README.md](../../README.md) file for different languages.
+
+## Purpose
+
+The `docs/i18n/` directory is dedicated to making Memvid accessible to developers worldwide by providing localized versions of the main README. Each translation helps non-English speakers understand and use Memvid more effectively.
+
+## Structure
+
+```
+docs/i18n/
+├── README.md          # This file
+├── README.zh-CN.md    # Chinese (Simplified) translation
+├── README.zh-TW.md    # Chinese (Traditional) translation
+├── README.es.md       # Spanish translation
+├── README.fr.md       # French translation
+├── README.de.md       # German translation
+├── README.ja.md       # Japanese translation
+├── README.ko.md       # Korean translation
+├── README.pt-BR.md    # Portuguese (Brazil) translation
+├── README.so.md       # Somali translation
+└── ...                # Additional languages
+```
+
+## Contributing Translations
+
+We welcome contributions of README translations! For detailed guidelines, see [Contributing Translations](CONTRIBUTING_TRANSLATIONS.md).
+
+Here's a quick overview:
+
+### 1. Check Existing Translations
+
+Before starting, check if a translation for your language already exists or is in progress.
+
+### 2. Create a Translation File
+
+- Use the format: `README.{language-code}.md`
+- Use standard language codes (e.g., `zh-CN`, `es`, `fr`, `de`, `ja`, `ko`, `pt-BR`, `so`)
+- Copy the main `README.md` as a starting point
+
+### 3. Translation Guidelines
+
+- **Keep the structure**: Maintain the same headings, sections, and formatting as the original
+- **Preserve links**: Keep all URLs and links unchanged
+- **Preserve code blocks**: Keep code examples, commands, and technical terms in English (or add comments in the target language)
+- **Update badges**: Keep badges and shields as they are (they're language-agnostic)
+- **Maintain accuracy**: Ensure technical accuracy while making the content natural in the target language
+
+### 4. Submit Your Translation
+
+1. Create a new file: `README.{language-code}.md` in this directory
+2. Translate the content while following the guidelines above
+3. Submit a pull request with:
+   - A clear description of the language being added
+   - Your name/username for attribution (if desired)
+
+## Language Codes
+
+Use standard ISO 639-1 or ISO 639-2 language codes:
+
+- `zh-CN` - Chinese (Simplified)
+- `zh-TW` - Chinese (Traditional)
+- `es` - Spanish
+- `fr` - French
+- `de` - German
+- `ja` - Japanese
+- `ko` - Korean
+- `pt-BR` - Portuguese (Brazil)
+- `ru` - Russian
+- `ar` - Arabic
+- `hi` - Hindi
+- `so` - Somali
+- And more...
+
+## Maintenance
+
+Translations should be updated when the main README is significantly changed. Contributors are encouraged to keep their translations in sync with the English version.
+
+## Questions?
+
+If you have questions about translations or want to coordinate with other translators, please:
+- See [Contributing Translations](CONTRIBUTING_TRANSLATIONS.md) for detailed guidelines
+- Open an issue on GitHub
+- Join our [Discussions](https://github.com/memvid/memvid/discussions)
+- Contact: contact@memvid.com
+
+---
+
+**Thank you for helping make Memvid accessible to developers worldwide! 🌍**

docs/i18n/README.nl.md

@@ -0,0 +1,347 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+  <strong>Memvid is een geheugenlaag van één bestand voor AI-agenten met directe toegang en langetermijnsgeheugen.</strong><br/>
+  Volhardend en draagbaar geheugen met versiebeheer en zonder databases.
+</p>
+
+<h2 align="center">⭐️ Laat een ster achter om het project te steunen ⭐️</h2>
+</p>
+
+## Wat is Memvid?
+
+Memvid is een draagbaar AI-geheugensysteem dat uw data, embedding, zoekstructuur en metadata in één bestand opslaat.
+
+In plaats van complexe RAG pijplijnen of servergebaseerde vectordatabases te gebruiken, zal Memvid snelle toegang recht vanuit het bestand toestaan.
+
+Het resultaat is een model-agnostische, infrastructuurvrije geheugenlaag die AI-agenten een volhardende langetermijnsgeheugen geeft, die ze overal kunnen meenemen.
+
+---
+
+## Waarom videoframes?
+
+Memvid neemt inspiratie uit videos encoderen, niet om de video op te slaan, maar om **het organiseren van AI-geheugen als een ultra-efficiënte sequentie van Smart Frames waarbij je enkel kan toevoegen.**
+
+Een Smart Frame is een immutabele eenheid die content opslaat samen met zijn tijdstempels, controlesommen en basismetadata.
+Frames worden gegroupeerd in een manier die voor efficiënte compressie, indexing en parallele lezingen zorgt.
+
+Dit frame-gebaseerde design maakt het volgende mogelijk:
+
+- Append-only bijschrijven van data zonder het aanpassen of corrumperen van bestaande data
+- Zoekopdrachten over vorige geheugenstaten
+- Tijdlijn-stijl inspectie van hoe kennis evolueert
+- Crashveiligheid door de vastgelegde immutabele frames
+- Efficiënte compressie gebruikmakend van technieken aangepast uit video encoderen
+
+Het resultaat is één bestand dat werkt als een terugspoelbare geheugentijdslijn van AI-systemen.
+
+---
+
+## Basisconcepten
+
+-   **Living Memory Engine**
+    Append, vertakt en evolueert geheugen continu over sessies.
+
+-   **Capsule Context (`.mv2`)**
+    Autonome, deelbaar geheugencapsules met regels en vervalling.
+
+-   **Time-Travel Debugging**
+    Spoel terug, herspeel, of vertak elke geheugenstatus.
+
+-   **Smart Recall**
+    Sub-5ms lokale geheugentoegang met voorspelbare caching.
+
+-   **Codec Intelligence**
+    Selecteert en verbetert automatisch de compressie doorheen de tijd.
+
+---
+
+## Gebruiksgevallen
+
+Memvid is een draagbare, serverloze geheugenlaag dat AI-agenten een volhardend geheugen en snelle herroepingen geeft. Door zijn model-agnostische, multi-modale en het feit dat het volledig offline werkt, gebruiken ontwikkelaars het over een wijd scala aan real-world applicaties.
+
+-   Lang werkende AI-agenten
+-   Kennisbanken voor ondernemingen
+-   Offline-First AI-systemen
+-   Codebase-begrip
+-   Klantenondersteuningsagenten
+-   Automatisering van de workflow
+-   Verkoop- en marketingcopiloten
+-   Persoonlijke Kennisassistenten
+-   Medische, juridische en financiële adviseurs
+-   Controleerbare en debugbare AI-workflows
+-   Aangepaste toepassingen
+
+---
+
+## SDKs & CLI
+
+Gebruik Memvid in je lievelingstaal:
+
+| Pakket          | Installatie                 | Links                                                                                                               |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## Installatie (Rust)
+
+### Benodigdheden
+
+-   **Rust 1.85.0+** — Installeer vanuit [rustup.rs](https://rustup.rs)
+
+### Voeg dit aan je project toe
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Feature Flags
+
+| Feature             | Beschrijving                                   |
+| ------------------- | ---------------------------------------------- |
+| `lex`               | Full-text search with BM25 ranking (Tantivy)   |
+| `pdf_extract`       | Pure Rust PDF text extraction                  |
+| `vec`               | Vector similarity search (HNSW + ONNX)         |
+| `clip`              | CLIP visual embeddings for image search        |
+| `whisper`           | Audio transcription with Whisper               |
+| `temporal_track`    | Natural language date parsing ("last Tuesday") |
+| `parallel_segments` | Multi-threaded ingestion                       |
+| `encryption`        | Password-based encryption capsules (.mv2e)     |
+
+Schakel functies in indien nodig:
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## Quick Start
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Create a new memory file
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Add documents with metadata
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // Search
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## Build
+
+Clone de repository:
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+Build in debug modus:
+
+```bash
+cargo build
+```
+
+Build in release modus (geoptimaliseerd):
+
+```bash
+cargo build --release
+```
+
+Build with specific features:
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## Tests uitvoeren
+
+Voer alle tests uit:
+
+```bash
+cargo test
+```
+
+Voer tests uit met uitvoer:
+
+```bash
+cargo test -- --nocapture
+```
+
+Voer een specifieke test uit:
+
+```bash
+cargo test test_name
+```
+
+Voer enkel integratie tests uit:
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## Voorbeelden
+
+De `examples/` map bedraagd werkende Voorbeelden:
+
+### Basisgebruik
+
+Beeldt create, put, search, and timeline operaties uit:
+
+```bash
+cargo run --example basic_usage
+```
+
+### PDF Ingestion
+
+PDF-documenten importeren en doorzoeken (gebruikt de "Attention Is All You Need" paper):
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### CLIP Visual Search
+
+Afbeeldingen zoeken met behulp van CLIP-integraties (gebruikt `clip` feature):
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Whisper Transcription
+
+Audio transcripties (gebruikt `whisper` feature):
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## Bestandsformaat
+
+Alles leeft in één `.mv2` bestand:
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  Magic, version, capacity
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  Crash recovery
+├────────────────────────────┤
+│ Data Segments              │  Compressed frames
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy full-text
+├────────────────────────────┤
+│ Vec Index                  │  HNSW vectors
+├────────────────────────────┤
+│ Time Index                 │  Chronological ordering
+├────────────────────────────┤
+│ TOC (Footer)               │  Segment offsets
+└────────────────────────────┘
+```
+
+Geen `.wal`, `.lock`, `.shm`, of sidecar-bestanden. Ooit.
+
+Zie [MV2_SPEC.md](MV2_SPEC.md) voor de complete bestandsformaat specificaties.
+
+---
+
+## Ondersteuning
+
+Heb je vragen of feedback?
+Email: contact@memvid.com
+
+**Laat een ⭐ om je ondersteuning te tonen**
+
+---
+
+## Licentie
+
+Apache License 2.0 — zie het [LICENSE](LICENSE) bestandvoor details.

docs/i18n/README.so.md

@@ -0,0 +1,345 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)" src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<!-- FLAGS:START -->
+<p align="center">
+ <a href="../../README.md">🇺🇸 English</a>
+ <a href="README.es.md">🇪🇸 Español</a>
+ <a href="README.fr.md">🇫🇷 Français</a>
+ <a href="README.so.md">🇸🇴 Soomaali</a>
+ <a href="README.ar.md">🇸🇦 العربية</a>
+ <a href="README.nl.md">🇧🇪/🇳🇱 Nederlands</a>
+ <a href="README.hi.md">🇮🇳 हिन्दी</a>
+ <a href="README.bn.md">🇧🇩 বাংলা</a>
+ <a href="README.cs.md">🇨🇿 Čeština</a>
+ <a href="README.ko.md">🇰🇷 한국어</a>
+ <a href="README.ja.md">🇯🇵 日本語</a>
+ <!-- Next Flag -->
+</p>
+<!-- FLAGS:END -->
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">Website</a>
+  ·
+  <a href="https://sandbox.memvid.com">Try Sandbox</a>
+  ·
+  <a href="https://docs.memvid.com">Docs</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">Discussions</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="License" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+  <strong>Memvid waa nidaam xusuuseed oo hal fayl ah kaas oo loogu talagalay wakiillada AI (AI agents), lehna soo-celin degdeg ah iyo xusuus fog.</strong><br/>
+  Xusuus joogto ah, la raadin karo, lana qaadan karo, iyadoo aan loo baahnayn database-yo kale.
+</p>
+
+<h2 align="center">⭐️ Noo saar STAR si aad mashruuca u taageerto ⭐️</h2>
+
+## Waa maxay Memvid?
+
+Memvid waa nidaam xusuuseed AI oo la qaadan karo kaas oo kuu keydinaya xogtaada, habka raadinta (embeddings), qaabdhismeedka iyo metadata-daba ku ururiya hal fayl oo keliya.
+
+Halkii aad ka isticmaali lahayd nidaamyada RAG-ga ee adag ama database-yada vector-ka ee ku shaqeeya server-ka, Memvid wuxuu kuu oggolaanayaa inaad xogta si toos ah uga soo ceshato faylka dhexdiisa si aad u degdeg badan.
+
+Natiijadu waa lakab xusuuseed ka madax-bannaan nooca modelka iyo kaabayaasha (infrastructure-free), kaas oo siiya wakiillada AI(AI agents) xusuus joogto ah oo fog oo ay meel walba u qaadan karaan.
+
+---
+
+## Maxay tahay sababta Frames-ka Muuqaalka (Video Frames)?
+
+Memvid wuxuu dhiirrigelin ka helayaa habka xogta muuqaalka loo kaydiyo (video encoding), isaga oo aan kaydinayn muuqaal balse u **habaynaya xusuusta AI si isku-xiga (sequence) oo aad u hufan oo "Smart Frames".**
+
+"Smart Frame" waa unug aan isbeddelayn oo kaydiya macluumaadka oo ay la socdaan waqtiga (timestamps), checksums iyo metadata aasaasi ah. Frames-ka waxaa loo ururiyaa qaab oggolaanaya isku-duubni (compression), tusmeyn (indexing), iyo akhris is-barbar-socda oo hufan.
+
+Qaabdhismeedkan ku salaysan frames-ka wuxuu suuragelinayaa:
+
+-   Qoraal kaliya oo lagu darayo (Append-only) iyadoo aan la beddelayn ama la kharribayn xogta jirtay
+-   Baaritaan lagu sameyn karo xaaladihii xusuusta ee hore
+-   Kormeeridda habka ay aqoontu u kobcayso iyadoo loo eegayo waqtiga
+-   Badbaadada xogta (crash safety) iyadoo la adeegsanayo frames go'an oo aan isbeddelayn
+-   Isku-duubni hufan oo loo adeegsanayo farsamooyin laga soo minguuriyay kaydinta muuqaallada
+
+Natiijadu waa hal fayl oo u dhaqmaya sidii jadwal xusuuseed oo dib loo celin karo oo loogu talagalay nidaamyada AI.
+
+---
+
+## Fikradaha Muhiimka ah
+
+-   **Living Memory Engine**
+  Si joogto ah ugu dar, u laameey (branch), una kobci xusuusta qeybo kala duwan.
+
+-   **Capsule Context (`.mv2`)**
+  Capsule xusuuseed oo isku-filan, la wadaagi karo, lehna sharciyo iyo waqti dhicitaan.
+
+-   **Time-Travel Debugging**
+  Dib u celi, ama qabeey xaalad kasta oo xusuusta ah.
+
+-   **Smart Recall**
+    Soo-celinta xusuusta gudaha wax ka yar 5ms iyadoo la adeegsanayo kaydinta saadaalinta (predictive caching).
+
+-   **Codec Intelligence**
+    Si otomaatig ah u doorta una casriyeeya isku-duubnida (compression) waqtiga ka dib.
+
+---
+
+## Meelaha loo adeegsado (Use Cases)
+
+Memvid waa nidaam xusuuseed oo la qaadan karo oo aan server u baahnayn, kaas oo siiya wakiillada AI(AI agents), xusuus joogto ah iyo soo-celin degdeg ah. Maadaama uu ka madax-bannaan yahay modelka, waxna ku akhriyo qaabab badan (multi-modal), una shaqeeyo si buuxda isagoo aan internet lahayn, horumariyayaashu waxay Memvid u adeegsanayaan hawlo badan:
+
+-   Wakiillada AI(AI Agents) ee muddada dheer shaqeeya 
+-   Keydka aqoonta ee shirkadaha
+-   Tageerida AI-ga ee ku shaqeeya offline-ka
+-   Fahamka nidaamyada koodhka (Codebase)
+-   Wakiillada adeegga macmiilka
+-   Otomaatigga shaqada (Workflow Automation)
+-   Kaaliyayaasha iibka iyo suuqgeynta
+-   Kaaliyayaasha aqoonta shakhsi ahaaneed
+-   Wakiillada caafimaadka, sharciga, iyo maaliyadda
+-   Hannaanka shaqada AI-ga oo la baari karo lana saxi karo (Auditable/Debuggable)
+-   Codsiyada gaarka ah (Custom Applications)
+
+---
+
+## SDKs & CLI
+
+Ku isticmaal Memvid luuqadda aad doorbidi lahayd:
+
+| Package         | Install                     | Links                                                                                                               |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI**         | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli)       |
+| **Node.js SDK** | `npm install @memvid/sdk`   | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk)     |
+| **Python SDK**  | `pip install memvid-sdk`    | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/)         |
+| **Rust**        | `cargo add memvid-core`     | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## Kushubashada (Installation) (Rust)
+
+### Shuruudaha
+
+-   **Rust 1.85.0+** — Si aad ugu shubato Guji linkagan [rustup.rs](https://rustup.rs)
+
+### Ku dar Mashruucaaga
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### Feature Flags
+
+| Feature             | Sharaxada                                    |
+| ------------------- | ---------------------------------------------- |
+| `lex`               | Raadinta qoraalka oo dhan oo leh darajada BM25 (Tantivy)   |
+| `pdf_extract`       | Soo saarista qoraalka PDF oo saafi ah                  |
+| `vec`               | Raadinta isku-midka ah ee Vector (HNSW + ONNX)         |
+| `clip`              | CLIP visual embeddings oo loogu talagalay raadinta sawirka        |
+| `whisper`           | Beddelka codka iyadoo loo baddelayo qoraal lana adeegsanayo Whisper               |
+| `temporal_track`    | Turjumidda taariikhda ee luuqadda caadiga ah ("Salaasadii hore") |
+| `parallel_segments` | Soo gelinta xogta iyadoo la adeegsanayo dhowr nuuc (Multi-threaded)                       |
+| `encryption`        | capsules-ka xusuusta ee ku xidhan sirta (password) (.mv2e)     |
+
+U furo sifooyinka (features) sida aad ugu baahan tahay:
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+---
+
+## Bilow Degdeg ah (Quick Start)
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // Create a new memory file
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // Add documents with metadata
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // Search
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## Dhisid (Build)
+
+Qeyb ka soo qaado (Clone) kaydka (Repository):
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+U dhis habka cilad-baadhista (debug mode):
+
+```bash
+cargo build
+```
+
+U dhis habka rasmiga ah (release mode - la hagaajiyay):
+
+```bash
+cargo build --release
+```
+
+Ku dhis sifooyin (features) gaar ah:
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## Tijaabi iskudayga (Run Tests)
+
+Tijaabi iskudayada oo dhan:
+
+```bash
+cargo test
+```
+
+Tijaabi iskudayga iyadoo natiijada la arkayo:
+
+```bash
+cargo test -- --nocapture
+```
+
+Tijaab iskuday gaar ah:
+
+```bash
+cargo test test_name
+```
+
+Tijaabi iskudayada isku-xirka (integration tests) ah oo keliya
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## Tusaalooyin (Examples)
+
+Tusaha `examples/` wuxuu ka kooban yahay tusaalooyin shaqaynaya:
+
+### Adeegsiga Fudud 
+
+Wuxuu muujinayaa samaynta, gelinta, raadinta, iyo hawlgallada waqtiga (timeline):
+
+```bash
+cargo run --example basic_usage
+```
+
+### Soo gelinta PDF
+
+Geli oo baadh dukumiintiyada PDF-ka ah (wuxuu isticmaalaa warqaddii aheyd "Attention Is All You Need"):
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### Raadinta Muuqaalka ee CLIP
+
+Raadinta sawirka iyadoo la adeegsanayo CLIP (waxay u baahan tahay clip feature):
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Beddelka Codka ee Whisper
+
+Beddelka codka (waxay u baahan tahay whisper feature):
+
+```bash
+cargo run --example test_whisper --features whisper
+```
+
+---
+
+## Qaabka Faylka (File Format)
+
+Wax walba waxay ku jiraan hal fayl oo .mv2 ah:
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  Magic, nooca, awoodda
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  Kasoo kabashada burburka
+├────────────────────────────┤
+│ Data Segments              │  Frames la isku-duubay
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy qoraal ka buuxo
+├────────────────────────────┤
+│ Vec Index                  │  HNSW vectors
+├────────────────────────────┤
+│ Time Index                 │  Siday u kala horreeyaan
+├────────────────────────────┤
+│ TOC (Footer)               │  Meeqaamka qaybaha
+└────────────────────────────┘
+```
+
+Ma jiraan faylal .wal, .lock, .shm, ama faylal dhinac socda. Weligaa.
+
+Fiiri [MV2_SPEC.md](MV2_SPEC.md) si aad u hesho faahfaahinta dhammaystiran ee qaabka faylka.
+
+---
+
+## Taageer (Support)
+
+Ma qabtaa su'aalo ama ra'yi?
+Email: contact@memvid.com
+
+**Noo saar ⭐ si aad u muujiso taageeradaada**
+
+---
+
+## Shatiga (License)
+
+Apache License 2.0 — Fiiri faylka [LICENSE](LICENSE) si aad u hesho faahfaahin dheeraad ah.

docs/i18n/README.zh-CN.md

@@ -0,0 +1,517 @@
+<!-- HEADER:START -->
+<img width="2000" height="524" alt="Social Cover (9)"
+     src="https://github.com/user-attachments/assets/cf66f045-c8be-494b-b696-b8d7e4fb709c" />
+<!-- HEADER:END -->
+
+<div style="height: 16px;"></div>
+
+<p align="center">
+    <a href="https://trendshift.io/repositories/17293" target="_blank"><img src="https://trendshift.io/api/badge/repositories/17293" alt="memvid%2Fmemvid | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+<!-- BADGES:END -->
+
+<p align="center">
+  <strong>Memvid 是专为 AI 智能体设计的单文件记忆层，具备即时检索和长期记忆能力。</strong><br/>
+  持久化、版本化、可移植的记忆，无需数据库。
+</p>
+
+<!-- NAV:START -->
+<p align="center">
+  <a href="https://www.memvid.com">官方网站</a>
+  ·
+  <a href="https://sandbox.memvid.com">尝试一下沙箱</a>
+  ·
+  <a href="https://docs.memvid.com">文档</a>
+  ·
+  <a href="https://github.com/memvid/memvid/discussions">讨论区</a>
+</p>
+<!-- NAV:END -->
+
+<!-- BADGES:START -->
+<p align="center">
+  <a href="https://crates.io/crates/memvid-core"><img src="https://img.shields.io/crates/v/memvid-core?style=flat-square&logo=rust" alt="Crates.io" /></a>
+  <a href="https://docs.rs/memvid-core"><img src="https://img.shields.io/docsrs/memvid-core?style=flat-square&logo=docs.rs" alt="docs.rs" /></a>
+  <a href="https://github.com/memvid/memvid/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue?style=flat-square" alt="许可证" /></a>
+</p>
+
+<p align="center">
+  <a href="https://github.com/memvid/memvid/stargazers"><img src="https://img.shields.io/github/stars/memvid/memvid?style=flat-square&logo=github" alt="Stars" /></a>
+  <a href="https://github.com/memvid/memvid/network/members"><img src="https://img.shields.io/github/forks/memvid/memvid?style=flat-square&logo=github" alt="Forks" /></a>
+  <a href="https://github.com/memvid/memvid/issues"><img src="https://img.shields.io/github/issues/memvid/memvid?style=flat-square&logo=github" alt="Issues" /></a>
+  <a href="https://discord.gg/2mynS7fcK7"><img src="https://img.shields.io/discord/1442910055233224745?style=flat-square&logo=discord&label=discord" alt="Discord" /></a>
+</p>
+
+
+
+<h2 align="center">⭐️ 给项目点个星标支持我们 ⭐️</h2>
+
+## 基准测试亮点
+
+**🚀 准确率超越其他记忆系统：** 在 LoCoMo 上领先 SOTA 35%，长时对话回忆与推理能力最佳
+
+**🧠 卓越的多跳与时序推理：** 比行业平均水平高出 76% 多跳推理，56% 时序推理
+
+**⚡ 超低延迟高吞吐：** P50 仅 0.025ms，P99 仅 0.075ms，吞吐量是标准的 1,372 倍
+
+**🔬 完全可复现的基准测试：** LoCoMo（10 次约 26K token 的对话）、开源评估、LLM-as-Judge
+
+
+## 什么是 Memvid？
+
+Memvid 是可移植的 AI 记忆系统，将数据、嵌入向量、搜索结构和元数据打包成单个文件。
+
+无需运行复杂的 RAG 管道或基于服务器的向量数据库，Memvid 支持直接从文件进行快速检索。
+
+结果是模型无关、无基础设施的记忆层，让 AI 智能体拥有可随身携带的持久化长期记忆。
+
+     
+## 什么是 Smart Frames？
+
+Memvid 借鉴视频编码的理念，不是为了存储视频，而是**将 AI 记忆组织为仅追加、超高效序列的 Smart Frames。**
+
+Smart Frame 是存储内容以及时间戳、校验和和基本元数据的不可变单元。
+帧以允许高效压缩、索引和并行读取的方式分组。
+
+这种基于帧的设计支持：
+
+-   仅追加写入，不修改或破坏现有数据
+-   对过去记忆状态的查询
+-   知识演化的时间轴式检查
+-   通过基于提交的不可变帧应对崩溃
+-   使用基于视频编码技术的高效压缩
+
+结果是一个表现为 AI 系统可追溯记忆时间线的单文件。
+
+
+## 核心概念
+
+-   **Living Memory Engine**
+    持续追加、分支和跨会话演进记忆。
+
+-   **Capsule Context (`.mv2`)**
+    自包含、可共享的记忆胶囊，带规则和过期时间。
+
+-   **Time-Travel Debugging**
+    回溯、重放或分支化任何记忆状态。
+
+-   **Smart Recall**
+    小于 5ms 本地记忆访问，具备预测性缓存。
+
+-   **Codec Intelligence**
+    随时间自动选择和升级压缩。
+
+
+## 使用场景
+
+Memvid 是无服务器便携记忆层，为 AI 智能体提供持久记忆和快速召回。由于它是模型无关、多模态且完全离线工作，开发者正在各种现实应用中广泛的使用 Memvid。
+
+-   长期运行的 AI 智能体
+-   企业知识库
+-   离线优先 AI 系统
+-   代码库理解
+-   客户支持智能体
+-   工作流自动化
+-   销售与营销助手
+-   个人知识助理
+-   医疗、法律和金融智能体
+-   可审计和可调试的 AI 工作流
+-   自定义应用
+
+
+## SDKs 与 CLI
+
+在您喜欢的语言中使用 Memvid：
+
+| 包 | 安装 | 链接 |
+| --------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| **CLI** | `npm install -g memvid-cli` | [![npm](https://img.shields.io/npm/v/memvid-cli?style=flat-square)](https://www.npmjs.com/package/memvid-cli) |
+| **Node.js SDK** | `npm install @memvid/sdk` | [![npm](https://img.shields.io/npm/v/@memvid/sdk?style=flat-square)](https://www.npmjs.com/package/@memvid/sdk) |
+| **Python SDK** | `pip install memvid-sdk` | [![PyPI](https://img.shields.io/pypi/v/memvid-sdk?style=flat-square)](https://pypi.org/project/memvid-sdk/) |
+| **Rust** | `cargo add memvid-core` | [![Crates.io](https://img.shields.io/crates/v/memvid-core?style=flat-square)](https://crates.io/crates/memvid-core) |
+
+---
+
+## 安装（Rust）
+
+### 要求
+
+-   **Rust 1.85.0+** — 从 [rustup.rs](https://rustup.rs) 安装
+
+### 添加到项目
+
+```toml
+[dependencies]
+memvid-core = "2.0"
+```
+
+### 功能标志
+
+| 功能 | 描述 |
+| -------------------- | ---------------------------------------------------------------- |
+| `lex` | 使用 BM25 排序的全文搜索（Tantivy） |
+| `pdf_extract` | 纯 Rust PDF 文本提取 |
+| `vec` | 向量相似搜索（HNSW + 通过 ONNX 的本地文本嵌入） |
+| `clip` | CLIP 视觉嵌入用于图像搜索 |
+| `whisper` | 使用 Whisper 进行音频转录 |
+| `api_embed` | 云 API 嵌入（OpenAI） |
+| `temporal_track` | 自然语言日期解析（"last Tuesday"） |
+| `parallel_segments` | 多线程摄取 |
+| `encryption` | 基于密码的加密胶囊（.mv2e） |
+| `symspell_cleanup` | 强大的 PDF 文本修复（修复 "emp lo yee" -> "employee"） |
+
+按需启用功能：
+
+```toml
+[dependencies]
+memvid-core = { version = "2.0", features = ["lex", "vec", "temporal_track"] }
+```
+
+
+## 快速开始
+
+```rust
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    // 创建新记忆文件
+    let mut mem = Memvid::create("knowledge.mv2")?;
+
+    // 添加带元数据的文档
+    let opts = PutOptions::builder()
+        .title("Meeting Notes")
+        .uri("mv2://meetings/2024-01-15")
+        .tag("project", "alpha")
+        .build();
+    mem.put_bytes_with_options(b"Q4 planning discussion...", opts)?;
+    mem.commit()?;
+
+    // 搜索
+    let response = mem.search(SearchRequest {
+        query: "planning".into(),
+        top_k: 10,
+        snippet_chars: 200,
+        ..Default::default()
+    })?;
+
+    for hit in response.hits {
+        println!("{}: {}", hit.title.unwrap_or_default(), hit.text);
+    }
+
+    Ok(())
+}
+```
+
+---
+
+## 构建
+
+克隆仓库：
+
+```bash
+git clone https://github.com/memvid/memvid.git
+cd memvid
+```
+
+以调试模式构建：
+
+```bash
+cargo build
+```
+
+以发布模式构建（优化）：
+
+```bash
+cargo build --release
+```
+
+使用特定功能构建：
+
+```bash
+cargo build --release --features "lex,vec,temporal_track"
+```
+
+---
+
+## 运行测试
+
+运行所有测试：
+
+```bash
+cargo test
+```
+
+带输出运行测试：
+
+```bash
+cargo test -- --nocapture
+```
+
+运行特定测试：
+
+```bash
+cargo test test_name
+```
+
+仅运行集成测试：
+
+```bash
+cargo test --test lifecycle
+cargo test --test search
+cargo test --test mutation
+```
+
+---
+
+## 示例
+
+`examples/` 目录包含可运行示例：
+
+### 基本用法
+
+演示创建、添加、搜索和时间线操作：
+
+```bash
+cargo run --example basic_usage
+```
+
+### PDF 提取
+
+提取和搜索 PDF 文档（使用 "Attention Is All You Need" 论文）：
+
+```bash
+cargo run --example pdf_ingestion
+```
+
+### CLIP 可视化搜索
+
+使用 CLIP 嵌入进行图像搜索（需要 `clip` 功能）：
+
+```bash
+cargo run --example clip_visual_search --features clip
+```
+
+### Whisper 转录
+
+音频转录（需要 `whisper` 功能）：
+
+```bash
+cargo run --example test_whisper --features whisper -- /path/to/audio.mp3
+```
+
+**可用模型：**
+
+| 模型 | 大小 | 速度 | 用例 |
+| ---------------------- | ------ | ------- | ----------------------------------- |
+| `whisper-small-en` | 244 MB | 最慢 | 最佳准确度（默认） |
+| `whisper-tiny-en` | 75 MB | 快 | 平衡 |
+| `whisper-tiny-en-q8k` | 19 MB | 最快 | 快速测试，资源受限 |
+
+**模型选择：**
+
+```bash
+# 默认（FP32 small，最高准确度）
+cargo run --example test_whisper --features whisper -- audio.mp3
+
+# 小型量化（小 75%，更快）
+MEMVID_WHISPER_MODEL=whisper-tiny-en-q8k cargo run --example test_whisper --features whisper -- audio.mp3
+```
+
+**可编程配置：**
+
+```rust
+use memvid_core::{WhisperConfig, WhisperTranscriber};
+
+// 默认 FP32 small 模型
+let config = WhisperConfig::default();
+
+// 小型量化模型（更快，更小）
+let config = WhisperConfig::with_quantization();
+
+// 特定模型
+let config = WhisperConfig::with_model("whisper-tiny-en-q8k");
+
+let transcriber = WhisperTranscriber::new(&config)?;
+let result = transcriber.transcribe_file("audio.mp3")?;
+println!("{}", result.text);
+```
+
+
+## 文本嵌入模型
+
+`vec` 功能包括使用 ONNX 模型的本地文本嵌入支持。在使用本地文本嵌入之前，您需要手动下载模型文件。
+
+### 快速开始：BGE-small（推荐）
+
+下载默认 BGE-small 模型（384 维，快速高效）：
+
+```bash
+mkdir -p ~/.cache/memvid/text-models
+
+# 下载 ONNX 模型
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5.onnx
+
+# 下载分词器
+curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-small-en-v1.5_tokenizer.json
+```
+
+### 可用模型
+
+| 模型 | 维度 | 大小 | 最适合 |
+| ------------------------ | ---------- | ------ | --------------- |
+| `bge-small-en-v1.5` | 384 | ~120MB | 默认，快速 |
+| `bge-base-en-v1.5` | 768 | ~420MB | 更好的质量 |
+| `nomic-embed-text-v1.5` | 768 | ~530MB | 多用途任务 |
+| `gte-large` | 1024 | ~1.3GB | 最高质量 |
+
+### 其他模型
+
+**BGE-base**（768 维）：
+```bash
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5.onnx
+curl -L 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/bge-base-en-v1.5_tokenizer.json
+```
+
+**Nomic**（768 维）：
+```bash
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5.onnx
+curl -L 'https://huggingface.co/nomic-ai/nomic-embed-text-v1.5/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/nomic-embed-text-v1.5_tokenizer.json
+```
+
+**GTE-large**（1024 维）：
+```bash
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/onnx/model.onnx' \
+  -o ~/.cache/memvid/text-models/gte-large.onnx
+curl -L 'https://huggingface.co/thenlper/gte-large/resolve/main/tokenizer.json' \
+  -o ~/.cache/memvid/text-models/gte-large_tokenizer.json
+```
+
+### 在代码中使用
+
+```rust
+use memvid_core::text_embed::{LocalTextEmbedder, TextEmbedConfig};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// 使用默认模型（BGE-small）
+let config = TextEmbedConfig::default();
+let embedder = LocalTextEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 384);
+
+// 使用不同模型
+let config = TextEmbedConfig::bge_base();
+let embedder = LocalTextEmbedder::new(config)?;
+```
+
+有关相似度计算和搜索排名的完整示例，请参见 `examples/text_embedding.rs`。
+
+### 模型一致性
+
+为防止意外地模型混合（例如，使用 OpenAI 嵌入查询 BGE-small 索引），您可以将 Memvid 实例显式绑定到特定模型名称：
+
+```rust
+// 将索引绑定到特定模型。
+// 如果之前使用不同模型创建索引将返回错误。
+mem.set_vec_model("bge-small-en-v1.5")?;
+```
+
+绑定是持久化的。一旦设置，将来尝试使用不同模型名称将快速失败并返回 `ModelMismatch` 错误。
+
+
+
+## API 嵌入（OpenAI）
+
+`api_embed` 功能使用 OpenAI 的 API 启用基于云的嵌入生成。
+
+### 设置
+
+设置您的 OpenAI API 密钥：
+
+```bash
+export OPENAI_API_KEY="sk-..."
+```
+
+### 用法
+
+```rust
+use memvid_core::api_embed::{OpenAIConfig, OpenAIEmbedder};
+use memvid_core::types::embedding::EmbeddingProvider;
+
+// 使用默认模型（text-embedding-3-small）
+let config = OpenAIConfig::default();
+let embedder = OpenAIEmbedder::new(config)?;
+
+let embedding = embedder.embed_text("hello world")?;
+assert_eq!(embedding.len(), 1536);
+
+// 使用更高质量模型
+let config = OpenAIConfig::large();  // text-embedding-3-large (3072 维)
+let embedder = OpenAIEmbedder::new(config)?;
+```
+
+### 可用模型
+
+| 模型 | 维度 | 最适合 |
+| ------------------------ | ---------- | -------------------------- |
+| `text-embedding-3-small` | 1536 | 默认，最快，最便宜 |
+| `text-embedding-3-large` | 3072 | 最高质量 |
+| `text-embedding-ada-002` | 1536 | 传统模型 |
+
+有关完整示例，请参见 `examples/openai_embedding.rs`。
+
+
+
+## 文件格式
+
+所有内容都存储在单个 `.mv2` 文件中：
+
+```
+┌────────────────────────────┐
+│ Header (4KB)               │  魔数，版本，容量
+├────────────────────────────┤
+│ Embedded WAL (1-64MB)      │  崩溃恢复
+├────────────────────────────┤
+│ Data Segments              │  压缩帧
+├────────────────────────────┤
+│ Lex Index                  │  Tantivy 全文
+├────────────────────────────┤
+│ Vec Index                  │  HNSW 向量
+├────────────────────────────┤
+│ Time Index                 │  时序排序
+├────────────────────────────┤
+│ TOC (Footer)               │  段偏移
+└────────────────────────────┘
+```
+
+不会有 `.wal`、`.lock`、`.shm` 或附带文件。永远不会。
+
+有关完整文件格式规范，请参见 [MV2_SPEC.md](MV2_SPEC.md)。
+
+
+
+## 支持
+
+有问题或反馈？
+邮箱：contact@memvid.com
+
+**点个 ⭐ 支持我们**
+
+---
+
+> **Memvid v1（基于 QR 码的记忆）已弃用**
+>
+> 如果您参考的是 QR 码，那么您正在使用过时信息。
+>
+> 参见：https://docs.memvid.com/memvid-v1-deprecation
+
+---
+
+## 许可证
+
+Apache License 2.0 — 详细信息请参见 [LICENSE](LICENSE) 文件。

docs/i18n/scripts/auto-add-flags.js

@@ -0,0 +1,249 @@
+const fs = require('fs');
+const path = require('path');
+
+const I18N_DIR = path.join(__dirname, '..');
+const ROOT_DIR = path.join(I18N_DIR, '..', '..');
+const README_PATH = path.join(ROOT_DIR, 'README.md');
+
+const LANG_MAP = {
+    'aa': { emoji: '🌐', name: 'Afar' },
+    'ab': { emoji: '🌐', name: 'Abkhazian' },
+    'ae': { emoji: '🌐', name: 'Avestan' },
+    'af': { emoji: '🇿🇦', name: 'Afrikaans' },
+    'ak': { emoji: '🌐', name: 'Akan' },
+    'am': { emoji: '🇪🇹', name: 'Amharic' },
+    'an': { emoji: '🌐', name: 'Aragonese' },
+    'ar': { emoji: '🇸🇦', name: 'العربية' },
+    'as': { emoji: '🌐', name: 'Assamese' },
+    'av': { emoji: '🌐', name: 'Avaric' },
+    'ay': { emoji: '🌐', name: 'Aymara' },
+    'az': { emoji: '🇦🇿', name: 'Azerbaijani' },
+    'ba': { emoji: '🌐', name: 'Bashkir' },
+    'be': { emoji: '🇧🇾', name: 'Belarusian' },
+    'bg': { emoji: '🇧🇬', name: 'Bulgarian' },
+    'bi': { emoji: '🌐', name: 'Bislama' },
+    'bm': { emoji: '🌐', name: 'Bambara' },
+    'bn': { emoji: '🇧🇩', name: 'বাংলা' },
+    'bo': { emoji: '🌐', name: 'Tibetan' },
+    'br': { emoji: '🌐', name: 'Breton' },
+    'bs': { emoji: '🇧🇦', name: 'Bosnian' },
+    'ca': { emoji: '🇪🇸', name: 'Catalan' },
+    'ce': { emoji: '🌐', name: 'Chechen' },
+    'ch': { emoji: '🌐', name: 'Chamorro' },
+    'co': { emoji: '🌐', name: 'Corsican' },
+    'cr': { emoji: '🌐', name: 'Cree' },
+    'cs': { emoji: '🇨🇿', name: 'Česko' },
+    'cu': { emoji: '🌐', name: 'Church Slavonic' },
+    'cv': { emoji: '🌐', name: 'Chuvash' },
+    'cy': { emoji: '🇬🇧', name: 'Welsh' },
+    'da': { emoji: '🇩🇰', name: 'Danish' },
+    'de': { emoji: '🇩🇪', name: 'Deutsch' },
+    'dv': { emoji: '🌐', name: 'Divehi' },
+    'dz': { emoji: '🇧🇹', name: 'Dzongkha' },
+    'ee': { emoji: '🌐', name: 'Ewe' },
+    'el': { emoji: '🇬🇷', name: 'Greek' },
+    'en': { emoji: '🇺🇸', name: 'English' },
+    'eo': { emoji: '🌐', name: 'Esperanto' },
+    'es': { emoji: '🇪🇸', name: 'Español' },
+    'et': { emoji: '🇪🇪', name: 'Estonian' },
+    'eu': { emoji: '🌐', name: 'Basque' },
+    'fa': { emoji: '🇮🇷', name: 'Persian' },
+    'ff': { emoji: '🌐', name: 'Fulah' },
+    'fi': { emoji: '🇫🇮', name: 'Finnish' },
+    'fj': { emoji: '🌐', name: 'Fijian' },
+    'fo': { emoji: '🌐', name: 'Faroese' },
+    'fr': { emoji: '🇫🇷', name: 'Français' },
+    'fy': { emoji: '🌐', name: 'Western Frisian' },
+    'ga': { emoji: '🇮🇪', name: 'Irish' },
+    'gd': { emoji: '🌐', name: 'Gaelic' },
+    'gl': { emoji: '🌐', name: 'Galician' },
+    'gn': { emoji: '🌐', name: 'Guarani' },
+    'gu': { emoji: '🌐', name: 'Gujarati' },
+    'gv': { emoji: '🌐', name: 'Manx' },
+    'ha': { emoji: '🇳🇬', name: 'Hausa' },
+    'he': { emoji: '🇮🇱', name: 'Hebrew' },
+    'hi': { emoji: '🇮🇳', name: 'हिन्दी' },
+    'ho': { emoji: '🌐', name: 'Hiri Motu' },
+    'hr': { emoji: '🇭🇷', name: 'Croatian' },
+    'ht': { emoji: '🌐', name: 'Haitian' },
+    'hu': { emoji: '🇭🇺', name: 'Hungarian' },
+    'hy': { emoji: '🇦🇲', name: 'Armenian' },
+    'hz': { emoji: '🌐', name: 'Herero' },
+    'ia': { emoji: '🌐', name: 'Interlingua' },
+    'id': { emoji: '🇮🇩', name: 'Bahasa' },
+    'ie': { emoji: '🌐', name: 'Interlingue' },
+    'ig': { emoji: '🇳🇬', name: 'Igbo' },
+    'ii': { emoji: '🌐', name: 'Sichuan Yi' },
+    'ik': { emoji: '🌐', name: 'Inupiaq' },
+    'io': { emoji: '🌐', name: 'Ido' },
+    'is': { emoji: '🇮🇸', name: 'Icelandic' },
+    'it': { emoji: '🇮🇹', name: 'Italiano' },
+    'iu': { emoji: '🌐', name: 'Inuktitut' },
+    'ja': { emoji: '🇯🇵', name: '日本語' },
+    'jv': { emoji: '🌐', name: 'Javanese' },
+    'ka': { emoji: '🇬🇪', name: 'Georgian' },
+    'kg': { emoji: '🌐', name: 'Kongo' },
+    'ki': { emoji: '🌐', name: 'Kikuyu' },
+    'kj': { emoji: '🌐', name: 'Kuanyama' },
+    'kk': { emoji: '🇰🇿', name: 'Kazakh' },
+    'kl': { emoji: '🌐', name: 'Kalaallisut' },
+    'km': { emoji: '🇰🇭', name: 'Central Khmer' },
+    'kn': { emoji: '🌐', name: 'Kannada' },
+    'ko': { emoji: '🇰🇷', name: '한국어' },
+    'kr': { emoji: '🌐', name: 'Kanuri' },
+    'ks': { emoji: '🌐', name: 'Kashmiri' },
+    'ku': { emoji: '🇮🇶', name: 'Kurdish' },
+    'kv': { emoji: '🌐', name: 'Komi' },
+    'kw': { emoji: '🌐', name: 'Cornish' },
+    'ky': { emoji: '🇰🇬', name: 'Kyrgyz' },
+    'la': { emoji: '🌐', name: 'Latin' },
+    'lb': { emoji: '🌐', name: 'Luxembourgish' },
+    'lg': { emoji: '🌐', name: 'Ganda' },
+    'li': { emoji: '🌐', name: 'Limburgan' },
+    'ln': { emoji: '🌐', name: 'Lingala' },
+    'lo': { emoji: '🇱🇦', name: 'Lao' },
+    'lt': { emoji: '🇱🇹', name: 'Lithuanian' },
+    'lu': { emoji: '🌐', name: 'Luba-Katanga' },
+    'lv': { emoji: '🇱🇻', name: 'Latvian' },
+    'mg': { emoji: '🌐', name: 'Malagasy' },
+    'mh': { emoji: '🌐', name: 'Marshallese' },
+    'mi': { emoji: '🌐', name: 'Maori' },
+    'mk': { emoji: '🇲🇰', name: 'Macedonian' },
+    'ml': { emoji: '🌐', name: 'Malayalam' },
+    'mn': { emoji: '🇲🇳', name: 'Mongolian' },
+    'mr': { emoji: '🌐', name: 'Marathi' },
+    'ms': { emoji: '🇲🇾', name: 'Malay' },
+    'mt': { emoji: '🇲🇹', name: 'Maltese' },
+    'my': { emoji: '🇲🇲', name: 'Burmese' },
+    'na': { emoji: '🌐', name: 'Nauru' },
+    'nb': { emoji: '🌐', name: 'Norwegian Bokmål' },
+    'nd': { emoji: '🌐', name: 'North Ndebele' },
+    'ne': { emoji: '🇳🇵', name: 'Nepali' },
+    'ng': { emoji: '🌐', name: 'Ndonga' },
+    'nl': { emoji: '🇧🇪/🇳🇱', name: 'Nederlands' },
+    'nn': { emoji: '🌐', name: 'Norwegian Nynorsk' },
+    'no': { emoji: '🇳🇴', name: 'Norwegian' },
+    'nr': { emoji: '🌐', name: 'South Ndebele' },
+    'nv': { emoji: '🌐', name: 'Navajo' },
+    'ny': { emoji: '🌐', name: 'Chichewa' },
+    'oc': { emoji: '🌐', name: 'Occitan' },
+    'oj': { emoji: '🌐', name: 'Ojibwa' },
+    'om': { emoji: '🌐', name: 'Oromo' },
+    'or': { emoji: '🌐', name: 'Oriya' },
+    'os': { emoji: '🌐', name: 'Ossetian' },
+    'pa': { emoji: '🌐', name: 'Punjabi' },
+    'pi': { emoji: '🌐', name: 'Pali' },
+    'pl': { emoji: '🇵��', name: 'Polski' },
+    'ps': { emoji: '🇦🇫', name: 'Pashto' },
+    'pt': { emoji: '🇵🇹', name: 'Português' },
+    'qu': { emoji: '🌐', name: 'Quechua' },
+    'rm': { emoji: '🌐', name: 'Romansh' },
+    'rn': { emoji: '🌐', name: 'Rundi' },
+    'ro': { emoji: '🇷🇴', name: 'Romanian' },
+    'ru': { emoji: '🇷🇺', name: 'Русский' },
+    'rw': { emoji: '🌐', name: 'Kinyarwanda' },
+    'sa': { emoji: '🌐', name: 'Sanskrit' },
+    'sc': { emoji: '🌐', name: 'Sardinian' },
+    'sd': { emoji: '🌐', name: 'Sindhi' },
+    'se': { emoji: '🌐', name: 'Northern Sami' },
+    'sg': { emoji: '🌐', name: 'Sango' },
+    'si': { emoji: '🇱🇰', name: 'Sinhala' },
+    'sk': { emoji: '🇸🇰', name: 'Slovak' },
+    'sl': { emoji: '🇸🇮', name: 'Slovenian' },
+    'sm': { emoji: '🌐', name: 'Samoan' },
+    'sn': { emoji: '🌐', name: 'Shona' },
+    'so': { emoji: '🇸🇴', name: 'Soomaali' },
+    'sq': { emoji: '🇦🇱', name: 'Albanian' },
+    'sr': { emoji: '🇷🇸', name: 'Serbian' },
+    'ss': { emoji: '🌐', name: 'Swati' },
+    'st': { emoji: '🇿🇦', name: 'Southern Sotho' },
+    'su': { emoji: '🌐', name: 'Sundanese' },
+    'sv': { emoji: '🇸🇪', name: 'Swedish' },
+    'sw': { emoji: '🇰🇪', name: 'Swahili' },
+    'ta': { emoji: '🌐', name: 'Tamil' },
+    'te': { emoji: '🌐', name: 'Telugu' },
+    'tg': { emoji: '🇹🇯', name: 'Tajik' },
+    'th': { emoji: '🇹🇭', name: 'Thai' },
+    'ti': { emoji: '🌐', name: 'Tigrinya' },
+    'tk': { emoji: '🇹🇲', name: 'Turkmen' },
+    'tl': { emoji: '🇵🇭', name: 'Tagalog' },
+    'tn': { emoji: '🌐', name: 'Tswana' },
+    'to': { emoji: '🌐', name: 'Tonga' },
+    'tr': { emoji: '🇹🇷', name: 'Türkçe' },
+    'ts': { emoji: '🌐', name: 'Tsonga' },
+    'tt': { emoji: '🌐', name: 'Tatar' },
+    'tw': { emoji: '🌐', name: 'Twi' },
+    'ty': { emoji: '🌐', name: 'Tahitian' },
+    'ug': { emoji: '🌐', name: 'Uighur' },
+    'uk': { emoji: '🇺🇦', name: 'Ukrainian' },
+    'ur': { emoji: '🇵🇰', name: 'Urdu' },
+    'uz': { emoji: '🇺🇿', name: 'Uzbek' },
+    've': { emoji: '🌐', name: 'Venda' },
+    'vi': { emoji: '🇻🇳', name: 'Tiếng Việt' },
+    'vo': { emoji: '🌐', name: 'Volapük' },
+    'wa': { emoji: '🌐', name: 'Walloon' },
+    'wo': { emoji: '🌐', name: 'Wolof' },
+    'xh': { emoji: '🇿🇦', name: 'Xhosa' },
+    'yi': { emoji: '🌐', name: 'Yiddish' },
+    'yo': { emoji: '🇳🇬', name: 'Yoruba' },
+    'za': { emoji: '🌐', name: 'Zhuang' },
+    'zh': { emoji: '🇨🇳', name: '中文' },
+    'zh-CN': { emoji: '🇨🇳', name: '中文 (简体)' },
+    'zh-HK': { emoji: '🇭🇰', name: '中文 (繁體)' },
+    'zh-Hans': { emoji: '🇨🇳', name: '中文 (简体)' },
+    'zh-Hant': { emoji: '🇹🇼', name: '中文 (繁體)' },
+    'zh-MO': { emoji: '🇲🇴', name: '中文 (繁體)' },
+    'zh-SG': { emoji: '🇸🇬', name: '中文 (繁體)' },
+    'zh-TW': { emoji: '🇹🇼', name: '中文 (繁體)' },
+    'zu': { emoji: '🇿🇦', name: 'Zulu' },
+};
+
+function autoAddFlags() {
+    if (!fs.existsSync(README_PATH)) {
+        console.error('Error: Cannot find ' + README_PATH);
+        process.exit(1);
+    }
+
+    let readmeContent = fs.readFileSync(README_PATH, 'utf-8');
+    const marker = ' <!-- Next Flag -->';
+
+    if (!readmeContent.includes(marker)) {
+        console.warn('Error: <!-- Next Flag --> marker not found in ' + README_PATH);
+        console.warn('Please add the marker where you want new flags to be inserted.');
+        return;
+    }
+
+    const files = fs.readdirSync(I18N_DIR);
+
+    const translationFiles = files.filter(f =>
+        f.startsWith('README') &&
+        f.endsWith('.md') &&
+        f !== 'README.md'
+    );
+
+    let updated = false;
+
+    translationFiles.forEach(file => {
+        const code = file.split('.')[1];
+        const lang = LANG_MAP[code];
+
+        if (!lang) return;
+
+        const flagLink = ` <a href="docs/i18n/${file}">${lang.emoji} ${lang.name}</a>`;
+
+        if (!readmeContent.includes(file)) {
+            readmeContent = readmeContent.replace(marker, flagLink + '\n' + marker);
+            updated = true;
+            console.log('Added ' + code);
+        }
+    });
+
+    if (updated) {
+        fs.writeFileSync(README_PATH, readmeContent, 'utf-8');
+        console.log('Main README updated.');
+    } else {
+        console.log('No new flags to add.');
+    }
+}
+
+autoAddFlags();

docs/i18n/scripts/update-localized-readmes.js

@@ -0,0 +1,114 @@
+const fs = require('fs');
+const path = require('path');
+
+const I18N_DIR = path.join(__dirname, '..');
+const ROOT_DIR = path.join(I18N_DIR, '..', '..');
+const README_PATH = path.join(ROOT_DIR, 'README.md');
+
+const MARKERS = {
+    HEADER: {
+        START: '<!-- HEADER:START -->',
+        END: '<!-- HEADER:END -->'
+    },
+    FLAGS: {
+        START: '<!-- FLAGS:START -->',
+        END: '<!-- FLAGS:END -->'
+    },
+    NAV: {
+        START: '<!-- NAV:START -->',
+        END: '<!-- NAV:END -->'
+    },
+    BADGES: {
+        START: '<!-- BADGES:START -->',
+        END: '<!-- BADGES:END -->'
+    }
+};
+
+function extractBlock(content, marker) {
+    const startIdx = content.indexOf(marker.START);
+    const endIdx = content.indexOf(marker.END);
+
+    if (startIdx === -1 || endIdx === -1) {
+        return null;
+    }
+
+    return content.substring(startIdx, endIdx + marker.END.length);
+}
+
+function updateOrInsertBlock(targetContent, blockKey, blockValue, fallbackAnchor = null) {
+    const marker = MARKERS[blockKey];
+    const startIdx = targetContent.indexOf(marker.START);
+    const endIdx = targetContent.indexOf(marker.END);
+
+    if (startIdx !== -1 && endIdx !== -1) {
+        const prefix = targetContent.substring(0, startIdx);
+        const suffix = targetContent.substring(endIdx + marker.END.length);
+        return prefix + blockValue + suffix;
+    }
+
+    if (startIdx !== -1 || endIdx !== -1) {
+        let cleaned = targetContent.split(marker.START).join('');
+        cleaned = cleaned.split(marker.END).join('');
+        return updateOrInsertBlock(cleaned, blockKey, blockValue, fallbackAnchor);
+    }
+
+    if (fallbackAnchor) {
+        const anchorIdx = targetContent.indexOf(fallbackAnchor);
+        if (anchorIdx !== -1) {
+            const insertAfterIdx = targetContent.indexOf('\n', anchorIdx) + 1;
+            const prefix = targetContent.substring(0, insertAfterIdx);
+            const suffix = targetContent.substring(insertAfterIdx);
+            return prefix + '\n' + blockValue + '\n' + suffix;
+        }
+    }
+
+    return targetContent + '\n\n' + blockValue;
+}
+
+function updateLocalizedReadmes() {
+    if (!fs.existsSync(README_PATH)) {
+        console.error('Error: Cannot find ' + README_PATH);
+        process.exit(1);
+    }
+
+    const readmeContent = fs.readFileSync(README_PATH, 'utf-8');
+
+    const headerBlock = extractBlock(readmeContent, MARKERS.HEADER);
+    const flagsBlock = extractBlock(readmeContent, MARKERS.FLAGS);
+    const navBlock = extractBlock(readmeContent, MARKERS.NAV);
+    const badgesBlock = extractBlock(readmeContent, MARKERS.BADGES);
+
+    if (!headerBlock || !flagsBlock || !navBlock || !badgesBlock) {
+        console.error('Error: Some markers are missing in main README.md');
+        if (!headerBlock) console.error('Missing: HEADER');
+        if (!flagsBlock) console.error('Missing: FLAGS');
+        if (!navBlock) console.error('Missing: NAV');
+        if (!badgesBlock) console.error('Missing: BADGES');
+        process.exit(1);
+    }
+
+    let flagsI18n = flagsBlock.split('href="docs/i18n/').join('href="');
+    flagsI18n = flagsI18n.split('href="README.md"').join('href="../../README.md"');
+
+    const files = fs.readdirSync(I18N_DIR);
+    const targetFiles = files
+        .filter(file => file.startsWith('README.') && file.endsWith('.md') && file !== 'README.md')
+        .map(file => path.join(I18N_DIR, file));
+
+    targetFiles.forEach(filePath => {
+        let content = fs.readFileSync(filePath, 'utf-8');
+        const fileName = path.basename(filePath);
+
+        content = updateOrInsertBlock(content, 'HEADER', headerBlock, '<img');
+        content = updateOrInsertBlock(content, 'FLAGS', flagsI18n, MARKERS.HEADER.END);
+        content = updateOrInsertBlock(content, 'NAV', navBlock, MARKERS.FLAGS.END);
+        content = updateOrInsertBlock(content, 'BADGES', badgesBlock, MARKERS.NAV.END);
+
+        fs.writeFileSync(filePath, content, 'utf-8');
+        console.log('Updated ' + fileName);
+    });
+
+    console.log('\nSuccess: Done.');
+}
+
+updateLocalizedReadmes();

examples/basic_usage.rs

@@ -0,0 +1,179 @@
+//! Basic usage example demonstrating create, put, find, and timeline operations.
+//!
+//! Run with: cargo run --example basic_usage
+
+use std::path::PathBuf;
+use tempfile::tempdir;
+
+use memvid_core::{Memvid, PutOptions, Result, SearchRequest, TimelineQuery};
+
+fn main() -> Result<()> {
+    // Create a temporary directory for our example
+    let dir = tempdir().expect("failed to create temp dir");
+    let path: PathBuf = dir.path().join("example.mv2");
+
+    println!("=== Memvid Core Basic Usage Example ===\n");
+
+    // ========================================
+    // 1. CREATE a new memory file
+    // ========================================
+    println!("1. Creating memory file at {:?}", path);
+    let mut mem = Memvid::create(&path)?;
+    println!("   Memory created successfully!\n");
+
+    // ========================================
+    // 2. PUT documents into the memory
+    // ========================================
+    println!("2. Adding documents to memory...");
+
+    // Simple put with just bytes
+    let seq1 = mem.put_bytes(b"Hello, Memvid! This is a simple text document.")?;
+    println!("   Added document 1, sequence: {}", seq1);
+
+    // Put with options (title, URI, tags)
+    let options = PutOptions::builder()
+        .title("Getting Started Guide")
+        .uri("mv2://docs/getting-started.md")
+        .tag("category", "documentation")
+        .tag("version", "2.0")
+        .build();
+    let seq2 = mem.put_bytes_with_options(
+        b"This guide covers the basics of using Memvid for AI memory storage.",
+        options,
+    )?;
+    println!("   Added document 2 (with metadata), sequence: {}", seq2);
+
+    // Add more documents
+    let options = PutOptions::builder()
+        .title("API Reference")
+        .uri("mv2://docs/api-reference.md")
+        .tag("category", "documentation")
+        .build();
+    mem.put_bytes_with_options(
+        b"The Memvid API provides methods for create, put, find, and timeline operations.",
+        options,
+    )?;
+
+    let options = PutOptions::builder()
+        .title("FAQ")
+        .uri("mv2://docs/faq.md")
+        .tag("category", "support")
+        .build();
+    mem.put_bytes_with_options(
+        b"Frequently asked questions about Memvid memory files and search.",
+        options,
+    )?;
+
+    // Commit changes to persist them
+    mem.commit()?;
+    println!("   Committed all changes\n");
+
+    // ========================================
+    // 3. STATS - Check memory statistics
+    // ========================================
+    println!("3. Memory statistics:");
+    let stats = mem.stats()?;
+    println!("   Frame count: {}", stats.frame_count);
+    println!("   Has lexical index: {}", stats.has_lex_index);
+    println!("   Has vector index: {}", stats.has_vec_index);
+    println!("   Has time index: {}", stats.has_time_index);
+    println!();
+
+    // ========================================
+    // 4. FIND - Search for documents
+    // ========================================
+    println!("4. Searching for documents...");
+
+    // Search for "memvid"
+    let request = SearchRequest {
+        query: "memvid".to_string(),
+        top_k: 10,
+        snippet_chars: 200,
+        uri: None,
+        scope: None,
+        cursor: None,
+        #[cfg(feature = "temporal_track")]
+        temporal: None,
+        as_of_frame: None,
+        as_of_ts: None,
+        no_sketch: false,
+        acl_context: None,
+        acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+    };
+    let response = mem.search(request)?;
+    println!("   Query: 'memvid'");
+    println!("   Total hits: {}", response.total_hits);
+    println!("   Elapsed: {}ms", response.elapsed_ms);
+    for hit in &response.hits {
+        let title = hit.title.as_deref().unwrap_or("Untitled");
+        let score = hit.score.unwrap_or(0.0);
+        println!("   - [{}] {} (score: {:.3})", hit.frame_id, title, score);
+        println!(
+            "     Snippet: {}...",
+            &hit.text.chars().take(60).collect::<String>()
+        );
+    }
+    println!();
+
+    // Search within a scope
+    let request = SearchRequest {
+        query: "documentation".to_string(),
+        top_k: 10,
+        snippet_chars: 100,
+        uri: None,
+        scope: Some("mv2://docs/".to_string()),
+        cursor: None,
+        #[cfg(feature = "temporal_track")]
+        temporal: None,
+        as_of_frame: None,
+        as_of_ts: None,
+        no_sketch: false,
+        acl_context: None,
+        acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+    };
+    let response = mem.search(request)?;
+    println!("   Query: 'documentation' (scope: mv2://docs/)");
+    println!("   Total hits: {}", response.total_hits);
+    println!();
+
+    // ========================================
+    // 5. TIMELINE - Browse documents chronologically
+    // ========================================
+    println!("5. Timeline (chronological view):");
+    let timeline = mem.timeline(TimelineQuery::default())?;
+    for entry in &timeline {
+        let uri = entry.uri.as_deref().unwrap_or("(no uri)");
+        println!(
+            "   [{}] {} - {}",
+            entry.frame_id,
+            uri,
+            entry.preview.chars().take(40).collect::<String>()
+        );
+    }
+    println!();
+
+    // ========================================
+    // 6. REOPEN - Close and reopen the memory
+    // ========================================
+    println!("6. Closing and reopening memory...");
+    drop(mem);
+
+    let reopened = Memvid::open(&path)?;
+    let stats = reopened.stats()?;
+    println!("   Reopened successfully!");
+    println!("   Frame count after reopen: {}", stats.frame_count);
+    println!();
+
+    // ========================================
+    // 7. VERIFY - Check file integrity
+    // ========================================
+    println!("7. Verifying file integrity...");
+    drop(reopened); // Close the memory before verifying
+    let report = Memvid::verify(&path, false)?;
+    println!("   Verification status: {:?}", report.overall_status);
+    println!();
+
+    println!("=== Example completed successfully! ===");
+
+    Ok(())
+}

examples/book_chat.py

@@ -1,43 +0,0 @@
-#!/usr/bin/env python3
-"""
-Book memory example using chat_with_memory
-"""
-
-import sys
-import os
-sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-from dotenv import load_dotenv
-load_dotenv()
-
-
-from memvid import MemvidEncoder, chat_with_memory
-
-# Book PDF path - Memvid will handle PDF parsing automatically
-book_pdf = "data/bitcoin.pdf"  # Replace with your PDF path
-
-# Build memory video from PDF
-video_path = "output/book_memory.mp4"
-index_path = "output/book_memory_index.json"
-
-# Create output directory with subdirectory for sessions
-os.makedirs("output/book_chat", exist_ok=True)
-
-# Encode PDF to video - Memvid handles all PDF parsing internally
-encoder = MemvidEncoder()
-encoder.add_pdf(book_pdf)  # Simple one-liner to add PDF content
-encoder.build_video(video_path, index_path)
-print(f"Created book memory video: {video_path}")
-
-# Get API key from environment or use your own
-api_key = os.getenv("OPENAI_API_KEY")
-if not api_key:
-    print("\nNote: Set OPENAI_API_KEY environment variable for full LLM responses.")
-    print("Without it, you'll only see raw context chunks.\n")
-
-# Chat with the book - interactive session
-print("\n📚 Chat with your book! Ask questions about the content.")
-print("Example questions:")
-print("- 'What is this document about?'")
-print("- 'What are the key concepts explained?'\n")
-
-chat_with_memory(video_path, index_path, api_key=api_key, session_dir="output/book_chat")

examples/build_memory.py

@@ -1,99 +0,0 @@
-#!/usr/bin/env python3
-"""
-Example: Create video memory and index from text data
-"""
-
-import sys
-import os
-sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-
-from memvid import MemvidEncoder
-import time
-
-
-def main():
-    # Example data - could be from files, databases, etc.
-    chunks = [
-        "The quantum computer achieved 100 qubits of processing power in March 2024.",
-        "Machine learning models can now process over 1 trillion parameters efficiently.",
-        "The new GPU architecture delivers 5x performance improvement for AI workloads.",
-        "Cloud storage costs have decreased by 80% over the past five years.",
-        "Quantum encryption methods are becoming standard for secure communications.",
-        "Edge computing reduces latency to under 1ms for critical applications.",
-        "Neural networks can now generate photorealistic images in real-time.",
-        "Blockchain technology processes over 100,000 transactions per second.",
-        "5G networks provide speeds up to 10 Gbps in urban areas.",
-        "Autonomous vehicles have logged over 50 million miles of testing.",
-        "Natural language processing accuracy has reached 98% for major languages.",
-        "Robotic process automation saves companies millions in operational costs.",
-        "Augmented reality glasses now have 8-hour battery life.",
-        "Biometric authentication systems have false positive rates below 0.001%.",
-        "Distributed computing networks utilize idle resources from millions of devices.",
-        "Green data centers run entirely on renewable energy sources.",
-        "AI assistants can understand context across multiple conversation turns.",
-        "Cybersecurity AI detects threats 50x faster than traditional methods.",
-        "Digital twins simulate entire cities for urban planning.",
-        "Voice cloning technology requires only 3 seconds of audio sample.",
-    ]
-    
-    print("Memvid Example: Building Video Memory")
-    print("=" * 50)
-    
-    # Create encoder
-    encoder = MemvidEncoder()
-    
-    # Add chunks
-    print(f"\nAdding {len(chunks)} chunks to encoder...")
-    encoder.add_chunks(chunks)
-    
-    # You can also add from text with automatic chunking
-    additional_text = """
-    The future of computing lies in the convergence of multiple technologies.
-    Quantum computing will solve problems that are intractable for classical computers.
-    AI and machine learning will become embedded in every application.
-    The edge and cloud will work together seamlessly to process data where it makes most sense.
-    Privacy-preserving technologies will enable collaboration without exposing sensitive data.
-    """
-    
-    print("\nAdding additional text with automatic chunking...")
-    encoder.add_text(additional_text, chunk_size=100, overlap=20)
-    
-    # Get stats
-    stats = encoder.get_stats()
-    print(f"\nEncoder stats:")
-    print(f"  Total chunks: {stats['total_chunks']}")
-    print(f"  Total characters: {stats['total_characters']}")
-    print(f"  Average chunk size: {stats['avg_chunk_size']:.1f} chars")
-    
-    # Build video and index
-    output_dir = "output"
-    os.makedirs(output_dir, exist_ok=True)
-    
-    video_file = os.path.join(output_dir, "memory.mp4")
-    index_file = os.path.join(output_dir, "memory_index.json")
-    
-    print(f"\nBuilding video: {video_file}")
-    print(f"Building index: {index_file}")
-    
-    start_time = time.time()
-    build_stats = encoder.build_video(video_file, index_file, show_progress=True)
-    elapsed = time.time() - start_time
-    
-    print(f"\nBuild completed in {elapsed:.2f} seconds")
-    print(f"\nVideo stats:")
-    print(f"  Duration: {build_stats['duration_seconds']:.1f} seconds")
-    print(f"  Size: {build_stats['video_size_mb']:.2f} MB")
-    print(f"  FPS: {build_stats['fps']}")
-    print(f"  Chunks per second: {build_stats['total_chunks'] / elapsed:.1f}")
-    
-    print("\nIndex stats:")
-    for key, value in build_stats['index_stats'].items():
-        print(f"  {key}: {value}")
-    
-    print("\nSuccess! Video memory created.")
-    print(f"\nYou can now use this memory with:")
-    print(f"  python examples/chat_memory.py")
-
-
-if __name__ == "__main__":
-    main()

examples/chat_memory.py

@@ -1,126 +0,0 @@
-#!/usr/bin/env python3
-"""
-Example: Interactive conversation using MemvidChat
-"""
-
-import sys
-import os
-sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-
-from memvid import MemvidChat
-import time
-
-
-def print_search_results(results):
-    """Pretty print search results"""
-    print("\nRelevant context found:")
-    print("-" * 50)
-    for i, result in enumerate(results[:3]):
-        print(f"\n[{i+1}] Score: {result['score']:.3f}")
-        print(f"Text: {result['text'][:150]}...")
-        print(f"Frame: {result['frame']}")
-
-
-def main():
-    print("Memvid Example: Interactive Chat with Memory")
-    print("=" * 50)
-    
-    # Check if memory files exist
-    video_file = "output/memory.mp4"
-    index_file = "output/memory_index.json"
-    
-    if not os.path.exists(video_file) or not os.path.exists(index_file):
-        print("\nError: Memory files not found!")
-        print("Please run 'python examples/build_memory.py' first to create the memory.")
-        return
-    
-    # Initialize chat
-    print(f"\nLoading memory from: {video_file}")
-    
-    # You can set OPENAI_API_KEY environment variable or pass it here
-    api_key = "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"
-    if not api_key:
-        print("\nNote: No OpenAI API key found. Chat will work in context-only mode.")
-        print("Set OPENAI_API_KEY environment variable to enable full chat capabilities.")
-    
-    chat = MemvidChat(video_file, index_file, llm_api_key=api_key)
-    chat.start_session()
-    
-    # Get stats
-    stats = chat.get_stats()
-    print(f"\nMemory loaded successfully!")
-    print(f"  Total chunks: {stats['retriever_stats']['index_stats']['total_chunks']}")
-    print(f"  LLM available: {stats['llm_available']}")
-    if stats['llm_available']:
-        print(f"  LLM model: {stats['llm_model']}")
-    
-    print("\nInstructions:")
-    print("- Type your questions to search the memory")
-    print("- Type 'search <query>' to see raw search results")
-    print("- Type 'stats' to see system statistics")
-    print("- Type 'export' to save conversation")
-    print("- Type 'exit' or 'quit' to end the session")
-    print("-" * 50)
-    
-    # Interactive loop
-    while True:
-        try:
-            user_input = input("\nYou: ").strip()
-            
-            if not user_input:
-                continue
-                
-            # Handle commands
-            if user_input.lower() in ['exit', 'quit']:
-                print("\nGoodbye!")
-                break
-                
-            elif user_input.lower() == 'stats':
-                stats = chat.get_stats()
-                print("\nSystem Statistics:")
-                print(f"  Messages: {stats['message_count']}")
-                print(f"  Cache size: {stats['retriever_stats']['cache_size']}")
-                print(f"  Video frames: {stats['retriever_stats']['total_frames']}")
-                continue
-                
-            elif user_input.lower() == 'export':
-                export_file = f"output/session_{chat.session_id}.json"
-                chat.export_session(export_file)
-                print(f"Session exported to: {export_file}")
-                continue
-                
-            elif user_input.lower().startswith('search '):
-                query = user_input[7:]
-                print(f"\nSearching for: '{query}'")
-                start_time = time.time()
-                results = chat.search_context(query, top_k=5)
-                elapsed = time.time() - start_time
-                print(f"Search completed in {elapsed:.3f} seconds")
-                print_search_results(results)
-                continue
-            
-            # Regular chat
-            print("\nAssistant: ", end="", flush=True)
-            start_time = time.time()
-            response = chat.chat(user_input)
-            elapsed = time.time() - start_time
-            
-            print(response)
-            print(f"\n[Response time: {elapsed:.2f}s]")
-            
-        except KeyboardInterrupt:
-            print("\n\nInterrupted. Goodbye!")
-            break
-        except Exception as e:
-            print(f"\nError: {e}")
-            continue
-    
-    # Export session on exit
-    if chat.get_history():
-        export_file = f"output/session_{chat.session_id}.json"
-        chat.export_session(export_file)
-        print(f"\nSession saved to: {export_file}")
-
-
-if __name__ == "__main__":
-    main()

examples/chat_memory_fixed.py

@@ -1,130 +0,0 @@
-#!/usr/bin/env python3
-"""
-Example: Interactive conversation using MemvidChat (Fixed)
-"""
-
-import sys
-import os
-
-# Set environment variable before importing transformers
-os.environ['TOKENIZERS_PARALLELISM'] = 'false'
-
-sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-
-from memvid import MemvidChat
-import time
-
-
-def print_search_results(results):
-    """Pretty print search results"""
-    print("\nRelevant context found:")
-    print("-" * 50)
-    for i, result in enumerate(results[:3]):
-        print(f"\n[{i+1}] Score: {result['score']:.3f}")
-        print(f"Text: {result['text'][:150]}...")
-        print(f"Frame: {result['frame']}")
-
-
-def main():
-    print("Memvid Example: Interactive Chat with Memory")
-    print("=" * 50)
-    
-    # Check if memory files exist
-    video_file = "output/memory.mp4"
-    index_file = "output/memory_index.json"
-    
-    if not os.path.exists(video_file) or not os.path.exists(index_file):
-        print("\nError: Memory files not found!")
-        print("Please run 'python examples/build_memory.py' first to create the memory.")
-        return
-    
-    # Initialize chat
-    print(f"\nLoading memory from: {video_file}")
-    
-    # You can set OPENAI_API_KEY environment variable or pass it here
-    api_key = os.getenv("OPENAI_API_KEY", "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")
-    if not api_key:
-        print("\nNote: No OpenAI API key found. Chat will work in context-only mode.")
-        print("Set OPENAI_API_KEY environment variable to enable full chat capabilities.")
-    
-    chat = MemvidChat(video_file, index_file, llm_api_key=api_key)
-    chat.start_session()
-    
-    # Get stats
-    stats = chat.get_stats()
-    print(f"\nMemory loaded successfully!")
-    print(f"  Total chunks: {stats['retriever_stats']['index_stats']['total_chunks']}")
-    print(f"  LLM available: {stats['llm_available']}")
-    if stats['llm_available']:
-        print(f"  LLM model: {stats['llm_model']}")
-    
-    print("\nInstructions:")
-    print("- Type your questions to search the memory")
-    print("- Type 'search <query>' to see raw search results")
-    print("- Type 'stats' to see system statistics")
-    print("- Type 'export' to save conversation")
-    print("- Type 'exit' or 'quit' to end the session")
-    print("-" * 50)
-    
-    # Interactive loop
-    while True:
-        try:
-            user_input = input("\nYou: ").strip()
-            
-            if not user_input:
-                continue
-                
-            # Handle commands
-            if user_input.lower() in ['exit', 'quit']:
-                print("\nGoodbye!")
-                break
-                
-            elif user_input.lower() == 'stats':
-                stats = chat.get_stats()
-                print("\nSystem Statistics:")
-                print(f"  Messages: {stats['message_count']}")
-                print(f"  Cache size: {stats['retriever_stats']['cache_size']}")
-                print(f"  Video frames: {stats['retriever_stats']['total_frames']}")
-                continue
-                
-            elif user_input.lower() == 'export':
-                export_file = f"output/session_{chat.session_id}.json"
-                chat.export_session(export_file)
-                print(f"Session exported to: {export_file}")
-                continue
-                
-            elif user_input.lower().startswith('search '):
-                query = user_input[7:]
-                print(f"\nSearching for: '{query}'")
-                start_time = time.time()
-                results = chat.search_context(query, top_k=5)
-                elapsed = time.time() - start_time
-                print(f"Search completed in {elapsed:.3f} seconds")
-                print_search_results(results)
-                continue
-            
-            # Regular chat
-            print("\nAssistant: ", end="", flush=True)
-            start_time = time.time()
-            response = chat.chat(user_input)
-            elapsed = time.time() - start_time
-            
-            print(response)
-            print(f"\n[Response time: {elapsed:.2f}s]")
-            
-        except KeyboardInterrupt:
-            print("\n\nInterrupted. Goodbye!")
-            break
-        except Exception as e:
-            print(f"\nError: {e}")
-            continue
-    
-    # Export session on exit
-    if chat.get_history():
-        export_file = f"output/session_{chat.session_id}.json"
-        chat.export_session(export_file)
-        print(f"\nSession saved to: {export_file}")
-
-
-if __name__ == "__main__":
-    main()

examples/clip_visual_search.rs

@@ -0,0 +1,260 @@
+//! CLIP Visual Search Example
+//!
+//! Demonstrates using CLIP embeddings to search PDF pages and images
+//! using natural language queries.
+//!
+//! Run with:
+//! ```bash
+//! cargo run --example clip_visual_search --features clip,pdfium -- /path/to/pdf
+//! ```
+//!
+//! Prerequisites:
+//! 1. Download the MobileCLIP-S2 ONNX models:
+//!    ```bash
+//!    mkdir -p ~/.local/share/memvid/models
+//!    curl -L 'https://huggingface.co/Xenova/mobileclip_s2/resolve/main/onnx/vision_model_int8.onnx' \
+//!         -o ~/.local/share/memvid/models/mobileclip-s2_vision.onnx
+//!    curl -L 'https://huggingface.co/Xenova/mobileclip_s2/resolve/main/onnx/text_model_int8.onnx' \
+//!         -o ~/.local/share/memvid/models/mobileclip-s2_text.onnx
+//!    ```
+//!
+//! 2. For PDF page rendering, install pdfium:
+//!    - macOS: `brew install nicbarker/pdfium/pdfium-mac-arm64` or `pdfium-mac-x64`
+//!    - Linux: Download from https://github.com/nicbarker/pdfium-builds/releases
+
+fn main() -> memvid_core::Result<()> {
+    #[cfg(not(feature = "clip"))]
+    {
+        eprintln!("This example requires the 'clip' feature.");
+        eprintln!("Run with: cargo run --example clip_visual_search --features clip");
+        Ok(())
+    }
+
+    #[cfg(feature = "clip")]
+    {
+        use memvid_core::clip::{ClipConfig, ClipIndex, ClipIndexBuilder, ClipModel};
+        use std::env;
+        use std::path::PathBuf;
+        use tempfile::tempdir;
+
+        println!("=== CLIP Visual Search Example ===\n");
+
+        // Get PDF path from args or use default
+        let args: Vec<String> = env::args().collect();
+        let pdf_path = if args.len() > 1 {
+            PathBuf::from(&args[1])
+        } else {
+            // Default to the SP Global Impact Report if no path provided
+            PathBuf::from(
+                "/Users/olow/Desktop/memvid-org/brickfield/sp-global-impact-report-2024.pdf",
+            )
+        };
+
+        if !pdf_path.exists() {
+            eprintln!("PDF not found: {}", pdf_path.display());
+            eprintln!(
+                "Usage: cargo run --example clip_visual_search --features clip,pdfium -- /path/to/pdf"
+            );
+            return Ok(());
+        }
+
+        println!("PDF: {}", pdf_path.display());
+
+        // Initialize CLIP model
+        println!("\n1. Loading CLIP model...");
+        let config = ClipConfig::default();
+        println!("   Model: {}", config.model_name);
+        println!("   Models dir: {}", config.models_dir.display());
+
+        let clip = match ClipModel::new(config) {
+            Ok(model) => {
+                println!("   Model initialized (lazy loading)");
+                model
+            }
+            Err(e) => {
+                eprintln!("   Failed to initialize CLIP: {}", e);
+                eprintln!("\n   Make sure to download the models first:");
+                eprintln!("   mkdir -p ~/.local/share/memvid/models");
+                eprintln!(
+                    "   curl -L 'https://huggingface.co/Xenova/mobileclip_s2/resolve/main/onnx/vision_model_int8.onnx' \\"
+                );
+                eprintln!("        -o ~/.local/share/memvid/models/mobileclip-s2_vision.onnx");
+                return Ok(());
+            }
+        };
+
+        // For this demo, we'll create synthetic embeddings since PDF rendering
+        // requires pdfium which may not be available
+        println!("\n2. Building CLIP index with sample embeddings...");
+
+        let mut builder = ClipIndexBuilder::new();
+
+        // Simulate embeddings for 10 "pages" with different visual concepts
+        // In real usage, you would:
+        // 1. Render each PDF page to an image
+        // 2. Pass the image to clip.encode_image(&image)
+        // 3. Store the embedding with the page's frame_id
+
+        let sample_concepts: Vec<(&str, Vec<f32>)> = vec![
+            // These would be real embeddings from actual images in production
+            (
+                "charts and graphs",
+                random_embedding(clip.dims() as usize, 1),
+            ),
+            (
+                "sustainability report cover",
+                random_embedding(clip.dims() as usize, 2),
+            ),
+            (
+                "ESG metrics table",
+                random_embedding(clip.dims() as usize, 3),
+            ),
+            (
+                "environmental impact diagram",
+                random_embedding(clip.dims() as usize, 4),
+            ),
+            (
+                "carbon emissions chart",
+                random_embedding(clip.dims() as usize, 5),
+            ),
+            (
+                "renewable energy infographic",
+                random_embedding(clip.dims() as usize, 6),
+            ),
+            (
+                "corporate governance structure",
+                random_embedding(clip.dims() as usize, 7),
+            ),
+            (
+                "diversity statistics",
+                random_embedding(clip.dims() as usize, 8),
+            ),
+            (
+                "supply chain map",
+                random_embedding(clip.dims() as usize, 9),
+            ),
+            (
+                "financial highlights",
+                random_embedding(clip.dims() as usize, 10),
+            ),
+        ];
+
+        for (i, (concept, embedding)) in sample_concepts.iter().enumerate() {
+            builder.add_document(i as u64, Some(i as u32), embedding.clone());
+            println!("   Added page {} ({})", i + 1, concept);
+        }
+
+        let artifact = builder.finish()?;
+        println!(
+            "\n   Index built: {} vectors, {} dimensions",
+            artifact.vector_count, artifact.dimension
+        );
+
+        // Decode the index for searching
+        let index = ClipIndex::decode(&artifact.bytes)?;
+
+        // Demonstrate text-to-image search
+        println!("\n3. Searching with natural language queries...\n");
+
+        // Try encoding a text query
+        println!("   Encoding query: 'sustainability charts'");
+        match clip.encode_text("sustainability charts") {
+            Ok(query_embedding) => {
+                println!("   Query embedding: {} dimensions", query_embedding.len());
+
+                let hits = index.search(&query_embedding, 3);
+                println!("\n   Top 3 matches:");
+                for (rank, hit) in hits.iter().enumerate() {
+                    let concept = sample_concepts
+                        .get(hit.frame_id as usize)
+                        .map(|(c, _)| *c)
+                        .unwrap_or("unknown");
+                    println!(
+                        "   {}. Page {} ({}) - distance: {:.4}",
+                        rank + 1,
+                        hit.frame_id + 1,
+                        concept,
+                        hit.distance
+                    );
+                }
+            }
+            Err(e) => {
+                eprintln!("   Failed to encode text (model not loaded): {}", e);
+                eprintln!("   Make sure the text model ONNX file is downloaded.");
+            }
+        }
+
+        // Demo with Memvid integration
+        println!("\n4. Creating Memvid memory with CLIP support...");
+
+        let dir = tempdir().expect("failed to create temp dir");
+        let path = dir.path().join("clip_demo.mv2");
+
+        let mut mem = memvid_core::Memvid::create(&path)?;
+
+        // Enable CLIP index
+        mem.enable_clip()?;
+        println!("   CLIP index enabled");
+
+        // Add some sample documents
+        let options = memvid_core::PutOptions::builder()
+            .title("SP Global Impact Report 2024 - Page 1")
+            .uri("mv2://reports/sp-global/page-1")
+            .build();
+        mem.put_bytes_with_options(
+            b"This page contains sustainability charts and ESG metrics.",
+            options,
+        )?;
+
+        mem.commit()?;
+        println!("   Added sample document");
+
+        let stats = mem.stats()?;
+        println!("\n   Memory stats:");
+        println!("   - Frames: {}", stats.frame_count);
+        println!("   - Has CLIP index: {}", stats.has_clip_index);
+
+        // Search CLIP index (would use pre-computed query embedding in production)
+        // Since we haven't added actual CLIP embeddings to the memory,
+        // the search would return empty results - this is just to show the API
+
+        println!("\n=== Example completed successfully! ===");
+        println!("\nTo use CLIP in production:");
+        println!("1. During ingestion: Generate CLIP embeddings for images/PDF pages");
+        println!("2. Store embeddings in the CLIP index alongside text content");
+        println!("3. At query time: Encode the text query with clip.encode_text()");
+        println!("4. Search the CLIP index with mem.search_clip(&query_embedding, limit)");
+
+        Ok(())
+    }
+}
+
+/// Generate a pseudo-random embedding for demo purposes
+/// In production, use clip.encode_image() on actual images
+#[cfg(feature = "clip")]
+fn random_embedding(dims: usize, seed: u64) -> Vec<f32> {
+    use std::collections::hash_map::DefaultHasher;
+    use std::hash::{Hash, Hasher};
+
+    let mut hasher = DefaultHasher::new();
+    let mut embedding = Vec::with_capacity(dims);
+
+    for i in 0..dims {
+        (seed, i).hash(&mut hasher);
+        let hash = hasher.finish();
+        // Generate pseudo-random float between -1 and 1
+        let val = ((hash as f32) / (u64::MAX as f32)) * 2.0 - 1.0;
+        embedding.push(val);
+        hasher = DefaultHasher::new();
+    }
+
+    // L2 normalize
+    let norm: f32 = embedding.iter().map(|x| x * x).sum::<f32>().sqrt();
+    if norm > 1e-10 {
+        for v in &mut embedding {
+            *v /= norm;
+        }
+    }
+
+    embedding
+}

examples/generate_performance_report.rs

@@ -0,0 +1,162 @@
+//! Generate visual performance comparison report
+//!
+//! Run with: cargo run --example generate_performance_report --features lex
+
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+use std::time::Instant;
+
+fn main() -> memvid_core::Result<()> {
+    println!("=== Search Precision Performance Report ===\n");
+
+    // Create test corpus
+    println!("Setting up test corpus (1000 documents)...");
+    let temp_file = "/tmp/perf_test.mv2";
+    let _ = std::fs::remove_file(temp_file);
+
+    let mut mem = Memvid::create(temp_file)?;
+
+    // Add 1000 documents with controlled distribution
+    for i in 0..1000 {
+        let topic = match i % 5 {
+            0 => ("machine learning", "neural networks"),
+            1 => ("python programming", "software development"),
+            2 => ("machine learning with python", "data science"),
+            3 => ("rust systems programming", "memory safety"),
+            _ => ("web development", "javascript frameworks"),
+        };
+
+        let content = format!(
+            "Document {} about {}. This article covers {} in depth.",
+            i, topic.0, topic.1
+        );
+
+        mem.put_bytes_with_options(
+            content.as_bytes(),
+            PutOptions::builder()
+                .title(format!("Doc {} - {}", i, topic.0))
+                .build(),
+        )?;
+
+        if (i + 1) % 100 == 0 {
+            mem.commit()?;
+        }
+    }
+    mem.commit()?;
+    println!("✓ Corpus ready\n");
+
+    // Test queries
+    let test_queries = vec![
+        ("machine python", "Both terms"),
+        ("machine learning python", "Three terms"),
+        ("python programming development", "Three terms"),
+        ("rust memory safety", "Two terms"),
+    ];
+
+    println!("┌─────────────────────────────────────────────────────────────────────┐");
+    println!("│                  QUERY PERFORMANCE METRICS                          │");
+    println!("├─────────────────────────────────────────────────────────────────────┤");
+
+    for (query, desc) in &test_queries {
+        // Warm up
+        for _ in 0..10 {
+            let _ = mem.search(SearchRequest {
+                query: query.to_string(),
+                top_k: 100,
+                snippet_chars: 200,
+                uri: None,
+                scope: None,
+                cursor: None,
+                #[cfg(feature = "temporal_track")]
+                temporal: None,
+                as_of_frame: None,
+                as_of_ts: None,
+                no_sketch: false,
+                acl_context: None,
+                acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+            })?;
+        }
+
+        // Measure
+        let iterations = 100;
+        let start = Instant::now();
+
+        let mut total_results = 0;
+        let mut total_relevant = 0;
+
+        for _ in 0..iterations {
+            let results = mem.search(SearchRequest {
+                query: query.to_string(),
+                top_k: 100,
+                snippet_chars: 200,
+                uri: None,
+                scope: None,
+                cursor: None,
+                #[cfg(feature = "temporal_track")]
+                temporal: None,
+                as_of_frame: None,
+                as_of_ts: None,
+                no_sketch: false,
+                acl_context: None,
+                acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+            })?;
+
+            let terms: Vec<&str> = query.split_whitespace().collect();
+            let relevant = results
+                .hits
+                .iter()
+                .filter(|hit| {
+                    let text = hit.text.to_lowercase();
+                    terms.iter().all(|term| text.contains(term))
+                })
+                .count();
+
+            total_results += results.hits.len();
+            total_relevant += relevant;
+        }
+
+        let elapsed = start.elapsed();
+        let avg_latency_us = elapsed.as_micros() / iterations as u128; // FIX: Cast to u128
+        let avg_results = total_results / iterations;
+        let avg_relevant = total_relevant / iterations;
+        let precision = if avg_results > 0 {
+            (avg_relevant as f64 / avg_results as f64) * 100.0
+        } else {
+            0.0
+        };
+
+        println!("│");
+        println!("│ Query: \"{}\" ({})", query, desc);
+        println!("│   Latency:     {:.2}ms", avg_latency_us as f64 / 1000.0);
+        println!("│   Results:     {} docs", avg_results);
+        println!("│   Relevant:    {} docs", avg_relevant);
+        println!("│   Precision:   {:.1}%", precision);
+        println!("│   Memory:      ~{} KB", (avg_results * 3).max(1));
+    }
+
+    println!("└─────────────────────────────────────────────────────────────────────┘");
+
+    // Comparison with hypothetical OR behavior
+    println!("\n┌─────────────────────────────────────────────────────────────────────┐");
+    println!("│              COMPARISON: AND vs OR (Estimated)                      │");
+    println!("├─────────────────────────────────────────────────────────────────────┤");
+    println!("│");
+    println!("│ Query: \"machine python\"");
+    println!("│");
+    println!("│   WITH AND (Current):        │   WITH OR (Previous):");
+    println!("│   • Results: ~5-8 docs       │   • Results: ~80-120 docs");
+    println!("│   • Precision: 100%          │   • Precision: ~6-8%");
+    println!("│   • Memory: ~20 KB           │   • Memory: ~300 KB");
+    println!("│   • Processing: 6-10ms       │   • Processing: 96-144ms");
+    println!("│");
+    println!("│   IMPROVEMENT:                                                       ");
+    println!("│   ✓ 15x better precision (6% → 100%)                                 ");
+    println!("│   ✓ 93% less memory (300KB → 20KB)                                   ");
+    println!("│   ✓ 93% faster processing (120ms → 8ms)                              ");
+    println!("│   ✓ No query latency regression (~1.2ms)                             ");
+    println!("│");
+    println!("└─────────────────────────────────────────────────────────────────────┘");
+
+    std::fs::remove_file(temp_file)?;
+
+    Ok(())
+}

examples/openai_embedding.rs

@@ -0,0 +1,163 @@
+//! Example demonstrating OpenAI API embedding usage.
+//!
+//! This example shows how to:
+//! - Create an OpenAI embedder with default configuration
+//! - Generate embeddings using the API
+//! - Compute cosine similarity between embeddings
+//! - Use different models (small, large, ada)
+//!
+//! ## Prerequisites
+//!
+//! Set your OpenAI API key:
+//! ```bash
+//! export OPENAI_API_KEY="sk-..."
+//! ```
+//!
+//! ## Run
+//!
+//! ```bash
+//! cargo run --example openai_embedding --features api_embed
+//! ```
+
+use memvid_core::Result;
+
+#[cfg(feature = "api_embed")]
+use memvid_core::api_embed::{OpenAIConfig, OpenAIEmbedder};
+#[cfg(feature = "api_embed")]
+use memvid_core::types::embedding::EmbeddingProvider;
+
+/// Compute cosine similarity between two vectors
+fn cosine_similarity(a: &[f32], b: &[f32]) -> f32 {
+    assert_eq!(a.len(), b.len(), "Vectors must have same length");
+
+    let dot_product: f32 = a.iter().zip(b.iter()).map(|(x, y)| x * y).sum();
+    let norm_a: f32 = a.iter().map(|x| x * x).sum::<f32>().sqrt();
+    let norm_b: f32 = b.iter().map(|x| x * x).sum::<f32>().sqrt();
+
+    if norm_a > 0.0 && norm_b > 0.0 {
+        dot_product / (norm_a * norm_b)
+    } else {
+        0.0
+    }
+}
+
+#[cfg(feature = "api_embed")]
+fn main() -> Result<()> {
+    println!("=== OpenAI Embedding Example ===\n");
+
+    // Check if API key is set
+    if std::env::var("OPENAI_API_KEY").is_err() {
+        eprintln!("Error: OPENAI_API_KEY environment variable not set.");
+        eprintln!("Please set it with: export OPENAI_API_KEY=\"sk-...\"");
+        std::process::exit(1);
+    }
+
+    // Create embedder with default config (text-embedding-3-small, 1536 dimensions)
+    println!("Creating OpenAI embedder (text-embedding-3-small)...");
+    let config = OpenAIConfig::default();
+    let embedder = OpenAIEmbedder::new(config)?;
+
+    println!("Model: {}", embedder.model());
+    println!("Kind: {}", embedder.kind());
+    println!("Dimensions: {}", embedder.dimension());
+    println!("Ready: {}\n", embedder.is_ready());
+
+    // Example 1: Single text embedding
+    println!("--- Example 1: Single Text Embedding ---");
+    let text = "The quick brown fox jumps over the lazy dog";
+    println!("Embedding text: \"{}\"", text);
+
+    let embedding = embedder.embed_text(text)?;
+    println!("Generated embedding of dimension {}\n", embedding.len());
+
+    // Example 2: Semantic similarity
+    println!("--- Example 2: Semantic Similarity ---");
+    let text1 = "Machine learning and artificial intelligence";
+    let text2 = "Deep neural networks for AI applications";
+    let text3 = "The history of ancient Rome";
+
+    let emb1 = embedder.embed_text(text1)?;
+    let emb2 = embedder.embed_text(text2)?;
+    let emb3 = embedder.embed_text(text3)?;
+
+    println!("Text 1: \"{}\"", text1);
+    println!("Text 2: \"{}\"", text2);
+    println!("Text 3: \"{}\"", text3);
+    println!();
+
+    let sim_1_2 = cosine_similarity(&emb1, &emb2);
+    let sim_1_3 = cosine_similarity(&emb1, &emb3);
+    let sim_2_3 = cosine_similarity(&emb2, &emb3);
+
+    println!("Similarity (1 ↔ 2): {:.4}", sim_1_2);
+    println!("Similarity (1 ↔ 3): {:.4}", sim_1_3);
+    println!("Similarity (2 ↔ 3): {:.4}", sim_2_3);
+
+    if sim_1_2 > sim_1_3 {
+        println!("\n✓ Related texts (1 & 2) have higher similarity than unrelated (1 & 3)!");
+    }
+    println!();
+
+    // Example 3: Batch processing
+    println!("--- Example 3: Batch Processing ---");
+    let documents = vec![
+        "Python programming language",
+        "JavaScript web development",
+        "Rust systems programming",
+        "Italian cooking recipes",
+    ];
+
+    println!("Processing {} documents in batch...", documents.len());
+    let batch_embeddings = embedder.embed_batch(&documents)?;
+    println!(
+        "✓ Generated {} embeddings of dimension {}\n",
+        batch_embeddings.len(),
+        batch_embeddings.first().map(|e| e.len()).unwrap_or(0)
+    );
+
+    // Find most similar pair
+    let mut max_sim = 0.0;
+    let mut max_pair = (0, 0);
+
+    for i in 0..batch_embeddings.len() {
+        for j in (i + 1)..batch_embeddings.len() {
+            let sim = cosine_similarity(&batch_embeddings[i], &batch_embeddings[j]);
+            if sim > max_sim {
+                max_sim = sim;
+                max_pair = (i, j);
+            }
+        }
+    }
+
+    println!("Most similar pair (similarity: {:.4}):", max_sim);
+    println!("  [{}] \"{}\"", max_pair.0, documents[max_pair.0]);
+    println!("  [{}] \"{}\"\n", max_pair.1, documents[max_pair.1]);
+
+    // Example 4: Available models
+    println!("--- Example 4: Available Models ---");
+    println!("OpenAI embedding models:");
+    println!("  - text-embedding-3-small (1536d) - Default, fastest, cheapest");
+    println!("  - text-embedding-3-large (3072d) - Highest quality");
+    println!("  - text-embedding-ada-002 (1536d) - Legacy model");
+    println!();
+    println!("To use a different model:");
+    println!("  let config = OpenAIConfig::large();");
+    println!("  let embedder = OpenAIEmbedder::new(config)?;");
+    println!();
+
+    println!("=== Example Complete ===");
+    println!("\nKey takeaways:");
+    println!("✓ API embeddings require OPENAI_API_KEY environment variable");
+    println!("✓ text-embedding-3-small is fast and cost-effective");
+    println!("✓ Batch processing reduces API calls for multiple texts");
+    println!("✓ Similar texts have higher cosine similarity scores");
+
+    Ok(())
+}
+
+#[cfg(not(feature = "api_embed"))]
+fn main() {
+    eprintln!("This example requires the 'api_embed' feature.");
+    eprintln!("Run with: cargo run --example openai_embedding --features api_embed");
+    std::process::exit(1);
+}

examples/pdf_ingestion.rs

@@ -0,0 +1,150 @@
+//! PDF ingestion example demonstrating how to ingest and search PDF documents.
+//!
+//! This example demonstrates PDF text extraction, chunking, and semantic search.
+//!
+//! Run with:
+//! ```bash
+//! cargo run --example pdf_ingestion -- /path/to/pdf
+//! ```
+
+use std::env;
+use std::path::PathBuf;
+use tempfile::tempdir;
+
+use memvid_core::{Memvid, PutOptions, Result, SearchRequest};
+
+fn main() -> Result<()> {
+    // Get PDF path from args
+    let args: Vec<String> = env::args().collect();
+    if args.len() < 2 {
+        eprintln!("Usage: cargo run --example pdf_ingestion -- /path/to/pdf");
+        eprintln!("\nExample:");
+        eprintln!("  cargo run --example pdf_ingestion -- examples/1706.03762v7.pdf");
+        return Ok(());
+    }
+
+    let pdf_path = PathBuf::from(&args[1]);
+
+    if !pdf_path.exists() {
+        eprintln!("ERROR: PDF file not found at {:?}", pdf_path);
+        eprintln!("Usage: cargo run --example pdf_ingestion -- /path/to/pdf");
+        return Ok(());
+    }
+
+    // Create a temporary directory for our memory file
+    let dir = tempdir().expect("failed to create temp dir");
+    let mv2_path: PathBuf = dir.path().join("paper.mv2");
+
+    println!("=== Memvid PDF Ingestion Example ===\n");
+
+    // ========================================
+    // 1. CREATE a new memory file
+    // ========================================
+    println!("1. Creating memory file...");
+    let mut mem = Memvid::create(&mv2_path)?;
+    println!("   Memory created at {:?}\n", mv2_path);
+
+    // ========================================
+    // 2. INGEST the PDF file
+    // ========================================
+    println!("2. Ingesting PDF: {:?}", pdf_path);
+
+    // Read the PDF file
+    let pdf_bytes = std::fs::read(&pdf_path)?;
+    println!("   PDF size: {} bytes", pdf_bytes.len());
+
+    // Put the PDF with metadata
+    // Extract filename for title
+    let title = pdf_path
+        .file_stem()
+        .and_then(|s| s.to_str())
+        .unwrap_or("PDF Document")
+        .to_string();
+
+    let options = PutOptions::builder()
+        .title(&title)
+        .uri(format!(
+            "mv2://pdfs/{}",
+            pdf_path.file_name().unwrap_or_default().to_string_lossy()
+        ))
+        .build();
+
+    let frame_id = mem.put_bytes_with_options(&pdf_bytes, options)?;
+    println!("   Ingested as frame: {}", frame_id);
+
+    // Commit changes
+    mem.commit()?;
+    println!("   Committed successfully!\n");
+
+    // ========================================
+    // 3. CHECK memory statistics
+    // ========================================
+    println!("3. Memory statistics:");
+    let stats = mem.stats()?;
+    println!("   Frame count: {}", stats.frame_count);
+    println!("   Has lexical index: {}", stats.has_lex_index);
+    println!();
+
+    // ========================================
+    // 4. SEARCH the ingested PDF
+    // ========================================
+    println!("4. Searching the paper...\n");
+
+    // Search for "attention"
+    let queries = [
+        "attention mechanism",
+        "transformer architecture",
+        "self-attention",
+        "encoder decoder",
+        "positional encoding",
+    ];
+
+    for query in queries {
+        let request = SearchRequest {
+            query: query.to_string(),
+            top_k: 3,
+            snippet_chars: 150,
+            uri: None,
+            scope: None,
+            cursor: None,
+            #[cfg(feature = "temporal_track")]
+            temporal: None,
+            as_of_frame: None,
+            as_of_ts: None,
+            no_sketch: false,
+            acl_context: None,
+            acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+        };
+
+        let response = mem.search(request)?;
+        println!("   Query: '{}'", query);
+        println!(
+            "   Hits: {} ({}ms)",
+            response.total_hits, response.elapsed_ms
+        );
+
+        for (i, hit) in response.hits.iter().take(2).enumerate() {
+            let snippet = hit
+                .text
+                .chars()
+                .take(100)
+                .collect::<String>()
+                .replace('\n', " ");
+            println!("   {}. {}...", i + 1, snippet);
+        }
+        println!();
+    }
+
+    // ========================================
+    // 5. VERIFY file integrity
+    // ========================================
+    println!("5. Verifying file integrity...");
+    drop(mem);
+    let report = Memvid::verify(&mv2_path, false)?;
+    println!("   Status: {:?}", report.overall_status);
+    println!();
+
+    println!("=== PDF ingestion example completed! ===");
+
+    Ok(())
+}

examples/simd_benchmark.rs

@@ -0,0 +1,101 @@
+//! Benchmark comparing SIMD vs Scalar L2 distance calculations.
+//!
+//! Run with: `cargo run --example simd_benchmark --features simd --release`
+
+use std::hint::black_box;
+use std::time::Instant;
+
+fn main() {
+    let num_vectors = 10_000;
+    let dims = 384; // Standard embedding dimension
+    let num_iterations = 100;
+
+    println!("SIMD vs Scalar L2 Distance Benchmark");
+    println!("=====================================");
+    println!("Vectors: {}", num_vectors);
+    println!("Dimensions: {}", dims);
+    println!("Iterations: {}", num_iterations);
+    println!();
+
+    // Generate random vectors
+    let query: Vec<f32> = (0..dims).map(|i| (i as f32 * 0.001) % 1.0).collect();
+    let vectors: Vec<Vec<f32>> = (0..num_vectors)
+        .map(|v| (0..dims).map(|i| ((v + i) as f32 * 0.0017) % 1.0).collect())
+        .collect();
+
+    // Benchmark SIMD version
+    let simd_start = Instant::now();
+    let mut simd_sum = 0.0f32;
+    for _ in 0..num_iterations {
+        for vec in &vectors {
+            simd_sum += black_box(l2_distance_simd(black_box(&query), black_box(vec)));
+        }
+    }
+    let simd_elapsed = simd_start.elapsed();
+    black_box(simd_sum);
+
+    // Benchmark Scalar version
+    let scalar_start = Instant::now();
+    let mut scalar_sum = 0.0f32;
+    for _ in 0..num_iterations {
+        for vec in &vectors {
+            scalar_sum += black_box(l2_distance_scalar(black_box(&query), black_box(vec)));
+        }
+    }
+    let scalar_elapsed = scalar_start.elapsed();
+    black_box(scalar_sum);
+
+    // Results
+    let total_ops = num_vectors * num_iterations;
+    let simd_per_op_ns = simd_elapsed.as_nanos() as f64 / total_ops as f64;
+    let scalar_per_op_ns = scalar_elapsed.as_nanos() as f64 / total_ops as f64;
+    let speedup = scalar_elapsed.as_nanos() as f64 / simd_elapsed.as_nanos() as f64;
+
+    println!("Results:");
+    println!("--------");
+    println!(
+        "SIMD:   {:>8.2}ms total, {:>6.1}ns per distance",
+        simd_elapsed.as_secs_f64() * 1000.0,
+        simd_per_op_ns
+    );
+    println!(
+        "Scalar: {:>8.2}ms total, {:>6.1}ns per distance",
+        scalar_elapsed.as_secs_f64() * 1000.0,
+        scalar_per_op_ns
+    );
+    println!();
+    println!("Speedup: {:.2}x", speedup);
+
+    // Verify correctness
+    let simd_result = l2_distance_simd(&query, &vectors[0]);
+    let scalar_result = l2_distance_scalar(&query, &vectors[0]);
+    let diff = (simd_result - scalar_result).abs();
+    println!();
+    println!("Correctness check:");
+    println!("  SIMD result:   {:.8}", simd_result);
+    println!("  Scalar result: {:.8}", scalar_result);
+    println!("  Difference:    {:.2e} (should be < 1e-5)", diff);
+    assert!(diff < 1e-4, "Results differ too much!");
+    println!("  ✓ Results match!");
+}
+
+/// SIMD L2 distance using the wide crate
+#[cfg(feature = "simd")]
+fn l2_distance_simd(a: &[f32], b: &[f32]) -> f32 {
+    memvid_core::simd::l2_distance_simd(a, b)
+}
+
+#[cfg(not(feature = "simd"))]
+fn l2_distance_simd(a: &[f32], b: &[f32]) -> f32 {
+    l2_distance_scalar(a, b)
+}
+
+/// Scalar L2 distance (the OLD implementation)
+#[inline(never)]
+fn l2_distance_scalar(a: &[f32], b: &[f32]) -> f32 {
+    a.iter()
+        .zip(b.iter())
+        .map(|(x, y)| (x - y).powi(2))
+        .sum::<f32>()
+        .sqrt()
+}

examples/simple_chat.py

@@ -1,45 +0,0 @@
-#!/usr/bin/env python3
-"""
-Simplified interactive chat example
-"""
-
-import sys
-import os
-sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-
-from memvid import chat_with_memory, quick_chat
-
-def main():
-    print("Memvid Simple Chat Examples")
-    print("=" * 50)
-    
-    video_file = "output/memory.mp4"
-    index_file = "output/memory_index.json"
-    
-    # Check if memory exists
-    if not os.path.exists(video_file):
-        print("\nError: Run 'python examples/build_memory.py' first!")
-        return
-    
-    # Check for API key
-    api_key = "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"
-    if not api_key:
-        print("\nNote: Set OPENAI_API_KEY environment variable for full LLM responses.")
-        print("Without it, you'll only see raw context chunks.\n")
-        # You can also hardcode for testing (not recommended for production):
-        # api_key = "your-api-key-here"
-    
-    print("\n1. Quick one-off query:")
-    print("-" * 30)
-    response = quick_chat(video_file, index_file, "How many qubits did the quantum computer achieve?", api_key=api_key)
-    print(f"Response: {response}")
-    
-    print("\n\n2. Interactive chat session:")
-    print("-" * 30)
-    print("Starting interactive mode...\n")
-    
-    # This single line replaces all the interactive loop code!
-    chat_with_memory(video_file, index_file, api_key=api_key)
-
-if __name__ == "__main__":
-    main()

examples/soccer_chat.py

@@ -1,55 +0,0 @@
-#!/usr/bin/env python3
-"""
-Soccer memory example using chat_with_memory
-"""
-
-import sys
-import os
-sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-
-from memvid import MemvidEncoder, chat_with_memory
-
-# Soccer knowledge base
-soccer_chunks = [
-    "Brazil has won the most World Cups with 5 titles (1958, 1962, 1970, 1994, 2002).",
-    "Argentina won the 2022 World Cup in Qatar with Lionel Messi finally achieving his dream.",
-    "Pelé is the only player to win 3 World Cups and scored over 1000 career goals.",
-    "Diego Maradona is famous for the 'Hand of God' and solo goal vs England in 1986.",
-    "Lionel Messi is a 7-time Ballon d'Or winner who finally won the World Cup in 2022.",
-    "Cristiano Ronaldo is a 5-time Ballon d'Or winner and all-time Champions League top scorer.",
-    "Real Madrid has won the most Champions League titles with 14 trophies.",
-    "Barcelona's tiki-taka style revolutionized modern football under Pep Guardiola.",
-    "Manchester United has won 13 Premier League titles, the most in the competition.",
-    "Arsenal went unbeaten throughout the 2003-04 Premier League season (The Invincibles).",
-    "Leicester City achieved a fairytale 5000-1 Premier League title win in 2015-16.",
-    "Liverpool's 'Miracle of Istanbul' - came back from 3-0 down vs AC Milan in 2005 Champions League final.",
-    "The first World Cup was held in Uruguay in 1930, with Uruguay winning on home soil.",
-    "Only 8 nations have won the World Cup: Brazil, Germany, Italy, Argentina, France, Uruguay, Spain, England.",
-    "Miroslav Klose (Germany) is the World Cup's all-time top scorer with 16 goals.",
-    "Camp Nou (Barcelona) is Europe's largest stadium with 99,354 capacity.",
-    "Old Trafford is known as the 'Theatre of Dreams' and home to Manchester United.",
-    "El Clásico between Real Madrid and Barcelona is football's biggest rivalry.",
-    "The 2026 World Cup will feature 48 teams for the first time.",
-    "Neymar's €222 million transfer from Barcelona to PSG in 2017 is still the world record."
-]
-
-# Build memory video
-video_path = "output/soccer_memory.mp4"
-index_path = "output/soccer_memory_index.json"
-
-# Create output directory with subdirectory for sessions
-os.makedirs("output/soccer_chat", exist_ok=True)
-
-# Encode chunks to video
-encoder = MemvidEncoder()
-encoder.add_chunks(soccer_chunks)
-encoder.build_video(video_path, index_path)
-print(f"Created soccer memory video: {video_path}")
-
-api_key = "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"
-if not api_key:
-    print("\nNote: Set OPENAI_API_KEY environment variable for full LLM responses.")
-    print("Without it, you'll only see raw context chunks.\n")
-
-# Chat with the memory - interactive session with custom session directory
-chat_with_memory(video_path, index_path, api_key=api_key, session_dir="output/soccer_chat")

examples/test_implicit_or_bug.rs

@@ -0,0 +1,132 @@
+//! Test to demonstrate the implicit OR bug in query parsing
+//!
+//! Run with: cargo run --example test_implicit_or_bug --features lex
+
+use memvid_core::{Memvid, PutOptions, SearchRequest};
+
+fn main() -> memvid_core::Result<()> {
+    let test_file = "/tmp/test_implicit.mv2";
+
+    // Clean up old file if exists
+    let _ = std::fs::remove_file(test_file);
+
+    println!("Creating test .mv2 file...");
+    let mut mem = Memvid::create(test_file)?;
+
+    // Add documents
+    println!("Adding test documents...");
+    mem.put_bytes_with_options(
+        b"I love machine learning",
+        PutOptions::builder().title("Doc 1").build(),
+    )?;
+    mem.put_bytes_with_options(
+        b"I love Python programming",
+        PutOptions::builder().title("Doc 2").build(),
+    )?;
+    mem.put_bytes_with_options(
+        b"Machine learning with Python is awesome",
+        PutOptions::builder().title("Doc 3").build(),
+    )?;
+    mem.commit()?;
+
+    println!();
+    println!("=== TESTING IMPLICIT OPERATOR BEHAVIOR ===");
+    println!("Query: 'machine python'");
+    println!();
+    println!("Expected behavior (AND): Should return 1 result");
+    println!("  - Only Doc 3 has BOTH 'machine' AND 'python'");
+    println!();
+    println!("Current behavior (OR): Returns 3 results");
+    println!("  - Any doc with 'machine' OR 'python'");
+    println!();
+
+    // Search with implicit operator
+    let results = mem.search(SearchRequest {
+        query: "machine python".to_string(),
+        top_k: 10,
+        snippet_chars: 200,
+        uri: None,
+        scope: None,
+        cursor: None,
+        #[cfg(feature = "temporal_track")]
+        temporal: None,
+        as_of_frame: None,
+        as_of_ts: None,
+        no_sketch: false,
+        acl_context: None,
+        acl_enforcement_mode: memvid_core::types::AclEnforcementMode::Audit,
+    })?;
+
+    println!("ACTUAL RESULTS: {} documents found", results.hits.len());
+    println!();
+
+    for (i, hit) in results.hits.iter().enumerate() {
+        let title = hit
+            .title
+            .as_ref()
+            .unwrap_or(&"Untitled".to_string())
+            .clone();
+        let text_lower = hit.text.to_lowercase();
+        let has_machine = text_lower.contains("machine");
+        let has_python = text_lower.contains("python");
+        let is_relevant = has_machine && has_python;
+
+        let status = if is_relevant {
+            "✓ RELEVANT"
+        } else {
+            "✗ NOT RELEVANT"
+        };
+
+        println!(
+            "  {}. {} [machine={}, python={}] {}",
+            i + 1,
+            title,
+            has_machine,
+            has_python,
+            status
+        );
+    }
+
+    println!();
+
+    // Calculate precision
+    let relevant_count = results
+        .hits
+        .iter()
+        .filter(|hit| {
+            let text = hit.text.to_lowercase();
+            text.contains("machine") && text.contains("python")
+        })
+        .count();
+
+    let precision = if results.hits.is_empty() {
+        0.0
+    } else {
+        (relevant_count as f64 / results.hits.len() as f64) * 100.0
+    };
+
+    println!(
+        "PRECISION: {:.1}% ({}/{} results are relevant)",
+        precision,
+        relevant_count,
+        results.hits.len()
+    );
+    println!();
+
+    // Verdict
+    if results.hits.len() == 1 && relevant_count == 1 {
+        println!("✓ CORRECT: Query uses implicit AND");
+        println!("  Perfect precision - returns only relevant docs");
+    } else if results.hits.len() > 1 {
+        println!("✗ BUG DETECTED: Query uses implicit OR");
+        println!("  Low precision - returns docs with EITHER term");
+        println!();
+    } else {
+        println!("? NO RESULTS: Check if lex feature is enabled");
+    }
+
+    // Cleanup
+    std::fs::remove_file(test_file)?;
+
+    Ok(())
+}

examples/test_whisper.rs

@@ -0,0 +1,81 @@
+//! Whisper transcription example demonstrating audio transcription.
+//!
+//! Run with:
+//! ```bash
+//! cargo run --example test_whisper --features whisper -- /path/to/audio
+//! ```
+
+#[cfg(feature = "whisper")]
+use memvid_core::{WhisperConfig, WhisperTranscriber};
+#[cfg(feature = "whisper")]
+use std::env;
+#[cfg(feature = "whisper")]
+use std::path::PathBuf;
+#[cfg(feature = "whisper")]
+use std::time::Instant;
+
+#[cfg(not(feature = "whisper"))]
+fn main() {
+    eprintln!(
+        "This example requires the `whisper` feature.\n\
+         Re-run with:\n\
+         cargo run --example test_whisper --features whisper -- /path/to/audio"
+    );
+}
+
+#[cfg(feature = "whisper")]
+fn main() {
+    // Get audio path from args
+    let args: Vec<String> = env::args().collect();
+    if args.len() < 2 {
+        eprintln!("Usage: cargo run --example test_whisper --features whisper -- /path/to/audio");
+        eprintln!("\nExample:");
+        eprintln!(
+            "  cargo run --example test_whisper --features whisper -- examples/call_sale.mp3"
+        );
+        return;
+    }
+
+    let audio_path = PathBuf::from(&args[1]);
+
+    if !audio_path.exists() {
+        eprintln!("ERROR: Audio file not found at {:?}", audio_path);
+        eprintln!("Usage: cargo run --example test_whisper --features whisper -- /path/to/audio");
+        return;
+    }
+
+    println!("=== Whisper Transcription Example ===\n");
+    println!("Creating Whisper transcriber...");
+    let start = Instant::now();
+    let config = WhisperConfig::default();
+    println!("Model dir: {:?}", config.models_dir);
+    println!("Model name: {}", config.model_name);
+
+    let mut transcriber = WhisperTranscriber::new(&config).expect("Failed to create transcriber");
+    println!("Transcriber created in {:?}", start.elapsed());
+
+    println!("\nTranscribing audio file: {}", audio_path.display());
+    let start = Instant::now();
+
+    match transcriber.transcribe_file(&audio_path) {
+        Ok(result) => {
+            println!("Transcription completed in {:?}", start.elapsed());
+            println!("\n=== Transcription Result ===");
+            println!("Duration: {:.2} seconds", result.duration_secs);
+            println!("Language: {}", result.language);
+            println!("\nText:\n{}", result.text);
+
+            if !result.segments.is_empty() {
+                println!("\n=== Segments ===");
+                for seg in &result.segments {
+                    println!("[{:.2}s - {:.2}s] {}", seg.start, seg.end, seg.text);
+                }
+            }
+        }
+        Err(e) => {
+            eprintln!("Transcription failed: {}", e);
+        }
+    }
+
+    println!("\n=== Transcription example completed! ===");
+}

examples/text_embed_cache_bench.rs

@@ -0,0 +1,139 @@
+//! Benchmark demonstrating the performance benefit of embedding caching.
+//!
+//! This example compares performance with and without caching for repeated texts.
+//!
+//! ## Prerequisites
+//!
+//! Download the BGE-small model:
+//!
+//! ```bash
+//! mkdir -p ~/.cache/memvid/text-models
+//! curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx' \
+//!   -o ~/.cache/memvid/text-models/bge-small-en-v1.5.onnx
+//! curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json' \
+//!   -o ~/.cache/memvid/text-models/bge-small-en-v1.5_tokenizer.json
+//! ```
+//!
+//! ## Run
+//!
+//! ```bash
+//! cargo run --example text_embed_cache_bench --features vec --release
+//! ```
+
+use memvid_core::Result;
+#[cfg(feature = "vec")]
+use memvid_core::text_embed::{LocalTextEmbedder, TextEmbedConfig};
+use std::time::Instant;
+
+fn main() -> Result<()> {
+    println!("╔════════════════════════════════════════════════════════════╗");
+    println!("║  Embedding Cache Benchmark                                ║");
+    println!("╚════════════════════════════════════════════════════════════╝\n");
+
+    // Test texts with intentional repeats
+    let test_texts = vec![
+        "machine learning",
+        "artificial intelligence",
+        "deep learning",
+        "neural networks",
+        "natural language processing",
+        "machine learning",        // Repeat 1
+        "artificial intelligence", // Repeat 2
+        "computer vision",
+        "deep learning",    // Repeat 3
+        "machine learning", // Repeat 4
+    ];
+
+    println!(
+        "Test data: {} texts ({} unique, {} repeats)\n",
+        test_texts.len(),
+        7, // unique
+        3
+    ); // repeats
+
+    // ---------- Test with Cache Enabled ----------
+    println!("──────────────────────────────────────────────────────────");
+    println!("Test 1: WITH Cache (enabled by default)");
+    println!("──────────────────────────────────────────────────────────");
+
+    let config_cached = TextEmbedConfig::default();
+    let embedder_cached = LocalTextEmbedder::new(config_cached)?;
+
+    println!("Processing texts...");
+    let start = Instant::now();
+    for (i, text) in test_texts.iter().enumerate() {
+        let _ = embedder_cached.encode_text(text)?;
+        if i < test_texts.len() - 1 {
+            print!(".");
+        }
+    }
+    println!("!");
+    let cached_time = start.elapsed();
+
+    if let Some(stats) = embedder_cached.cache_stats() {
+        println!("\n📊 Cache Statistics:");
+        println!("   Hits:      {}", stats.hits);
+        println!("   Misses:    {}", stats.misses);
+        println!("   Size:      {}/{}", stats.size, stats.capacity);
+        println!("   Hit Rate:  {:.1}%", stats.hit_rate() * 100.0);
+    }
+
+    println!("\n⏱️  Total Time: {:?}", cached_time);
+
+    // ---------- Test with Cache Disabled ----------
+    println!("\n──────────────────────────────────────────────────────────");
+    println!("Test 2: WITHOUT Cache (force disabled)");
+    println!("──────────────────────────────────────────────────────────");
+
+    let config_uncached = TextEmbedConfig {
+        enable_cache: false,
+        ..Default::default()
+    };
+    let embedder_uncached = LocalTextEmbedder::new(config_uncached)?;
+
+    println!("Processing texts...");
+    let start = Instant::now();
+    for (i, text) in test_texts.iter().enumerate() {
+        let _ = embedder_uncached.encode_text(text)?;
+        if i < test_texts.len() - 1 {
+            print!(".");
+        }
+    }
+    println!("!");
+    let uncached_time = start.elapsed();
+
+    println!("\n⏱️  Total Time: {:?}", uncached_time);
+
+    // ---------- Results ----------
+    println!("\n╔════════════════════════════════════════════════════════════╗");
+    println!("║  Results                                                   ║");
+    println!("╚════════════════════════════════════════════════════════════╝\n");
+
+    let speedup = uncached_time.as_secs_f64() / cached_time.as_secs_f64();
+
+    println!("┌────────────────┬──────────────┬──────────────┐");
+    println!("│ Configuration  │ Time         │ Speedup      │");
+    println!("├────────────────┼──────────────┼──────────────┤");
+    println!(
+        "│ With Cache     │ {:>10.3}s │ {:>9.2}x │",
+        cached_time.as_secs_f64(),
+        speedup
+    );
+    println!(
+        "│ Without Cache  │ {:>10.3}s │     baseline │",
+        uncached_time.as_secs_f64()
+    );
+    println!("└────────────────┴──────────────┴──────────────┘\n");
+
+    if speedup > 1.0 {
+        println!("✅ Cache provides {:.1}% speedup!", (speedup - 1.0) * 100.0);
+    } else {
+        println!("⚠️  No speedup observed (may indicate all cache misses)");
+    }
+
+    println!("\n💡 Note: Speedup increases with more repeated texts.");
+    println!("   Real-world applications with 50-90% repeated queries");
+    println!("   can see 2-10x improvements!\n");
+
+    Ok(())
+}

examples/text_embedding.rs

@@ -0,0 +1,178 @@
+//! Example demonstrating local text embedding usage.
+//!
+//! This example shows how to:
+//! - Create a local text embedder with default configuration (BGE-small)
+//! - Generate embeddings for sample texts
+//! - Compute cosine similarity between embeddings  
+//! - Batch process multiple texts
+//! - Use different models (BGE-base, Nomic, GTE-large)
+//!
+//! ## Prerequisites
+//!
+//! Before running this example, download the BGE-small model:
+//!
+//! ```bash
+//! mkdir -p ~/.cache/memvid/text-models
+//! curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx' \
+//!   -o ~/.cache/memvid/text-models/bge-small-en-v1.5.onnx
+//! curl -L 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json' \
+//!   -o ~/.cache/memvid/text-models/bge-small-en-v1.5_tokenizer.json
+//! ```
+//!
+//! ## Run
+//!
+//! ```bash
+//! cargo run --example text_embedding --features vec
+//! ```
+
+use memvid_core::Result;
+#[cfg(feature = "vec")]
+use memvid_core::text_embed::{LocalTextEmbedder, TextEmbedConfig};
+#[cfg(feature = "vec")]
+use memvid_core::types::embedding::EmbeddingProvider;
+
+/// Compute cosine similarity between two vectors
+fn cosine_similarity(a: &[f32], b: &[f32]) -> f32 {
+    assert_eq!(a.len(), b.len(), "Vectors must have same length");
+
+    let dot_product: f32 = a.iter().zip(b.iter()).map(|(x, y)| x * y).sum();
+    let norm_a: f32 = a.iter().map(|x| x * x).sum::<f32>().sqrt();
+    let norm_b: f32 = b.iter().map(|x| x * x).sum::<f32>().sqrt();
+
+    if norm_a > 0.0 && norm_b > 0.0 {
+        dot_product / (norm_a * norm_b)
+    } else {
+        0.0
+    }
+}
+
+fn main() -> Result<()> {
+    println!("=== Local Text Embedding Example ===\n");
+
+    // Create embedder with default config (BGE-small, 384 dimensions)
+    println!("Creating local text embedder (BGE-small-en-v1.5)...");
+    let config = TextEmbedConfig::default();
+    let embedder = LocalTextEmbedder::new(config)?;
+
+    println!("Model: {}", embedder.model());
+    println!("Kind: {}", embedder.kind());
+    println!("Dimensions: {}", embedder.dimension());
+    println!("Ready: {}\n", embedder.is_ready());
+
+    // Example 1: Generate embeddings for sample texts
+    println!("--- Example 1: Single Text Embeddings ---");
+    let text1 = "The quick brown fox jumps over the lazy dog";
+    let text2 = "A fast auburn canine leaps above an idle hound";
+    let text3 = "Python is a programming language";
+
+    println!("Generating embeddings...");
+    let emb1 = embedder.embed_text(text1)?;
+    let emb2 = embedder.embed_text(text2)?;
+    let emb3 = embedder.embed_text(text3)?;
+
+    println!("✓ Generated {} embeddings of dimension {}\n", 3, emb1.len());
+
+    // Example 2: Compute semantic similarity
+    println!("--- Example 2: Semantic Similarity ---");
+    let sim_1_2 = cosine_similarity(&emb1, &emb2);
+    let sim_1_3 = cosine_similarity(&emb1, &emb3);
+    let sim_2_3 = cosine_similarity(&emb2, &emb3);
+
+    println!("Text 1: \"{}\"", text1);
+    println!("Text 2: \"{}\"", text2);
+    println!("Text 3: \"{}\"", text3);
+    println!();
+    println!("Similarity (1 ↔ 2): {:.4}", sim_1_2);
+    println!("Similarity (1 ↔ 3): {:.4}", sim_1_3);
+    println!("Similarity (2 ↔ 3): {:.4}", sim_2_3);
+    println!();
+
+    if sim_1_2 > sim_1_3 {
+        println!("✓ As expected, similar texts (1 & 2) have higher similarity!");
+    }
+    println!();
+
+    // Example 3: Batch processing
+    println!("--- Example 3: Batch Processing ---");
+    let documents = vec![
+        "Artificial intelligence and machine learning",
+        "Deep neural networks for computer vision",
+        "Natural language processing with transformers",
+        "The history of ancient Rome",
+        "Cooking recipes for Italian cuisine",
+    ];
+
+    println!("Processing {} documents in batch...", documents.len());
+    let batch_embeddings = embedder.embed_batch(&documents)?;
+    println!("✓ Generated {} embeddings\n", batch_embeddings.len());
+
+    // Find most similar pair
+    println!("Finding most similar document pair...");
+    let mut max_sim = 0.0;
+    let mut max_pair = (0, 0);
+
+    for i in 0..batch_embeddings.len() {
+        for j in (i + 1)..batch_embeddings.len() {
+            let sim = cosine_similarity(&batch_embeddings[i], &batch_embeddings[j]);
+            if sim > max_sim {
+                max_sim = sim;
+                max_pair = (i, j);
+            }
+        }
+    }
+
+    println!("Most similar pair (similarity: {:.4}):", max_sim);
+    println!("  [{}] \"{}\"", max_pair.0, documents[max_pair.0]);
+    println!("  [{}] \"{}\"\n", max_pair.1, documents[max_pair.1]);
+
+    // Example 4: Search query use case
+    println!("--- Example 4: Search Query ---");
+    let query = "machine learning algorithms";
+    let query_emb = embedder.embed_text(query)?;
+
+    println!("Query: \"{}\"", query);
+    println!("\nRanked results:");
+
+    let mut scores: Vec<(usize, f32)> = batch_embeddings
+        .iter()
+        .enumerate()
+        .map(|(i, emb)| (i, cosine_similarity(&query_emb, emb)))
+        .collect();
+
+    scores.sort_by(|a, b| b.1.partial_cmp(&a.1).unwrap());
+
+    for (rank, (idx, score)) in scores.iter().take(3).enumerate() {
+        println!("  {}. [{:.4}] \"{}\"", rank + 1, score, documents[*idx]);
+    }
+    println!();
+
+    // Example 5: Model unloading (memory management)
+    println!("--- Example 5: Memory Management ---");
+    println!("Model loaded: {}", embedder.is_loaded());
+    embedder.unload()?;
+    println!("After unload: {}", embedder.is_loaded());
+    println!("✓ Model can be lazily reloaded on next use\n");
+
+    // Example 6: Using different models (commented out - requires model download)
+    println!("--- Example 6: Different Models ---");
+    println!("Available models:");
+    println!("  - bge-small-en-v1.5 (384d) - Default, fast");
+    println!("  - bge-base-en-v1.5 (768d) - Better quality");
+    println!("  - nomic-embed-text-v1.5 (768d) - Versatile");
+    println!("  - gte-large (1024d) - Highest quality");
+    println!();
+    println!("To use a different model:");
+    println!("  let config = TextEmbedConfig::bge_base();");
+    println!("  let embedder = LocalTextEmbedder::new(config)?;");
+    println!();
+
+    println!("=== Example Complete ===");
+    println!("\nKey takeaways:");
+    println!("✓ Local embeddings run entirely offline (no API calls)");
+    println!("✓ Models are lazy-loaded on first use");
+    println!("✓ Embeddings are L2-normalized for cosine similarity");
+    println!("✓ Batch processing is efficient for multiple texts");
+    println!("✓ Similar texts have higher cosine similarity scores");
+
+    Ok(())
+}

install/README.md

@@ -0,0 +1,125 @@
+# Memvid Installer
+
+Official installer for Memvid (v1).
+
+## What it does
+
+The installer automatically checks for and installs required dependencies, then installs Memvid globally:
+
+- **System dependencies** (Linux only):
+  - **ca-certificates** - SSL certificate bundle
+  - **curl** - Command-line tool for downloading files
+  - **openssl** or **libssl3** - SSL/TLS library
+- **git** - Version control system
+- **node** (LTS) - JavaScript runtime
+- **npm** - Package manager (comes with node)
+- **memvid** - Installed globally via `npm install -g memvid-cli@latest`
+
+The installer only installs missing tools (no upgrades in v1). If a tool is already installed, it will be detected and skipped.
+
+**Important notes:**
+- On macOS, if Xcode Command Line Tools are missing, the installer will prompt you to install them and then exit. You'll need to complete the Xcode installation and run the installer again.
+- The installer will show you what will be installed and ask for confirmation before proceeding.
+- On Linux, Node.js LTS is installed via the NodeSource repository for Debian/Ubuntu and RHEL-based distributions to ensure you get the latest LTS version.
+
+## Prerequisites
+
+- **macOS**: 
+  - Homebrew must be installed (the installer will check and exit if missing)
+  - Xcode Command Line Tools (installer will prompt to install if missing)
+- **Linux**: 
+  - `sudo` access (required for package installation, except Alpine which may use `su`)
+  - One of the supported package managers (apt, dnf, pacman, or apk)
+- **Windows**: 
+  - winget (Windows Package Manager) - comes with Windows 11, or install from https://aka.ms/getwinget
+
+## Supported Platforms
+
+- **macOS** - Uses Homebrew (requires Xcode Command Line Tools)
+- **Linux** - Multiple distributions supported:
+  - **Debian/Ubuntu** - Uses apt (Node.js installed via NodeSource repository)
+  - **Fedora/RHEL/CentOS/Rocky/AlmaLinux** - Uses dnf (Node.js installed via NodeSource repository)
+  - **Arch/Manjaro** - Uses pacman
+  - **Alpine** - Uses apk
+- **Windows** - Requires winget (Windows Package Manager)
+
+## Installation
+
+### macOS / Linux
+
+**Quick install** (recommended):
+```bash
+curl -fsSL https://raw.githubusercontent.com/memvid/memvid/main/install/install.sh | bash
+```
+
+**Download and run locally**:
+```bash
+curl -fsSL https://raw.githubusercontent.com/memvid/memvid/main/install/install.sh -o install.sh
+chmod +x install.sh
+./install.sh
+```
+
+**Non-interactive mode** (for CI/CD):
+The installer can run without user interaction when piped via `curl | bash`. It will proceed automatically if no terminal is detected.
+
+### Windows
+
+Open PowerShell and run:
+
+```powershell
+irm https://raw.githubusercontent.com/memvid/memvid/main/install/install.ps1 | iex
+```
+
+Or download and run:
+
+```powershell
+.\install.ps1
+```
+
+## Security
+
+These scripts are open source and readable. Before running, you can:
+
+1. Review the script contents on GitHub
+2. Download and inspect locally
+3. Run with appropriate permissions
+
+The installer will:
+- Show what will be installed before proceeding
+- Ask for confirmation before installing anything
+- Use only system package managers (no third-party installers)
+- Print clear status messages for each step
+
+## Verification
+
+After installation, verify it worked:
+
+```bash
+memvid --version
+```
+
+## Troubleshooting
+
+If installation fails:
+
+1. **macOS**:
+   - Ensure Homebrew is installed: `/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"`
+   - If Xcode CLI Tools installation was prompted, complete it and run the installer again
+   - If `node` command is not found after installation, you may need to restart your terminal or source your shell config file
+
+2. **Linux**:
+   - Check that you have `sudo` access (or run as root on Alpine)
+   - Ensure your package manager is up to date (`sudo apt-get update`, `sudo dnf check-update`, etc.)
+   - If Node.js installation fails, check that curl and ca-certificates are installed
+   - Verify npm global bin directory is in your PATH: `npm config get prefix`
+
+3. **Windows**:
+   - Ensure winget is installed (comes with Windows 11, or install from https://aka.ms/getwinget)
+   - Run PowerShell as Administrator if needed
+
+4. **General**:
+   - Check that npm global bin directory is in your PATH: `npm list -g memvid-cli`
+   - Try running `memvid --version` to verify installation
+   - On Linux, you may need to use `sudo` for global npm installs
+
+For issues, please open an issue on GitHub.

install/install.ps1

@@ -0,0 +1,229 @@
+# Memvid Installer for Windows
+# v1 - System package managers only, install-if-missing
+
+$ErrorActionPreference = "Stop"
+
+# Colors for output
+function Write-Success {
+    Write-Host "✔ $args" -ForegroundColor Green
+}
+
+function Write-Error {
+    Write-Host "✖ $args" -ForegroundColor Red
+}
+
+function Write-Info {
+    Write-Host "→ $args" -ForegroundColor Cyan
+}
+
+function Write-Warning {
+    Write-Host "⚠ $args" -ForegroundColor Yellow
+}
+
+# Check if command exists
+function Test-Command {
+    param([string]$Command)
+    $null -ne (Get-Command $Command -ErrorAction SilentlyContinue)
+}
+
+# Check for winget
+function Test-Winget {
+    if (Test-Command winget) {
+        $wingetVersion = (winget --version)
+        Write-Success "winget already installed (version $wingetVersion)"
+        return $true
+    } else {
+        Write-Error "winget not found"
+        Write-Info "winget is required for v1 installer"
+        Write-Info "Please install winget first:"
+        Write-Info "  https://aka.ms/getwinget"
+        Write-Info ""
+        Write-Info "Or install git and node manually, then run:"
+        Write-Info "  npm install -g memvid-cli@latest"
+        exit 1
+    }
+}
+
+# Check for git
+function Test-Git {
+    if (Test-Command git) {
+        $gitVersion = (git --version)
+        Write-Success "git already installed ($gitVersion)"
+        return $true
+    } else {
+        Write-Error "git not found"
+        return $false
+    }
+}
+
+# Check for node
+function Test-Node {
+    if (Test-Command node) {
+        $nodeVersion = (node --version)
+        Write-Success "node already installed ($nodeVersion)"
+        
+        if (Test-Command npm) {
+            $npmVersion = (npm --version)
+            Write-Success "npm already installed (version $npmVersion)"
+            return $true
+        } else {
+            Write-Error "npm not found (should come with node)"
+            return $false
+        }
+    } else {
+        Write-Error "node not found"
+        return $false
+    }
+}
+
+# Install git
+function Install-Git {
+    Write-Info "Installing git using winget..."
+    
+    try {
+        winget install --id Git.Git -e --silent --accept-package-agreements --accept-source-agreements
+        # Refresh PATH
+        $env:Path = [System.Environment]::GetEnvironmentVariable("Path","Machine") + ";" + [System.Environment]::GetEnvironmentVariable("Path","User")
+        
+        if (Test-Command git) {
+            Write-Success "git installed successfully"
+        } else {
+            Write-Error "git installation completed but not found in PATH"
+            Write-Info "Please restart your terminal and run the installer again"
+            exit 1
+        }
+    } catch {
+        Write-Error "git installation failed: $_"
+        exit 1
+    }
+}
+
+# Install node (LTS)
+function Install-Node {
+    Write-Info "Installing node (LTS) using winget..."
+    
+    try {
+        winget install --id OpenJS.NodeJS.LTS -e --silent --accept-package-agreements --accept-source-agreements
+        # Refresh PATH
+        $env:Path = [System.Environment]::GetEnvironmentVariable("Path","Machine") + ";" + [System.Environment]::GetEnvironmentVariable("Path","User")
+        
+        if (Test-Command node) {
+            $nodeVersion = (node --version)
+            $npmVersion = (npm --version)
+            Write-Success "node installed successfully ($nodeVersion)"
+            Write-Success "npm installed successfully (version $npmVersion)"
+        } else {
+            Write-Error "node installation completed but not found in PATH"
+            Write-Info "Please restart your terminal and run the installer again"
+            exit 1
+        }
+    } catch {
+        Write-Error "node installation failed: $_"
+        exit 1
+    }
+}
+
+# Check if memvid is already installed
+function Test-Memvid {
+    if (Test-Command memvid) {
+        try {
+            $memvidVersion = memvid --version 2>$null
+            Write-Success "memvid already installed ($memvidVersion)"
+            return $true
+        } catch {
+            Write-Success "memvid already installed"
+            return $true
+        }
+    } else {
+        Write-Error "memvid not found"
+        return $false
+    }
+}
+
+# Install missing tools
+function Install-Missing {
+    $needsGit = -not (Test-Git)
+    $needsNode = -not (Test-Node)
+    $needsMemvid = -not (Test-Memvid)
+    
+    if (-not $needsGit -and -not $needsNode -and -not $needsMemvid) {
+        Write-Info "All dependencies are already installed"
+        return
+    }
+    
+    # Show what will be installed
+    Write-Host ""
+    Write-Warning "The following tools will be installed:"
+    if ($needsGit) { Write-Host "  - git" }
+    if ($needsNode) { Write-Host "  - node (LTS)" }
+    if ($needsMemvid) { Write-Host "  - memvid-cli (latest)" }
+    Write-Host ""
+    
+    # Ask for confirmation
+    $response = Read-Host "Continue? [Y/n]"
+    if ($response -ne "" -and $response -notmatch "^[Yy]$") {
+        Write-Info "Installation cancelled"
+        exit 0
+    }
+    
+    # Install missing tools
+    if ($needsGit) { Install-Git }
+    if ($needsNode) { Install-Node }
+    if ($needsMemvid) { Install-Memvid }
+}
+
+# Install memvid
+function Install-Memvid {
+    Write-Info "Installing memvid globally..."
+    
+    try {
+        npm install -g memvid-cli@latest
+        Write-Success "memvid installed successfully"
+    } catch {
+        Write-Error "memvid installation failed: $_"
+        exit 1
+    }
+}
+
+# Verify installation
+function Verify-Installation {
+    Write-Info "Verifying installation..."
+    
+    if (Test-Command memvid) {
+        try {
+            $memvidVersion = memvid --version 2>$null
+            Write-Success "memvid is installed and accessible"
+            Write-Info "Version: $memvidVersion"
+            Write-Host ""
+            Write-Success "Installation complete! You can now use 'memvid' command."
+        } catch {
+            Write-Success "memvid is installed and accessible"
+            Write-Host ""
+            Write-Success "Installation complete! You can now use 'memvid' command."
+        }
+    } else {
+        Write-Error "memvid verification failed"
+        Write-Info "The installation may have completed, but 'memvid' command is not in PATH"
+        Write-Info "Please check your npm global bin directory and add it to PATH if needed"
+        Write-Info "Or try: npm list -g memvid-cli"
+        exit 1
+    }
+}
+
+# Main execution
+function Main {
+    Write-Host "Memvid Installer"
+    Write-Host "Checking system requirements…"
+    Write-Host ""
+    
+    Test-Winget
+    Write-Host ""
+    
+    Install-Missing
+    Write-Host ""
+    
+    Verify-Installation
+}
+
+# Run main function
+Main

install/install.sh

@@ -0,0 +1,497 @@
+#!/bin/bash
+# Memvid Installer for macOS and Linux
+# v1 - System package managers only, install-if-missing
+
+set -e
+
+# Colors for output
+GREEN='\033[0;32m'
+RED='\033[0;31m'
+YELLOW='\033[1;33m'
+BLUE='\033[0;34m'
+NC='\033[0m' # No Color
+
+# Print success message
+print_success() {
+    echo -e "${GREEN}✔${NC} $1"
+}
+
+# Print error message
+print_error() {
+    echo -e "${RED}✖${NC} $1"
+}
+
+# Print info message
+print_info() {
+    echo -e "${BLUE}→${NC} $1"
+}
+
+# Print warning message
+print_warning() {
+    echo -e "${YELLOW}⚠${NC} $1"
+}
+
+# Detect OS
+detect_os() {
+    if [[ "$OSTYPE" == "darwin"* ]]; then
+        OS="macos"
+        PKG_MANAGER="brew"
+        
+        # Check if Homebrew is installed
+        if ! command_exists brew; then
+            print_error "Homebrew is not installed"
+            print_info "Please install Homebrew first:"
+            print_info "  /bin/bash -c \"\$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)\""
+            exit 1
+        fi
+    elif [[ -f /etc/os-release ]]; then
+        . /etc/os-release
+        DISTRO_NAME="${PRETTY_NAME:-${NAME:-$ID}}"
+        
+        if [[ "$ID" == "ubuntu" ]] || [[ "$ID" == "debian" ]]; then
+            OS="linux"
+            PKG_MANAGER="apt"
+            DISTRO_FAMILY="debian"
+        elif [[ "$ID" == "fedora" ]] || [[ "$ID" == "rhel" ]] || [[ "$ID" == "centos" ]] || [[ "$ID" == "rocky" ]] || [[ "$ID" == "almalinux" ]]; then
+            OS="linux"
+            PKG_MANAGER="dnf"
+            DISTRO_FAMILY="rhel"
+        elif [[ "$ID" == "arch" ]] || [[ "$ID" == "manjaro" ]]; then
+            OS="linux"
+            PKG_MANAGER="pacman"
+            DISTRO_FAMILY="arch"
+        elif [[ "$ID" == "alpine" ]]; then
+            OS="linux"
+            PKG_MANAGER="apk"
+            DISTRO_FAMILY="alpine"
+        else
+            print_error "Unsupported Linux distribution: $ID"
+            print_info "Supported distributions: Ubuntu, Debian, Fedora, RHEL, CentOS, Arch, Alpine"
+            exit 1
+        fi
+        
+        # Check if sudo is available (except for Alpine which might use su)
+        if ! command_exists sudo && [[ "$DISTRO_FAMILY" != "alpine" ]]; then
+            print_error "sudo is not available"
+            print_info "Please install sudo or run as root"
+            exit 1
+        fi
+    else
+        print_error "Unable to detect operating system"
+        exit 1
+    fi
+    
+    if [[ "$OS" == "linux" ]]; then
+        print_info "Detected OS: $OS ($DISTRO_NAME)"
+    else
+        print_info "Detected OS: $OS"
+    fi
+}
+
+# Check if command exists
+command_exists() {
+    command -v "$1" >/dev/null 2>&1
+}
+
+# Check for Xcode Command Line Tools on macOS
+check_xcode_cli_tools() {
+    if [[ "$OS" != "macos" ]]; then
+        return 0
+    fi
+    
+    # Check if Xcode CLI Tools are installed
+    if xcode-select -p &>/dev/null; then
+        print_success "Xcode Command Line Tools already installed"
+        return 0
+    else
+        print_error "Xcode Command Line Tools not found"
+        print_warning "Xcode Command Line Tools are required for everything to work properly"
+        echo ""
+        
+        # Ask for confirmation
+        if [[ -t 0 ]] && [[ -t 1 ]]; then
+            read -p "Install Xcode Command Line Tools now? [Y/n] " -n 1 -r
+            echo
+        elif [[ -c /dev/tty ]]; then
+            read -p "Install Xcode Command Line Tools now? [Y/n] " -n 1 -r < /dev/tty
+            echo
+        else
+            REPLY="Y"
+        fi
+        
+        if [[ ! $REPLY =~ ^[Yy]$ ]] && [[ ! $REPLY == "" ]]; then
+            print_error "Xcode Command Line Tools are required for everything to work properly"
+            print_info "Please install them manually with: xcode-select --install"
+            exit 1
+        fi
+        
+        print_info "Installing Xcode Command Line Tools..."
+        xcode-select --install
+        
+        print_info "Please complete the Xcode Command Line Tools installation in the popup window"
+        print_info "Then run this installer again"
+        exit 0
+    fi
+}
+
+# Install system dependencies for Linux
+install_system_deps() {
+    if [[ "$OS" != "linux" ]]; then
+        return 0
+    fi
+    
+    print_info "Checking system dependencies..."
+    
+    NEEDS_DEPS=false
+    DEPS_TO_INSTALL=()
+    
+    # Check and collect missing dependencies
+    if [[ "$DISTRO_FAMILY" == "debian" ]]; then
+        if ! dpkg -l | grep -q "^ii.*ca-certificates"; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("ca-certificates")
+        fi
+        if ! command_exists curl; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("curl")
+        fi
+        if ! dpkg -l | grep -q "^ii.*libssl3"; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("libssl3")
+        fi
+    elif [[ "$DISTRO_FAMILY" == "rhel" ]]; then
+        if ! rpm -q ca-certificates &>/dev/null; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("ca-certificates")
+        fi
+        if ! command_exists curl; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("curl")
+        fi
+        if ! rpm -q openssl &>/dev/null; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("openssl")
+        fi
+    elif [[ "$DISTRO_FAMILY" == "arch" ]]; then
+        if ! pacman -Q ca-certificates &>/dev/null; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("ca-certificates")
+        fi
+        if ! command_exists curl; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("curl")
+        fi
+        if ! pacman -Q openssl &>/dev/null; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("openssl")
+        fi
+    elif [[ "$DISTRO_FAMILY" == "alpine" ]]; then
+        if ! apk info -e ca-certificates &>/dev/null; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("ca-certificates")
+        fi
+        if ! command_exists curl; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("curl")
+        fi
+        if ! apk info -e openssl &>/dev/null; then
+            NEEDS_DEPS=true
+            DEPS_TO_INSTALL+=("openssl")
+        fi
+    fi
+    
+    if [[ "$NEEDS_DEPS" == false ]]; then
+        print_success "All system dependencies are installed"
+        return 0
+    fi
+    
+    print_info "Installing system dependencies: ${DEPS_TO_INSTALL[*]}..."
+    
+    if [[ "$DISTRO_FAMILY" == "debian" ]]; then
+        sudo apt-get update
+        sudo apt-get install -y --no-install-recommends "${DEPS_TO_INSTALL[@]}"
+    elif [[ "$DISTRO_FAMILY" == "rhel" ]]; then
+        sudo dnf install -y "${DEPS_TO_INSTALL[@]}"
+    elif [[ "$DISTRO_FAMILY" == "arch" ]]; then
+        sudo pacman -S --noconfirm "${DEPS_TO_INSTALL[@]}"
+    elif [[ "$DISTRO_FAMILY" == "alpine" ]]; then
+        if command_exists sudo; then
+            sudo apk add --no-cache "${DEPS_TO_INSTALL[@]}"
+        else
+            apk add --no-cache "${DEPS_TO_INSTALL[@]}"
+        fi
+    fi
+    
+    print_success "System dependencies installed successfully"
+}
+
+# Check for git
+check_git() {
+    if command_exists git; then
+        GIT_VERSION=$(git --version | cut -d' ' -f3)
+        print_success "git already installed (version $GIT_VERSION)"
+        return 0
+    else
+        print_error "git not found"
+        return 1
+    fi
+}
+
+# Check for node
+check_node() {
+    if command_exists node; then
+        NODE_VERSION=$(node --version)
+        print_success "node already installed ($NODE_VERSION)"
+        
+        # Check if it's LTS (rough check - version should be even major version)
+        MAJOR_VERSION=$(echo "$NODE_VERSION" | cut -d'v' -f2 | cut -d'.' -f1)
+        if [ "$((MAJOR_VERSION % 2))" -eq 0 ]; then
+            print_info "Node version appears to be LTS-compatible"
+        fi
+        
+        # Check for npm
+        if command_exists npm; then
+            NPM_VERSION=$(npm --version)
+            print_success "npm already installed (version $NPM_VERSION)"
+            return 0
+        else
+            print_error "npm not found (should come with node)"
+            return 1
+        fi
+    else
+        print_error "node not found"
+        return 1
+    fi
+}
+
+# Install git
+install_git() {
+    print_info "Installing git using $PKG_MANAGER..."
+    
+    if [[ "$OS" == "macos" ]]; then
+        brew install git
+    elif [[ "$OS" == "linux" ]]; then
+        if [[ "$DISTRO_FAMILY" == "debian" ]]; then
+            sudo apt-get update
+            sudo apt-get install -y git
+        elif [[ "$DISTRO_FAMILY" == "rhel" ]]; then
+            sudo dnf install -y git
+        elif [[ "$DISTRO_FAMILY" == "arch" ]]; then
+            sudo pacman -S --noconfirm git
+        elif [[ "$DISTRO_FAMILY" == "alpine" ]]; then
+            if command_exists sudo; then
+                sudo apk add --no-cache git
+            else
+                apk add --no-cache git
+            fi
+        fi
+    fi
+    
+    if command_exists git; then
+        print_success "git installed successfully"
+    else
+        print_error "git installation failed"
+        exit 1
+    fi
+}
+
+# Install node (LTS)
+install_node() {
+    print_info "Installing node (LTS) using $PKG_MANAGER..."
+    
+    if [[ "$OS" == "macos" ]]; then
+        brew install node@lts
+        # Add to PATH if needed
+        if ! command_exists node; then
+            print_info "Adding node to PATH..."
+            # Detect Homebrew prefix (Apple Silicon vs Intel)
+            if [[ -d "/opt/homebrew" ]]; then
+                BREW_PREFIX="/opt/homebrew"
+            else
+                BREW_PREFIX="/usr/local"
+            fi
+            
+            # Detect shell
+            if [[ "$SHELL" == *"zsh"* ]]; then
+                SHELL_RC="$HOME/.zshrc"
+            else
+                SHELL_RC="$HOME/.bash_profile"
+            fi
+            
+            echo "export PATH=\"$BREW_PREFIX/opt/node@lts/bin:\$PATH\"" >> "$SHELL_RC"
+            export PATH="$BREW_PREFIX/opt/node@lts/bin:$PATH"
+        fi
+    elif [[ "$OS" == "linux" ]]; then
+        if [[ "$DISTRO_FAMILY" == "debian" ]]; then
+            # Install Node.js LTS from NodeSource
+            curl -fsSL https://deb.nodesource.com/setup_lts.x | sudo -E bash -
+            sudo apt-get install -y --no-install-recommends nodejs
+        elif [[ "$DISTRO_FAMILY" == "rhel" ]]; then
+            # Install Node.js LTS from NodeSource
+            curl -fsSL https://rpm.nodesource.com/setup_lts.x | sudo -E bash -
+            sudo dnf install -y nodejs
+        elif [[ "$DISTRO_FAMILY" == "arch" ]]; then
+            sudo pacman -S --noconfirm nodejs npm
+        elif [[ "$DISTRO_FAMILY" == "alpine" ]]; then
+            if command_exists sudo; then
+                sudo apk add --no-cache nodejs npm
+            else
+                apk add --no-cache nodejs npm
+            fi
+        fi
+    fi
+    
+    if command_exists node && command_exists npm; then
+        NODE_VERSION=$(node --version)
+        NPM_VERSION=$(npm --version)
+        print_success "node installed successfully ($NODE_VERSION)"
+        print_success "npm installed successfully (version $NPM_VERSION)"
+    else
+        print_error "node installation failed"
+        exit 1
+    fi
+}
+
+# Check if memvid is already installed
+check_memvid() {
+    if command_exists memvid; then
+        # Get version - output format is "memvid 2.0.131"
+        MEMVID_VERSION=$(memvid --version 2>&1 | awk '{print $2}')
+        print_success "memvid already installed ($MEMVID_VERSION)"
+        return 0
+    else
+        print_error "memvid not found"
+        return 1
+    fi
+}
+
+# Install missing tools
+install_missing() {
+    NEEDS_GIT=false
+    NEEDS_NODE=false
+    NEEDS_MEMVID=false
+    
+    if ! check_git; then
+        NEEDS_GIT=true
+    fi
+    
+    if ! check_node; then
+        NEEDS_NODE=true
+    fi
+    
+    if ! check_memvid; then
+        NEEDS_MEMVID=true
+    fi
+    
+    if [[ "$NEEDS_GIT" == false ]] && [[ "$NEEDS_NODE" == false ]] && [[ "$NEEDS_MEMVID" == false ]]; then
+        print_info "All dependencies are already installed"
+        return 0
+    fi
+    
+    # Show what will be installed
+    echo ""
+    print_warning "The following tools will be installed:"
+    [[ "$NEEDS_GIT" == true ]] && echo "  - git"
+    [[ "$NEEDS_NODE" == true ]] && echo "  - node (LTS)"
+    [[ "$NEEDS_MEMVID" == true ]] && echo "  - memvid-cli (latest)"
+    echo ""
+    
+    # Ask for confirmation
+    # Read from /dev/tty to ensure it works when piped via curl | bash
+    if [[ -t 0 ]] && [[ -t 1 ]]; then
+        # Interactive terminal - read normally
+        read -p "Continue? [Y/n] " -n 1 -r
+        echo
+    elif [[ -c /dev/tty ]]; then
+        # Piped input - read from terminal device
+        read -p "Continue? [Y/n] " -n 1 -r < /dev/tty
+        echo
+    else
+        # No terminal available - proceed automatically (non-interactive mode)
+        print_info "No terminal detected, proceeding with installation..."
+        REPLY="Y"
+    fi
+    if [[ ! $REPLY =~ ^[Yy]$ ]] && [[ ! $REPLY == "" ]]; then
+        print_info "Installation cancelled"
+        exit 0
+    fi
+    
+    # Install missing tools
+    [[ "$NEEDS_GIT" == true ]] && install_git
+    [[ "$NEEDS_NODE" == true ]] && install_node
+    [[ "$NEEDS_MEMVID" == true ]] && install_memvid
+}
+
+# Install memvid
+install_memvid() {
+    print_info "Installing memvid globally..."
+    
+    if [[ "$OS" == "linux" ]]; then
+        # Linux requires sudo for global npm installs
+        if sudo npm install -g memvid-cli@latest; then
+            print_success "memvid installed successfully"
+        else
+            print_error "memvid installation failed"
+            exit 1
+        fi
+    else
+        # macOS typically doesn't need sudo if npm was installed via Homebrew
+        if npm install -g memvid-cli@latest; then
+            print_success "memvid installed successfully"
+        else
+            print_error "memvid installation failed"
+            exit 1
+        fi
+    fi
+}
+
+# Verify installation
+verify() {
+    print_info "Verifying installation..."
+    
+    if command_exists memvid; then
+        # Get version - output format is "memvid 2.0.131"
+        MEMVID_VERSION=$(memvid --version 2>&1 | awk '{print $2}')
+        
+        print_success "memvid is installed and accessible"
+        print_info "Version: $MEMVID_VERSION"
+        echo ""
+        print_success "Installation complete! You can now use 'memvid' command."
+    else
+        print_error "memvid verification failed"
+        print_info "The installation may have completed, but 'memvid' command is not in PATH"
+        print_info "Please check your npm global bin directory and add it to PATH if needed"
+        print_info "Or try: npm list -g memvid-cli"
+        exit 1
+    fi
+}
+
+# Main execution
+main() {
+    echo "Memvid Installer"
+    echo "Checking system requirements…"
+    echo ""
+    
+    detect_os
+    echo ""
+    
+    # Check Xcode CLI Tools on macOS
+    if [[ "$OS" == "macos" ]]; then
+        check_xcode_cli_tools
+        echo ""
+    fi
+    
+    # Install system dependencies on Linux
+    if [[ "$OS" == "linux" ]]; then
+        install_system_deps
+        echo ""
+    fi
+    
+    install_missing
+    echo ""
+    
+    verify
+}
+
+# Run main function
+main

memvid.egg-info/PKG-INFO

@@ -1,337 +0,0 @@
-Metadata-Version: 2.4
-Name: memvid
-Version: 0.1.2
-Summary: Video-based AI memory library for fast semantic search and retrieval
-Home-page: https://github.com/olow304/memvid
-Author: Memvid Team
-Author-email: team@memvid.ai
-Classifier: Development Status :: 3 - Alpha
-Classifier: Intended Audience :: Developers
-Classifier: Topic :: Software Development :: Libraries :: Python Modules
-Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
-Classifier: Topic :: Multimedia :: Video
-Classifier: License :: OSI Approved :: MIT License
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.8
-Classifier: Programming Language :: Python :: 3.9
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3.11
-Requires-Python: >=3.8
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: qrcode[pil]>=7.3
-Requires-Dist: opencv-python>=4.5.0
-Requires-Dist: opencv-contrib-python>=4.5.0
-Requires-Dist: sentence-transformers>=2.2.0
-Requires-Dist: numpy<2.0.0,>=1.21.0
-Requires-Dist: openai>=1.0.0
-Requires-Dist: tqdm>=4.50.0
-Requires-Dist: faiss-cpu>=1.7.0
-Requires-Dist: Pillow>=9.0.0
-Requires-Dist: python-dotenv>=0.19.0
-Provides-Extra: dev
-Requires-Dist: pytest>=7.0.0; extra == "dev"
-Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
-Requires-Dist: black>=23.0.0; extra == "dev"
-Requires-Dist: flake8>=6.0.0; extra == "dev"
-Provides-Extra: pdf
-Requires-Dist: PyPDF2==3.0.1; extra == "pdf"
-Provides-Extra: web
-Requires-Dist: fastapi>=0.100.0; extra == "web"
-Requires-Dist: gradio>=4.0.0; extra == "web"
-Dynamic: author
-Dynamic: author-email
-Dynamic: classifier
-Dynamic: description
-Dynamic: description-content-type
-Dynamic: home-page
-Dynamic: license-file
-Dynamic: provides-extra
-Dynamic: requires-dist
-Dynamic: requires-python
-Dynamic: summary
-
-# Memvid - Video-Based AI Memory 🧠📹
-
-**The lightweight, game-changing solution for AI memory at scale**
-
-[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
-[![Python 3.8+](https://img.shields.io/badge/python-3.8+-blue.svg)](https://www.python.org/downloads/)
-
-Memvid revolutionizes AI memory management by encoding text data into videos, enabling **lightning-fast semantic search** across millions of text chunks with **sub-second retrieval times**. Unlike traditional vector databases that consume massive amounts of RAM and storage, Memvid compresses your knowledge base into compact video files while maintaining instant access to any piece of information.
-
-## 🚀 Why Memvid?
-
-### Game-Changing Innovation
-- **Video as Database**: Store millions of text chunks in a single MP4 file
-- **Instant Retrieval**: Sub-second semantic search across massive datasets
-- **10x Storage Efficiency**: Video compression reduces memory footprint dramatically
-- **Zero Infrastructure**: No database servers, just files you can copy anywhere
-- **Offline-First**: Works completely offline once videos are generated
-
-### Lightweight Architecture
-- **Minimal Dependencies**: Core functionality in ~1000 lines of Python
-- **CPU-Friendly**: Runs efficiently without GPU requirements
-- **Portable**: Single video file contains your entire knowledge base
-- **Streamable**: Videos can be streamed from cloud storage
-
-## 📦 Installation
-
-### Prerequisites
-```bash
-# Ubuntu/Debian
-sudo apt-get install libzbar0
-
-# macOS
-brew install zbar
-
-# Windows
-# Download and install from https://sourceforge.net/projects/zbarw/
-```
-
-### Install from PyPI
-```bash
-pip install memvid
-```
-
-### Install from Source
-```bash
-git clone https://github.com/yourusername/memvid.git
-cd memvid
-pip install -e .
-```
-
-### Development Installation
-```bash
-# Create virtual environment
-python -m venv .memvid
-source .memvid/bin/activate  # On Windows: .memvid\Scripts\activate
-
-# Install with development dependencies
-pip install -e ".[dev]"
-```
-
-## 🎯 Quick Start
-
-### Basic Usage
-```python
-from memvid import MemvidEncoder, MemvidChat
-
-# Create video memory from text chunks
-chunks = ["Important fact 1", "Important fact 2", "Historical event details", ...]
-encoder = MemvidEncoder()
-encoder.add_chunks(chunks)
-encoder.build_video("memory.mp4", "memory_index.json")
-
-# Chat with your memory
-chat = MemvidChat("memory.mp4", "memory_index.json")
-chat.start_session()
-response = chat.chat("What do you know about historical events?")
-print(response)
-```
-
-### Building Memory from Documents
-```python
-from memvid import MemvidEncoder
-import os
-
-# Load documents
-encoder = MemvidEncoder(chunk_size=512, overlap=50)
-
-# Add text files
-for file in os.listdir("documents"):
-    with open(f"documents/{file}", "r") as f:
-        encoder.add_text(f.read(), metadata={"source": file})
-
-# Build optimized video
-encoder.build_video(
-    "knowledge_base.mp4",
-    "knowledge_index.json",
-    fps=30,  # Higher FPS = more chunks per second
-    frame_size=512  # Larger frames = more data per frame
-)
-```
-
-### Advanced Search & Retrieval
-```python
-from memvid import MemvidRetriever
-
-# Initialize retriever
-retriever = MemvidRetriever("knowledge_base.mp4", "knowledge_index.json")
-
-# Semantic search
-results = retriever.search("machine learning algorithms", top_k=5)
-for chunk, score in results:
-    print(f"Score: {score:.3f} | {chunk[:100]}...")
-
-# Get context window
-context = retriever.get_context("explain neural networks", max_tokens=2000)
-print(context)
-```
-
-### Interactive Chat Interface
-```python
-from memvid import MemvidInteractive
-
-# Launch interactive chat UI
-interactive = MemvidInteractive("knowledge_base.mp4", "knowledge_index.json")
-interactive.run()  # Opens web interface at http://localhost:7860
-```
-
-## 📊 Performance Benchmarks
-
-| Dataset Size | Encoding Time | Search Time | Storage Size | RAM Usage |
-|-------------|--------------|-------------|--------------|-----------|
-| 10K chunks | 45 seconds | 0.05s | 12 MB | 50 MB |
-| 100K chunks | 8 minutes | 0.15s | 120 MB | 200 MB |
-| 1M chunks | 80 minutes | 0.8s | 1.2 GB | 800 MB |
-| 10M chunks | 13 hours | 1.5s | 12 GB | 2 GB |
-
-*Benchmarked on Intel i7-9700K, 16GB RAM, SSD storage*
-
-## 🔧 API Reference
-
-### MemvidEncoder
-```python
-encoder = MemvidEncoder(
-    chunk_size=512,      # Characters per chunk
-    overlap=50,          # Character overlap between chunks
-    model_name='all-MiniLM-L6-v2'  # Sentence transformer model
-)
-
-# Methods
-encoder.add_chunks(chunks: List[str], metadata: List[dict] = None)
-encoder.add_text(text: str, metadata: dict = None)
-encoder.build_video(video_path: str, index_path: str, fps: int = 30, qr_size: int = 512)
-```
-
-### MemvidRetriever
-```python
-retriever = MemvidRetriever(
-    video_path: str,
-    index_path: str,
-    cache_size: int = 100  # Number of frames to cache
-)
-
-# Methods
-results = retriever.search(query: str, top_k: int = 5)
-context = retriever.get_context(query: str, max_tokens: int = 2000)
-chunks = retriever.get_chunks_by_ids(chunk_ids: List[int])
-```
-
-### MemvidChat
-```python
-chat = MemvidChat(
-    video_path: str,
-    index_path: str,
-    llm_backend: str = 'openai',  # 'openai', 'anthropic', 'local'
-    model: str = 'gpt-4'
-)
-
-# Methods
-chat.start_session(system_prompt: str = None)
-response = chat.chat(message: str, stream: bool = False)
-chat.clear_history()
-chat.export_conversation(path: str)
-```
-
-## 🛠️ Advanced Configuration
-
-### Custom Embeddings
-```python
-from sentence_transformers import SentenceTransformer
-
-# Use custom embedding model
-custom_model = SentenceTransformer('sentence-transformers/all-mpnet-base-v2')
-encoder = MemvidEncoder(embedding_model=custom_model)
-```
-
-### Video Optimization
-```python
-# For maximum compression
-encoder.build_video(
-    "compressed.mp4",
-    "index.json",
-    fps=60,  # More frames per second
-    frame_size=256,  # Smaller frames
-    video_codec='h265',  # Better compression
-    crf=28  # Compression quality (lower = better quality)
-)
-```
-
-### Distributed Processing
-```python
-# Process large datasets in parallel
-encoder = MemvidEncoder(n_workers=8)
-encoder.add_chunks_parallel(massive_chunk_list)
-```
-
-## 🐛 Troubleshooting
-
-### Common Issues
-
-**Video Decode Errors**
-```bash
-# Increase error correction
-encoder.build_video(..., error_correction='H')  # Highest error correction
-```
-
-**Memory Issues with Large Datasets**
-```python
-# Use streaming mode
-encoder = MemvidEncoder(streaming=True)
-for batch in chunk_generator():
-    encoder.add_chunks_batch(batch)
-```
-
-**Slow Search Performance**
-```python
-# Pre-warm cache
-retriever.warm_cache(common_queries=['AI', 'machine learning', 'data'])
-```
-
-## 🤝 Contributing
-
-We welcome contributions! Please see our [Contributing Guide](CONTRIBUTING.md) for details.
-
-```bash
-# Run tests
-pytest tests/
-
-# Run with coverage
-pytest --cov=memvid tests/
-
-# Format code
-black memvid/
-```
-
-## 📚 Examples
-
-Check out the [examples/](examples/) directory for:
-- Building memory from Wikipedia dumps
-- Creating a personal knowledge base
-- Multi-language support
-- Real-time memory updates
-- Integration with popular LLMs
-
-## 🔗 Links
-
-- [Documentation](https://memvid.readthedocs.io)
-- [PyPI Package](https://pypi.org/project/memvid)
-- [GitHub Repository](https://github.com/yourusername/memvid)
-- [Discord Community](https://discord.gg/memvid)
-
-## 📄 License
-
-MIT License - see [LICENSE](LICENSE) file for details.
-
-## 🙏 Acknowledgments
-
-Built with ❤️ using:
-- [sentence-transformers](https://www.sbert.net/) for embeddings
-- [OpenCV](https://opencv.org/) for video processing
-- [qrcode](https://github.com/lincolnloop/python-qrcode) & [pyzbar](https://github.com/NaturalHistoryMuseum/pyzbar) for data encoding
-
----
-
-**Ready to revolutionize your AI memory management? Install Memvid and start building!** 🚀

memvid.egg-info/SOURCES.txt

@@ -1,28 +0,0 @@
-LICENSE
-MANIFEST.in
-README.md
-requirements.txt
-setup.py
-examples/book_chat.py
-examples/build_memory.py
-examples/chat_memory.py
-examples/chat_memory_fixed.py
-examples/simple_chat.py
-examples/soccer_chat.py
-memvid/__init__.py
-memvid/chat.py
-memvid/config.py
-memvid/encoder.py
-memvid/index.py
-memvid/interactive.py
-memvid/retriever.py
-memvid/utils.py
-memvid.egg-info/PKG-INFO
-memvid.egg-info/SOURCES.txt
-memvid.egg-info/dependency_links.txt
-memvid.egg-info/requires.txt
-memvid.egg-info/top_level.txt
-tests/__init__.py
-tests/test_encoder.py
-tests/test_retriever.py
-tests/test_utils.py

memvid.egg-info/dependency_links.txt

@@ -1 +0,0 @@
-

memvid.egg-info/requires.txt

@@ -1,23 +0,0 @@
-qrcode[pil]>=7.3
-opencv-python>=4.5.0
-opencv-contrib-python>=4.5.0
-sentence-transformers>=2.2.0
-numpy<2.0.0,>=1.21.0
-openai>=1.0.0
-tqdm>=4.50.0
-faiss-cpu>=1.7.0
-Pillow>=9.0.0
-python-dotenv>=0.19.0
-
-[dev]
-pytest>=7.0.0
-pytest-cov>=4.0.0
-black>=23.0.0
-flake8>=6.0.0
-
-[pdf]
-PyPDF2==3.0.1
-
-[web]
-fastapi>=0.100.0
-gradio>=4.0.0

memvid.egg-info/top_level.txt

@@ -1,2 +0,0 @@
-memvid
-tests

memvid/__init__.py

@@ -1,12 +0,0 @@
-"""
-Memvid - QR Code Video-Based AI Memory Library
-"""
-
-__version__ = "0.1.0"
-
-from .encoder import MemvidEncoder
-from .retriever import MemvidRetriever
-from .chat import MemvidChat
-from .interactive import chat_with_memory, quick_chat
-
-__all__ = ["MemvidEncoder", "MemvidRetriever", "MemvidChat", "chat_with_memory", "quick_chat"]

memvid/chat.py

@@ -1,268 +0,0 @@
-"""
-MemvidChat - Conversational interface with context-aware memory
-"""
-
-import os
-import logging
-from typing import List, Dict, Any, Optional, Tuple
-from datetime import datetime
-import json
-
-from .retriever import MemvidRetriever
-from .config import get_default_config
-
-logger = logging.getLogger(__name__)
-
-
-try:
-    from openai import OpenAI
-    OPENAI_AVAILABLE = True
-except ImportError:
-    OPENAI_AVAILABLE = False
-    logger.warning("OpenAI library not available. LLM features will be limited.")
-
-
-class MemvidChat:
-    """Manages conversations with context retrieval and LLM interface"""
-    
-    def __init__(self, video_file: str, index_file: str,
-                 llm_api_key: Optional[str] = None,
-                 llm_model: Optional[str] = None,
-                 config: Optional[Dict[str, Any]] = None):
-        """
-        Initialize MemvidChat
-        
-        Args:
-            video_file: Path to QR code video
-            index_file: Path to index file
-            llm_api_key: API key for LLM (or set via environment)
-            llm_model: LLM model to use
-            config: Optional configuration
-        """
-        self.config = config or get_default_config()
-        self.retriever = MemvidRetriever(video_file, index_file, self.config)
-        
-        # Initialize LLM
-        self.llm_model = llm_model or self.config["llm"]["model"]
-        self._init_llm(llm_api_key)
-        
-        # Conversation state
-        self.conversation_history = []
-        self.session_id = None
-        self.context_chunks = self.config["chat"]["context_chunks"]
-        self.max_history = self.config["chat"]["max_history"]
-        
-    def _init_llm(self, api_key: Optional[str] = None):
-        """Initialize LLM client"""
-        if not OPENAI_AVAILABLE:
-            self.llm_client = None
-            logger.warning("OpenAI not available. Chat will return context only.")
-            return
-        
-        # Get API key from parameter, env, or config
-        api_key = api_key or os.getenv("OPENAI_API_KEY")
-        if not api_key:
-            self.llm_client = None
-            logger.warning("No OpenAI API key provided. Chat will return context only.")
-            return
-        
-        try:
-            self.llm_client = OpenAI(api_key=api_key)
-            logger.info(f"Initialized OpenAI client with model: {self.llm_model}")
-        except Exception as e:
-            self.llm_client = None
-            logger.error(f"Failed to initialize OpenAI client: {e}")
-    
-    def start_session(self, session_id: Optional[str] = None):
-        """
-        Start a new chat session
-        
-        Args:
-            session_id: Optional session ID (generates one if not provided)
-        """
-        self.session_id = session_id or f"session_{datetime.now().strftime('%Y%m%d_%H%M%S')}"
-        self.conversation_history = []
-        logger.info(f"Started chat session: {self.session_id}")
-    
-    def chat(self, user_input: str) -> str:
-        """
-        Process user input and generate response
-        
-        Args:
-            user_input: User's message
-            
-        Returns:
-            Assistant's response
-        """
-        if not self.session_id:
-            self.start_session()
-        
-        # Retrieve relevant context
-        context_chunks = self.retriever.search(user_input, top_k=self.context_chunks)
-        
-        # Build context string
-        context = "\n\n".join([f"[Context {i+1}]: {chunk}" 
-                               for i, chunk in enumerate(context_chunks)])
-        
-        # Add to history
-        self.conversation_history.append({
-            "role": "user",
-            "content": user_input,
-            "timestamp": datetime.now().isoformat()
-        })
-        
-        # Generate response
-        if self.llm_client:
-            response = self._generate_llm_response(user_input, context)
-        else:
-            # Fallback: return context only
-            response = self._generate_context_response(context_chunks)
-        
-        # Add response to history
-        self.conversation_history.append({
-            "role": "assistant",
-            "content": response,
-            "timestamp": datetime.now().isoformat(),
-            "context_used": len(context_chunks)
-        })
-        
-        # Trim history if needed
-        if len(self.conversation_history) > self.max_history * 2:
-            self.conversation_history = self.conversation_history[-self.max_history * 2:]
-        
-        return response
-    
-    def _generate_llm_response(self, user_input: str, context: str) -> str:
-        """Generate response using LLM"""
-        try:
-            # Build messages
-            messages = [
-                {
-                    "role": "system",
-                    "content": (
-                        "You are a helpful assistant with access to a knowledge base stored in video memory. "
-                        "Use the provided context chunks to answer questions accurately. "
-                        "If asked about what the context or knowledge base contains, analyze and summarize "
-                        "the topics covered based on the context chunks provided. "
-                        "Always base your answers on the given context."
-                    )
-                }
-            ]
-            
-            # Add conversation history (last N messages)
-            history_start = max(0, len(self.conversation_history) - self.max_history)
-            for msg in self.conversation_history[history_start:-1]:  # Exclude current user message
-                messages.append({
-                    "role": msg["role"],
-                    "content": msg["content"]
-                })
-            
-            # Add current query with context
-            messages.append({
-                "role": "user",
-                "content": f"Context from knowledge base:\n{context}\n\nUser question: {user_input}"
-            })
-            
-            # Generate response
-            response = self.llm_client.chat.completions.create(
-                model=self.llm_model,
-                messages=messages,
-                max_tokens=self.config["llm"]["max_tokens"],
-                temperature=self.config["llm"]["temperature"]
-            )
-            
-            return response.choices[0].message.content
-            
-        except Exception as e:
-            logger.error(f"LLM generation failed: {e}")
-            return self._generate_context_response(context.split("\n\n"))
-    
-    def _generate_context_response(self, context_chunks: List[str]) -> str:
-        """Generate response without LLM (context only)"""
-        if not context_chunks:
-            return "I couldn't find any relevant information in the knowledge base."
-        
-        # Check if the chunks are actually relevant (not just random matches)
-        # If all chunks are very short or seem unrelated, indicate low relevance
-        avg_chunk_length = sum(len(chunk) for chunk in context_chunks) / len(context_chunks)
-        if avg_chunk_length < 50:  # Likely fragment matches
-            return "I couldn't find any relevant information about that topic in the knowledge base."
-        
-        response = "Based on the knowledge base, here's what I found:\n\n"
-        for i, chunk in enumerate(context_chunks[:3]):  # Limit to top 3
-            response += f"{i+1}. {chunk[:200]}...\n\n" if len(chunk) > 200 else f"{i+1}. {chunk}\n\n"
-        
-        return response.strip()
-    
-    def search_context(self, query: str, top_k: int = 5) -> List[Dict[str, Any]]:
-        """
-        Search for context without generating a response
-        
-        Args:
-            query: Search query
-            top_k: Number of results
-            
-        Returns:
-            List of search results with metadata
-        """
-        return self.retriever.search_with_metadata(query, top_k)
-    
-    def get_history(self) -> List[Dict[str, Any]]:
-        """Get conversation history"""
-        return self.conversation_history.copy()
-    
-    def export_session(self, output_file: str):
-        """
-        Export session to file
-        
-        Args:
-            output_file: Path to output file
-        """
-        session_data = {
-            "session_id": self.session_id,
-            "start_time": self.conversation_history[0]["timestamp"] if self.conversation_history else None,
-            "end_time": self.conversation_history[-1]["timestamp"] if self.conversation_history else None,
-            "message_count": len(self.conversation_history),
-            "history": self.conversation_history,
-            "config": {
-                "llm_model": self.llm_model,
-                "context_chunks": self.context_chunks,
-                "max_history": self.max_history
-            }
-        }
-        
-        with open(output_file, 'w') as f:
-            json.dump(session_data, f, indent=2)
-        
-        logger.info(f"Exported session to: {output_file}")
-    
-    def load_session(self, session_file: str):
-        """
-        Load session from file
-        
-        Args:
-            session_file: Path to session file
-        """
-        with open(session_file, 'r') as f:
-            session_data = json.load(f)
-        
-        self.session_id = session_data["session_id"]
-        self.conversation_history = session_data["history"]
-        
-        logger.info(f"Loaded session: {self.session_id}")
-    
-    def reset_session(self):
-        """Reset conversation history"""
-        self.conversation_history = []
-        logger.info("Reset conversation history")
-    
-    def get_stats(self) -> Dict[str, Any]:
-        """Get chat statistics"""
-        return {
-            "session_id": self.session_id,
-            "message_count": len(self.conversation_history),
-            "llm_available": self.llm_client is not None,
-            "llm_model": self.llm_model,
-            "context_chunks_per_query": self.context_chunks,
-            "retriever_stats": self.retriever.get_stats()
-        }

memvid/config.py

@@ -1,94 +0,0 @@
-"""
-Configuration defaults and constants for Memvid
-"""
-
-from typing import Dict, Any
-
-# QR Code settings
-QR_VERSION = 1  # 1-40, higher = more data capacity
-QR_ERROR_CORRECTION = 'M'  # L, M, Q, H
-QR_BOX_SIZE = 10
-QR_BORDER = 4
-QR_FILL_COLOR = "black"
-QR_BACK_COLOR = "white"
-
-# Video settings
-VIDEO_FPS = 30
-VIDEO_CODEC = 'mp4v'
-FRAME_WIDTH = 512
-FRAME_HEIGHT = 512
-
-# Retrieval settings
-DEFAULT_TOP_K = 5
-BATCH_SIZE = 100
-MAX_WORKERS = 4  # For parallel processing
-CACHE_SIZE = 1000  # Number of frames to cache
-
-# Embedding settings
-EMBEDDING_MODEL = "all-MiniLM-L6-v2"  # Fast and good quality
-EMBEDDING_DIMENSION = 384
-
-# Index settings
-INDEX_TYPE = "Flat"  # Can be "IVF" for larger datasets
-NLIST = 100  # Number of clusters for IVF index
-
-# LLM settings
-DEFAULT_LLM_MODEL = "gpt-3.5-turbo"
-MAX_TOKENS = 1000
-TEMPERATURE = 0.7
-CONTEXT_WINDOW = 4096
-
-# Chat settings
-MAX_HISTORY_LENGTH = 10
-CONTEXT_CHUNKS_PER_QUERY = 5
-
-# Performance settings
-PREFETCH_FRAMES = 50
-DECODE_TIMEOUT = 10  # seconds
-
-def get_default_config() -> Dict[str, Any]:
-    """Get default configuration dictionary"""
-    return {
-        "qr": {
-            "version": QR_VERSION,
-            "error_correction": QR_ERROR_CORRECTION,
-            "box_size": QR_BOX_SIZE,
-            "border": QR_BORDER,
-            "fill_color": QR_FILL_COLOR,
-            "back_color": QR_BACK_COLOR,
-        },
-        "video": {
-            "fps": VIDEO_FPS,
-            "codec": VIDEO_CODEC,
-            "frame_width": FRAME_WIDTH,
-            "frame_height": FRAME_HEIGHT,
-        },
-        "retrieval": {
-            "top_k": DEFAULT_TOP_K,
-            "batch_size": BATCH_SIZE,
-            "max_workers": MAX_WORKERS,
-            "cache_size": CACHE_SIZE,
-        },
-        "embedding": {
-            "model": EMBEDDING_MODEL,
-            "dimension": EMBEDDING_DIMENSION,
-        },
-        "index": {
-            "type": INDEX_TYPE,
-            "nlist": NLIST,
-        },
-        "llm": {
-            "model": DEFAULT_LLM_MODEL,
-            "max_tokens": MAX_TOKENS,
-            "temperature": TEMPERATURE,
-            "context_window": CONTEXT_WINDOW,
-        },
-        "chat": {
-            "max_history": MAX_HISTORY_LENGTH,
-            "context_chunks": CONTEXT_CHUNKS_PER_QUERY,
-        },
-        "performance": {
-            "prefetch_frames": PREFETCH_FRAMES,
-            "decode_timeout": DECODE_TIMEOUT,
-        }
-    }

memvid/encoder.py

@@ -1,232 +0,0 @@
-"""
-MemvidEncoder - Handles chunking and QR video creation
-"""
-
-import json
-import logging
-from pathlib import Path
-from typing import List, Optional, Dict, Any
-from tqdm import tqdm
-import cv2
-import numpy as np
-
-from .utils import encode_to_qr, qr_to_frame, create_video_writer, chunk_text
-from .index import IndexManager
-from .config import get_default_config
-
-logger = logging.getLogger(__name__)
-
-
-class MemvidEncoder:
-    """Encodes text chunks into QR code videos with searchable index"""
-    
-    def __init__(self, config: Optional[Dict[str, Any]] = None):
-        """
-        Initialize MemvidEncoder
-        
-        Args:
-            config: Optional configuration dictionary
-        """
-        self.config = config or get_default_config()
-        self.chunks = []
-        self.index_manager = IndexManager(self.config)
-        
-    def add_chunks(self, chunks: List[str]):
-        """
-        Add text chunks to be encoded
-        
-        Args:
-            chunks: List of text chunks
-        """
-        self.chunks.extend(chunks)
-        logger.info(f"Added {len(chunks)} chunks. Total: {len(self.chunks)}")
-    
-    def add_text(self, text: str, chunk_size: int = 500, overlap: int = 50):
-        """
-        Add text and automatically chunk it
-        
-        Args:
-            text: Text to chunk and add
-            chunk_size: Target chunk size
-            overlap: Overlap between chunks
-        """
-        chunks = chunk_text(text, chunk_size, overlap)
-        self.add_chunks(chunks)
-    
-    def add_pdf(self, pdf_path: str, chunk_size: int = 800, overlap: int = 100):
-        """
-        Extract text from PDF and add as chunks
-        
-        Args:
-            pdf_path: Path to PDF file
-            chunk_size: Target chunk size (default larger for books)
-            overlap: Overlap between chunks
-        """
-        try:
-            import PyPDF2
-        except ImportError:
-            raise ImportError("PyPDF2 is required for PDF support. Install with: pip install PyPDF2")
-        
-        if not Path(pdf_path).exists():
-            raise FileNotFoundError(f"PDF file not found: {pdf_path}")
-        
-        text = ""
-        with open(pdf_path, 'rb') as file:
-            pdf_reader = PyPDF2.PdfReader(file)
-            num_pages = len(pdf_reader.pages)
-            
-            logger.info(f"Extracting text from {num_pages} pages of {Path(pdf_path).name}")
-            
-            for page_num in range(num_pages):
-                page = pdf_reader.pages[page_num]
-                page_text = page.extract_text()
-                text += page_text + "\n\n"
-        
-        if text.strip():
-            self.add_text(text, chunk_size, overlap)
-            logger.info(f"Added PDF content: {len(text)} characters from {Path(pdf_path).name}")
-        else:
-            logger.warning(f"No text extracted from PDF: {pdf_path}")
-    
-    def build_video(self, output_file: str, index_file: str, 
-                    show_progress: bool = True) -> Dict[str, Any]:
-        """
-        Build QR code video and index from chunks
-        
-        Args:
-            output_file: Path to output video file
-            index_file: Path to output index file
-            show_progress: Show progress bar
-            
-        Returns:
-            Dictionary with build statistics
-        """
-        if not self.chunks:
-            raise ValueError("No chunks to encode. Use add_chunks() first.")
-        
-        output_path = Path(output_file)
-        index_path = Path(index_file)
-        
-        # Ensure output directory exists
-        output_path.parent.mkdir(parents=True, exist_ok=True)
-        index_path.parent.mkdir(parents=True, exist_ok=True)
-        
-        logger.info(f"Building video with {len(self.chunks)} chunks")
-        
-        # Create video writer
-        video_config = self.config["video"]
-        writer = create_video_writer(str(output_path), video_config)
-        
-        frame_numbers = []
-        
-        try:
-            # Generate QR codes and write to video
-            chunks_iter = enumerate(self.chunks)
-            if show_progress:
-                chunks_iter = tqdm(chunks_iter, total=len(self.chunks), desc="Encoding chunks to video")
-            
-            for frame_num, chunk in chunks_iter:
-                # Create metadata for chunk
-                chunk_data = {
-                    "id": frame_num,
-                    "text": chunk,
-                    "frame": frame_num
-                }
-                
-                # Encode to QR
-                qr_image = encode_to_qr(json.dumps(chunk_data), self.config)
-                
-                # Convert to video frame
-                frame = qr_to_frame(qr_image, (video_config["frame_width"], video_config["frame_height"]))
-                
-                # Write frame
-                writer.write(frame)
-                frame_numbers.append(frame_num)
-            
-            # Add chunks to index
-            logger.info("Building search index...")
-            self.index_manager.add_chunks(self.chunks, frame_numbers, show_progress)
-            
-            # Save index
-            self.index_manager.save(str(index_path.with_suffix('')))
-            
-            # Get statistics
-            stats = {
-                "total_chunks": len(self.chunks),
-                "total_frames": len(frame_numbers),
-                "video_file": str(output_path),
-                "index_file": str(index_path),
-                "video_size_mb": output_path.stat().st_size / (1024 * 1024) if output_path.exists() else 0,
-                "fps": video_config["fps"],
-                "duration_seconds": len(frame_numbers) / video_config["fps"],
-                "index_stats": self.index_manager.get_stats()
-            }
-            
-            logger.info(f"Successfully built video: {output_path}")
-            logger.info(f"Video duration: {stats['duration_seconds']:.1f} seconds")
-            logger.info(f"Video size: {stats['video_size_mb']:.1f} MB")
-            
-            return stats
-            
-        finally:
-            writer.release()
-    
-    def clear(self):
-        """Clear all chunks"""
-        self.chunks = []
-        self.index_manager = IndexManager(self.config)
-        logger.info("Cleared all chunks")
-    
-    def get_stats(self) -> Dict[str, Any]:
-        """Get encoder statistics"""
-        return {
-            "total_chunks": len(self.chunks),
-            "total_characters": sum(len(chunk) for chunk in self.chunks),
-            "avg_chunk_size": np.mean([len(chunk) for chunk in self.chunks]) if self.chunks else 0,
-            "config": self.config
-        }
-    
-    @classmethod
-    def from_file(cls, file_path: str, chunk_size: int = 500, 
-                  overlap: int = 50, config: Optional[Dict[str, Any]] = None) -> 'MemvidEncoder':
-        """
-        Create encoder from text file
-        
-        Args:
-            file_path: Path to text file
-            chunk_size: Target chunk size
-            overlap: Overlap between chunks
-            config: Optional configuration
-            
-        Returns:
-            MemvidEncoder instance with chunks loaded
-        """
-        encoder = cls(config)
-        
-        with open(file_path, 'r', encoding='utf-8') as f:
-            text = f.read()
-        
-        encoder.add_text(text, chunk_size, overlap)
-        return encoder
-    
-    @classmethod
-    def from_documents(cls, documents: List[str], chunk_size: int = 500,
-                      overlap: int = 50, config: Optional[Dict[str, Any]] = None) -> 'MemvidEncoder':
-        """
-        Create encoder from list of documents
-        
-        Args:
-            documents: List of document strings
-            chunk_size: Target chunk size
-            overlap: Overlap between chunks
-            config: Optional configuration
-            
-        Returns:
-            MemvidEncoder instance with chunks loaded
-        """
-        encoder = cls(config)
-        
-        for doc in documents:
-            encoder.add_text(doc, chunk_size, overlap)
-        
-        return encoder

memvid/index.py

@@ -1,216 +0,0 @@
-"""
-Index management for embeddings and vector search
-"""
-
-import json
-import numpy as np
-import faiss
-from sentence_transformers import SentenceTransformer
-from typing import List, Dict, Any, Tuple, Optional
-import logging
-from pathlib import Path
-import pickle
-from tqdm import tqdm
-
-from .config import get_default_config
-
-logger = logging.getLogger(__name__)
-
-
-class IndexManager:
-    """Manages embeddings, FAISS index, and metadata for fast retrieval"""
-    
-    def __init__(self, config: Optional[Dict[str, Any]] = None):
-        """
-        Initialize IndexManager
-        
-        Args:
-            config: Optional configuration dictionary
-        """
-        self.config = config or get_default_config()
-        self.embedding_model = SentenceTransformer(self.config["embedding"]["model"])
-        self.dimension = self.config["embedding"]["dimension"]
-        
-        # Initialize FAISS index
-        self.index = self._create_index()
-        
-        # Metadata storage
-        self.metadata = []
-        self.chunk_to_frame = {}  # Maps chunk ID to frame number
-        self.frame_to_chunks = {}  # Maps frame number to chunk IDs
-        
-    def _create_index(self) -> faiss.Index:
-        """Create FAISS index based on configuration"""
-        index_type = self.config["index"]["type"]
-        
-        if index_type == "Flat":
-            # Exact search - best quality, slower for large datasets
-            index = faiss.IndexFlatL2(self.dimension)
-        elif index_type == "IVF":
-            # Inverted file index - faster for large datasets
-            quantizer = faiss.IndexFlatL2(self.dimension)
-            index = faiss.IndexIVFFlat(quantizer, self.dimension, self.config["index"]["nlist"])
-        else:
-            raise ValueError(f"Unknown index type: {index_type}")
-            
-        # Add ID mapping for retrieval
-        index = faiss.IndexIDMap(index)
-        return index
-    
-    def add_chunks(self, chunks: List[str], frame_numbers: List[int], 
-                   show_progress: bool = True) -> List[int]:
-        """
-        Add chunks to index with their frame mappings
-        
-        Args:
-            chunks: List of text chunks
-            frame_numbers: Corresponding frame numbers for each chunk
-            show_progress: Show progress bar
-            
-        Returns:
-            List of chunk IDs
-        """
-        if len(chunks) != len(frame_numbers):
-            raise ValueError("Number of chunks must match number of frame numbers")
-        
-        # Generate embeddings
-        logger.info(f"Generating embeddings for {len(chunks)} chunks")
-        
-        # SentenceTransformer expects a list, not an iterator
-        embeddings = self.embedding_model.encode(
-            chunks, 
-            show_progress_bar=show_progress,
-            batch_size=32
-        )
-        embeddings = np.array(embeddings).astype('float32')
-        
-        # Assign IDs
-        start_id = len(self.metadata)
-        chunk_ids = list(range(start_id, start_id + len(chunks)))
-        
-        # Train index if needed (for IVF)
-        if isinstance(self.index.index, faiss.IndexIVFFlat) and not self.index.index.is_trained:
-            logger.info("Training FAISS index...")
-            self.index.index.train(embeddings)
-        
-        # Add to index
-        self.index.add_with_ids(embeddings, np.array(chunk_ids, dtype=np.int64))
-        
-        # Store metadata
-        for i, (chunk, frame_num, chunk_id) in enumerate(zip(chunks, frame_numbers, chunk_ids)):
-            metadata = {
-                "id": chunk_id,
-                "text": chunk,
-                "frame": frame_num,
-                "length": len(chunk)
-            }
-            self.metadata.append(metadata)
-            
-            # Update mappings
-            self.chunk_to_frame[chunk_id] = frame_num
-            if frame_num not in self.frame_to_chunks:
-                self.frame_to_chunks[frame_num] = []
-            self.frame_to_chunks[frame_num].append(chunk_id)
-        
-        logger.info(f"Added {len(chunks)} chunks to index")
-        return chunk_ids
-    
-    def search(self, query: str, top_k: int = 5) -> List[Tuple[int, float, Dict[str, Any]]]:
-        """
-        Search for similar chunks
-        
-        Args:
-            query: Search query
-            top_k: Number of results to return
-            
-        Returns:
-            List of (chunk_id, distance, metadata) tuples
-        """
-        # Generate query embedding
-        query_embedding = self.embedding_model.encode([query])
-        query_embedding = np.array(query_embedding).astype('float32')
-        
-        # Search
-        distances, indices = self.index.search(query_embedding, top_k)
-        
-        # Gather results
-        results = []
-        for dist, idx in zip(distances[0], indices[0]):
-            if idx >= 0:  # Valid result
-                metadata = self.metadata[idx]
-                results.append((idx, float(dist), metadata))
-        
-        return results
-    
-    def get_chunks_by_frame(self, frame_number: int) -> List[Dict[str, Any]]:
-        """Get all chunks associated with a frame"""
-        chunk_ids = self.frame_to_chunks.get(frame_number, [])
-        return [self.metadata[chunk_id] for chunk_id in chunk_ids]
-    
-    def get_chunk_by_id(self, chunk_id: int) -> Optional[Dict[str, Any]]:
-        """Get chunk metadata by ID"""
-        if 0 <= chunk_id < len(self.metadata):
-            return self.metadata[chunk_id]
-        return None
-    
-    def save(self, path: str):
-        """
-        Save index to disk
-        
-        Args:
-            path: Path to save index (without extension)
-        """
-        path = Path(path)
-        
-        # Save FAISS index
-        faiss.write_index(self.index, str(path.with_suffix('.faiss')))
-        
-        # Save metadata and mappings
-        data = {
-            "metadata": self.metadata,
-            "chunk_to_frame": self.chunk_to_frame,
-            "frame_to_chunks": self.frame_to_chunks,
-            "config": self.config
-        }
-        
-        with open(path.with_suffix('.json'), 'w') as f:
-            json.dump(data, f, indent=2)
-        
-        logger.info(f"Saved index to {path}")
-    
-    def load(self, path: str):
-        """
-        Load index from disk
-        
-        Args:
-            path: Path to load index from (without extension)
-        """
-        path = Path(path)
-        
-        # Load FAISS index
-        self.index = faiss.read_index(str(path.with_suffix('.faiss')))
-        
-        # Load metadata and mappings
-        with open(path.with_suffix('.json'), 'r') as f:
-            data = json.load(f)
-        
-        self.metadata = data["metadata"]
-        self.chunk_to_frame = {int(k): v for k, v in data["chunk_to_frame"].items()}
-        self.frame_to_chunks = {int(k): v for k, v in data["frame_to_chunks"].items()}
-        
-        # Update config if available
-        if "config" in data:
-            self.config.update(data["config"])
-        
-        logger.info(f"Loaded index from {path}")
-    
-    def get_stats(self) -> Dict[str, Any]:
-        """Get index statistics"""
-        return {
-            "total_chunks": len(self.metadata),
-            "total_frames": len(self.frame_to_chunks),
-            "index_type": self.config["index"]["type"],
-            "embedding_model": self.config["embedding"]["model"],
-            "dimension": self.dimension,
-            "avg_chunks_per_frame": np.mean([len(chunks) for chunks in self.frame_to_chunks.values()]) if self.frame_to_chunks else 0
-        }

memvid/interactive.py

@@ -1,181 +0,0 @@
-"""
-Interactive chat interface for Memvid
-"""
-
-import os
-import time
-from typing import Optional, Dict, Any
-from .chat import MemvidChat
-
-
-def chat_with_memory(
-    video_file: str,
-    index_file: str,
-    api_key: Optional[str] = None,
-    llm_model: Optional[str] = None,
-    show_stats: bool = True,
-    export_on_exit: bool = True,
-    session_dir: Optional[str] = None,
-    config: Optional[Dict[str, Any]] = None
-):
-    """
-    Start an interactive chat session with a video memory.
-    
-    Args:
-        video_file: Path to QR code video
-        index_file: Path to index file
-        api_key: OpenAI API key (or set OPENAI_API_KEY env var)
-        llm_model: LLM model to use (default: gpt-3.5-turbo)
-        show_stats: Show memory stats on startup
-        export_on_exit: Auto-export conversation on exit
-        session_dir: Directory to save session files (default: "output")
-        config: Optional configuration
-        
-    Commands:
-        - 'search <query>': Show raw search results
-        - 'stats': Show system statistics
-        - 'export': Save conversation
-        - 'clear': Clear conversation history
-        - 'help': Show commands
-        - 'exit' or 'quit': End session
-        
-    Example:
-        >>> from memvid import chat_with_memory
-        >>> chat_with_memory("knowledge.mp4", "knowledge_index.json")
-    """
-    # Set tokenizers parallelism to avoid warning
-    os.environ['TOKENIZERS_PARALLELISM'] = 'false'
-    
-    # Set default session directory
-    if session_dir is None:
-        session_dir = "output"
-    os.makedirs(session_dir, exist_ok=True)
-    
-    # Check if files exist
-    if not os.path.exists(video_file):
-        print(f"Error: Video file not found: {video_file}")
-        return
-    if not os.path.exists(index_file):
-        print(f"Error: Index file not found: {index_file}")
-        return
-    
-    # Initialize chat
-    api_key = api_key or os.getenv("OPENAI_API_KEY")
-    
-    print("Initializing Memvid Chat...")
-    chat = MemvidChat(video_file, index_file, llm_api_key=api_key, llm_model=llm_model, config=config)
-    chat.start_session()
-    
-    # Show stats if requested
-    if show_stats:
-        stats = chat.get_stats()
-        print(f"\nMemory loaded: {stats['retriever_stats']['index_stats']['total_chunks']} chunks")
-        if stats['llm_available']:
-            print(f"LLM: {stats['llm_model']}")
-        else:
-            print("LLM: Not available (context-only mode)")
-    
-    print("\nType 'help' for commands, 'exit' to quit")
-    print("-" * 50)
-    
-    # Interactive loop
-    while True:
-        try:
-            user_input = input("\nYou: ").strip()
-            
-            if not user_input:
-                continue
-                
-            # Handle commands
-            lower_input = user_input.lower()
-            
-            if lower_input in ['exit', 'quit', 'q']:
-                break
-                
-            elif lower_input == 'help':
-                print("\nCommands:")
-                print("  search <query> - Show raw search results")
-                print("  stats         - Show system statistics")
-                print("  export        - Save conversation")
-                print("  clear         - Clear conversation history")
-                print("  help          - Show this help")
-                print("  exit/quit     - End session")
-                continue
-                
-            elif lower_input == 'stats':
-                stats = chat.get_stats()
-                print(f"\nMessages: {stats['message_count']}")
-                print(f"Cache size: {stats['retriever_stats']['cache_size']}")
-                print(f"Video frames: {stats['retriever_stats']['total_frames']}")
-                continue
-                
-            elif lower_input == 'export':
-                export_file = os.path.join(session_dir, f"memvid_session_{chat.session_id}.json")
-                chat.export_session(export_file)
-                print(f"Exported to: {export_file}")
-                continue
-                
-            elif lower_input == 'clear':
-                chat.reset_session()
-                chat.start_session()
-                print("Conversation cleared.")
-                continue
-                
-            elif lower_input.startswith('search '):
-                query = user_input[7:]
-                print(f"\nSearching: '{query}'")
-                start_time = time.time()
-                results = chat.search_context(query, top_k=5)
-                elapsed = time.time() - start_time
-                print(f"Found {len(results)} results in {elapsed:.3f}s:\n")
-                for i, result in enumerate(results[:3]):
-                    print(f"{i+1}. [Score: {result['score']:.3f}] {result['text'][:100]}...")
-                continue
-            
-            # Regular chat
-            print("\nAssistant: ", end="", flush=True)
-            start_time = time.time()
-            response = chat.chat(user_input)
-            elapsed = time.time() - start_time
-            
-            print(response)
-            print(f"\n[{elapsed:.1f}s]", end="")
-            
-        except KeyboardInterrupt:
-            print("\n\nInterrupted.")
-            break
-        except Exception as e:
-            print(f"\nError: {e}")
-            continue
-    
-    # Export on exit if requested
-    if export_on_exit and chat.get_history():
-        export_file = os.path.join(session_dir, f"memvid_session_{chat.session_id}.json")
-        chat.export_session(export_file)
-        print(f"\nSession saved to: {export_file}")
-    
-    print("Goodbye!")
-
-
-def quick_chat(video_file: str, index_file: str, query: str, api_key: Optional[str] = None) -> str:
-    """
-    Quick one-off query without interactive loop.
-    
-    Args:
-        video_file: Path to QR code video
-        index_file: Path to index file
-        query: Question to ask
-        api_key: OpenAI API key (optional)
-        
-    Returns:
-        Response string
-        
-    Example:
-        >>> from memvid import quick_chat
-        >>> response = quick_chat("knowledge.mp4", "knowledge_index.json", "What is quantum computing?")
-        >>> print(response)
-    """
-    os.environ['TOKENIZERS_PARALLELISM'] = 'false'
-    
-    chat = MemvidChat(video_file, index_file, llm_api_key=api_key)
-    return chat.chat(query)

memvid/retriever.py

@@ -1,287 +0,0 @@
-"""
-MemvidRetriever - Fast semantic search, QR frame extraction, and context assembly
-"""
-
-import json
-import logging
-from pathlib import Path
-from typing import List, Dict, Any, Tuple, Optional
-from concurrent.futures import ThreadPoolExecutor, as_completed
-import time
-from functools import lru_cache
-import cv2
-
-from .utils import (
-    extract_frame, decode_qr, batch_extract_and_decode,
-    extract_and_decode_cached
-)
-from .index import IndexManager
-from .config import get_default_config
-
-logger = logging.getLogger(__name__)
-
-
-class MemvidRetriever:
-    """Fast retrieval from QR code videos using semantic search"""
-    
-    def __init__(self, video_file: str, index_file: str, 
-                 config: Optional[Dict[str, Any]] = None):
-        """
-        Initialize MemvidRetriever
-        
-        Args:
-            video_file: Path to QR code video
-            index_file: Path to index file
-            config: Optional configuration
-        """
-        self.video_file = str(Path(video_file).absolute())
-        self.index_file = str(Path(index_file).absolute())
-        self.config = config or get_default_config()
-        
-        # Load index
-        self.index_manager = IndexManager(self.config)
-        self.index_manager.load(str(Path(index_file).with_suffix('')))
-        
-        # Cache for decoded frames
-        self._frame_cache = {}
-        self._cache_size = self.config["retrieval"]["cache_size"]
-        
-        # Verify video file
-        self._verify_video()
-        
-        logger.info(f"Initialized retriever with {self.index_manager.get_stats()['total_chunks']} chunks")
-    
-    def _verify_video(self):
-        """Verify video file is accessible"""
-        cap = cv2.VideoCapture(self.video_file)
-        if not cap.isOpened():
-            raise ValueError(f"Cannot open video file: {self.video_file}")
-        
-        self.total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
-        self.fps = cap.get(cv2.CAP_PROP_FPS)
-        cap.release()
-        
-        logger.info(f"Video has {self.total_frames} frames at {self.fps} fps")
-    
-    def search(self, query: str, top_k: int = 5) -> List[str]:
-        """
-        Search for relevant chunks using semantic search
-        
-        Args:
-            query: Search query
-            top_k: Number of results to return
-            
-        Returns:
-            List of relevant text chunks
-        """
-        start_time = time.time()
-        
-        # Semantic search in index
-        search_results = self.index_manager.search(query, top_k)
-        
-        # Extract unique frame numbers
-        frame_numbers = list(set(result[2]["frame"] for result in search_results))
-        
-        # Decode frames in parallel
-        decoded_frames = self._decode_frames_parallel(frame_numbers)
-        
-        # Extract text from decoded data
-        results = []
-        for chunk_id, distance, metadata in search_results:
-            frame_num = metadata["frame"]
-            if frame_num in decoded_frames:
-                try:
-                    chunk_data = json.loads(decoded_frames[frame_num])
-                    results.append(chunk_data["text"])
-                except (json.JSONDecodeError, KeyError):
-                    # Fallback to metadata text
-                    results.append(metadata["text"])
-            else:
-                # Fallback to metadata text if decode failed
-                results.append(metadata["text"])
-        
-        elapsed = time.time() - start_time
-        logger.info(f"Search completed in {elapsed:.3f}s for query: '{query[:50]}...'")
-        
-        return results
-    
-    def get_chunk_by_id(self, chunk_id: int) -> Optional[str]:
-        """
-        Get specific chunk by ID
-        
-        Args:
-            chunk_id: Chunk ID
-            
-        Returns:
-            Chunk text or None
-        """
-        metadata = self.index_manager.get_chunk_by_id(chunk_id)
-        if metadata:
-            frame_num = metadata["frame"]
-            decoded = self._decode_single_frame(frame_num)
-            if decoded:
-                try:
-                    chunk_data = json.loads(decoded)
-                    return chunk_data["text"]
-                except (json.JSONDecodeError, KeyError):
-                    pass
-            return metadata["text"]
-        return None
-    
-    def _decode_single_frame(self, frame_number: int) -> Optional[str]:
-        """Decode single frame with caching"""
-        # Check cache
-        if frame_number in self._frame_cache:
-            return self._frame_cache[frame_number]
-        
-        # Decode frame
-        result = extract_and_decode_cached(self.video_file, frame_number)
-        
-        # Update cache
-        if result and len(self._frame_cache) < self._cache_size:
-            self._frame_cache[frame_number] = result
-        
-        return result
-    
-    def _decode_frames_parallel(self, frame_numbers: List[int]) -> Dict[int, str]:
-        """
-        Decode multiple frames in parallel
-        
-        Args:
-            frame_numbers: List of frame numbers to decode
-            
-        Returns:
-            Dict mapping frame number to decoded data
-        """
-        # Check cache first
-        results = {}
-        uncached_frames = []
-        
-        for frame_num in frame_numbers:
-            if frame_num in self._frame_cache:
-                results[frame_num] = self._frame_cache[frame_num]
-            else:
-                uncached_frames.append(frame_num)
-        
-        if not uncached_frames:
-            return results
-        
-        # Decode uncached frames in parallel
-        max_workers = self.config["retrieval"]["max_workers"]
-        decoded = batch_extract_and_decode(
-            self.video_file, 
-            uncached_frames, 
-            max_workers=max_workers
-        )
-        
-        # Update results and cache
-        for frame_num, data in decoded.items():
-            results[frame_num] = data
-            if len(self._frame_cache) < self._cache_size:
-                self._frame_cache[frame_num] = data
-        
-        return results
-    
-    def search_with_metadata(self, query: str, top_k: int = 5) -> List[Dict[str, Any]]:
-        """
-        Search with full metadata
-        
-        Args:
-            query: Search query
-            top_k: Number of results
-            
-        Returns:
-            List of result dictionaries with text, score, and metadata
-        """
-        start_time = time.time()
-        
-        # Semantic search
-        search_results = self.index_manager.search(query, top_k)
-        
-        # Extract frame numbers
-        frame_numbers = list(set(result[2]["frame"] for result in search_results))
-        
-        # Decode frames
-        decoded_frames = self._decode_frames_parallel(frame_numbers)
-        
-        # Build results with metadata
-        results = []
-        for chunk_id, distance, metadata in search_results:
-            frame_num = metadata["frame"]
-            
-            # Get text from decoded frame or metadata
-            if frame_num in decoded_frames:
-                try:
-                    chunk_data = json.loads(decoded_frames[frame_num])
-                    text = chunk_data["text"]
-                except (json.JSONDecodeError, KeyError):
-                    text = metadata["text"]
-            else:
-                text = metadata["text"]
-            
-            results.append({
-                "text": text,
-                "score": 1.0 / (1.0 + distance),  # Convert distance to similarity score
-                "chunk_id": chunk_id,
-                "frame": frame_num,
-                "metadata": metadata
-            })
-        
-        elapsed = time.time() - start_time
-        logger.info(f"Search with metadata completed in {elapsed:.3f}s")
-        
-        return results
-    
-    def get_context_window(self, chunk_id: int, window_size: int = 2) -> List[str]:
-        """
-        Get chunk with surrounding context
-        
-        Args:
-            chunk_id: Central chunk ID
-            window_size: Number of chunks before/after to include
-            
-        Returns:
-            List of chunks in context window
-        """
-        chunks = []
-        
-        # Get chunks in window
-        for offset in range(-window_size, window_size + 1):
-            target_id = chunk_id + offset
-            chunk = self.get_chunk_by_id(target_id)
-            if chunk:
-                chunks.append(chunk)
-        
-        return chunks
-    
-    def prefetch_frames(self, frame_numbers: List[int]):
-        """
-        Prefetch frames into cache for faster retrieval
-        
-        Args:
-            frame_numbers: List of frame numbers to prefetch
-        """
-        # Only prefetch frames not in cache
-        to_prefetch = [f for f in frame_numbers if f not in self._frame_cache]
-        
-        if to_prefetch:
-            logger.info(f"Prefetching {len(to_prefetch)} frames...")
-            decoded = self._decode_frames_parallel(to_prefetch)
-            logger.info(f"Prefetched {len(decoded)} frames")
-    
-    def clear_cache(self):
-        """Clear frame cache"""
-        self._frame_cache.clear()
-        extract_and_decode_cached.cache_clear()
-        logger.info("Cleared frame cache")
-    
-    def get_stats(self) -> Dict[str, Any]:
-        """Get retriever statistics"""
-        return {
-            "video_file": self.video_file,
-            "total_frames": self.total_frames,
-            "fps": self.fps,
-            "cache_size": len(self._frame_cache),
-            "max_cache_size": self._cache_size,
-            "index_stats": self.index_manager.get_stats()
-        }

memvid/utils.py

@@ -1,301 +0,0 @@
-"""
-Shared utility functions for Memvid
-"""
-
-import io
-import json
-import qrcode
-import cv2
-import numpy as np
-from PIL import Image
-from typing import List, Tuple, Optional, Dict, Any
-from concurrent.futures import ThreadPoolExecutor, ProcessPoolExecutor
-from functools import lru_cache
-import logging
-from tqdm import tqdm
-import base64
-import gzip
-
-from .config import get_default_config
-
-logger = logging.getLogger(__name__)
-
-
-def encode_to_qr(data: str, config: Optional[Dict[str, Any]] = None) -> Image.Image:
-    """
-    Encode data to QR code image
-    
-    Args:
-        data: String data to encode
-        config: Optional QR configuration
-        
-    Returns:
-        PIL Image of QR code
-    """
-    if config is None:
-        config = get_default_config()["qr"]
-    else:
-        # If config is provided but incomplete, merge with defaults
-        default_qr_config = get_default_config()["qr"]
-        config = {**default_qr_config, **config.get("qr", config)}
-    
-    # Compress data if it's large
-    if len(data) > 100:
-        compressed = gzip.compress(data.encode())
-        data = base64.b64encode(compressed).decode()
-        data = "GZ:" + data  # Prefix to indicate compression
-    
-    qr = qrcode.QRCode(
-        version=config["version"],
-        error_correction=getattr(qrcode.constants, f"ERROR_CORRECT_{config['error_correction']}"),
-        box_size=config["box_size"],
-        border=config["border"],
-    )
-    
-    qr.add_data(data)
-    qr.make(fit=True)
-    
-    img = qr.make_image(fill_color=config["fill_color"], back_color=config["back_color"])
-    return img
-
-
-def decode_qr(image: np.ndarray) -> Optional[str]:
-    """
-    Decode QR code from image
-    
-    Args:
-        image: OpenCV image array
-        
-    Returns:
-        Decoded string or None if decode fails
-    """
-    try:
-        # Initialize OpenCV QR code detector
-        detector = cv2.QRCodeDetector()
-        
-        # Detect and decode
-        data, bbox, straight_qrcode = detector.detectAndDecode(image)
-        
-        if data:
-            # Check if data was compressed
-            if data.startswith("GZ:"):
-                compressed_data = base64.b64decode(data[3:])
-                data = gzip.decompress(compressed_data).decode()
-            
-            return data
-    except Exception as e:
-        logger.warning(f"QR decode failed: {e}")
-    return None
-
-
-def create_video_writer(output_path: str, config: Optional[Dict[str, Any]] = None) -> cv2.VideoWriter:
-    """
-    Create OpenCV video writer
-    
-    Args:
-        output_path: Path to output video file
-        config: Optional video configuration
-        
-    Returns:
-        cv2.VideoWriter instance
-    """
-    if config is None:
-        config = get_default_config()["video"]
-    
-    fourcc = cv2.VideoWriter_fourcc(*config["codec"])
-    return cv2.VideoWriter(
-        output_path,
-        fourcc,
-        config["fps"],
-        (config["frame_width"], config["frame_height"])
-    )
-
-
-def qr_to_frame(qr_image: Image.Image, frame_size: Tuple[int, int]) -> np.ndarray:
-    """
-    Convert QR PIL image to video frame
-    
-    Args:
-        qr_image: PIL Image of QR code
-        frame_size: Target frame size (width, height)
-        
-    Returns:
-        OpenCV frame array
-    """
-    # Resize to fit frame while maintaining aspect ratio
-    qr_image = qr_image.resize(frame_size, Image.Resampling.LANCZOS)
-    
-    # Convert to RGB mode if necessary (handles L, P, etc. modes)
-    if qr_image.mode != 'RGB':
-        qr_image = qr_image.convert('RGB')
-    
-    # Convert to numpy array and ensure proper dtype
-    img_array = np.array(qr_image, dtype=np.uint8)
-    
-    # Convert to OpenCV format
-    frame = cv2.cvtColor(img_array, cv2.COLOR_RGB2BGR)
-    return frame
-
-
-def extract_frame(video_path: str, frame_number: int) -> Optional[np.ndarray]:
-    """
-    Extract single frame from video
-    
-    Args:
-        video_path: Path to video file
-        frame_number: Frame index to extract
-        
-    Returns:
-        OpenCV frame array or None
-    """
-    cap = cv2.VideoCapture(video_path)
-    try:
-        cap.set(cv2.CAP_PROP_POS_FRAMES, frame_number)
-        ret, frame = cap.read()
-        if ret:
-            return frame
-    finally:
-        cap.release()
-    return None
-
-
-@lru_cache(maxsize=1000)
-def extract_and_decode_cached(video_path: str, frame_number: int) -> Optional[str]:
-    """
-    Extract and decode frame with caching
-    """
-    frame = extract_frame(video_path, frame_number)
-    if frame is not None:
-        return decode_qr(frame)
-    return None
-
-
-def batch_extract_frames(video_path: str, frame_numbers: List[int], 
-                        max_workers: int = 4) -> List[Tuple[int, Optional[np.ndarray]]]:
-    """
-    Extract multiple frames in parallel
-    
-    Args:
-        video_path: Path to video file
-        frame_numbers: List of frame indices
-        max_workers: Number of parallel workers
-        
-    Returns:
-        List of (frame_number, frame) tuples
-    """
-    results = []
-    
-    # Sort frame numbers for sequential access
-    sorted_frames = sorted(frame_numbers)
-    
-    cap = cv2.VideoCapture(video_path)
-    try:
-        for frame_num in sorted_frames:
-            cap.set(cv2.CAP_PROP_POS_FRAMES, frame_num)
-            ret, frame = cap.read()
-            results.append((frame_num, frame if ret else None))
-    finally:
-        cap.release()
-    
-    return results
-
-
-def parallel_decode_qr(frames: List[Tuple[int, np.ndarray]], 
-                      max_workers: int = 4) -> List[Tuple[int, Optional[str]]]:
-    """
-    Decode multiple QR frames in parallel
-    
-    Args:
-        frames: List of (frame_number, frame) tuples
-        max_workers: Number of parallel workers
-        
-    Returns:
-        List of (frame_number, decoded_data) tuples
-    """
-    def decode_frame(item):
-        frame_num, frame = item
-        if frame is not None:
-            data = decode_qr(frame)
-            return (frame_num, data)
-        return (frame_num, None)
-    
-    with ThreadPoolExecutor(max_workers=max_workers) as executor:
-        results = list(executor.map(decode_frame, frames))
-    
-    return results
-
-
-def batch_extract_and_decode(video_path: str, frame_numbers: List[int], 
-                            max_workers: int = 4, show_progress: bool = False) -> Dict[int, str]:
-    """
-    Extract and decode multiple frames efficiently
-    
-    Args:
-        video_path: Path to video file
-        frame_numbers: List of frame indices
-        max_workers: Number of parallel workers
-        show_progress: Show progress bar
-        
-    Returns:
-        Dict mapping frame_number to decoded data
-    """
-    # Extract frames
-    frames = batch_extract_frames(video_path, frame_numbers)
-    
-    # Decode in parallel
-    if show_progress:
-        frames = tqdm(frames, desc="Decoding QR frames")
-    
-    decoded = parallel_decode_qr(frames, max_workers)
-    
-    # Build result dict
-    result = {}
-    for frame_num, data in decoded:
-        if data is not None:
-            result[frame_num] = data
-    
-    return result
-
-
-def chunk_text(text: str, chunk_size: int = 500, overlap: int = 50) -> List[str]:
-    """
-    Split text into overlapping chunks
-    
-    Args:
-        text: Text to chunk
-        chunk_size: Target chunk size in characters
-        overlap: Overlap between chunks
-        
-    Returns:
-        List of text chunks
-    """
-    chunks = []
-    start = 0
-    
-    while start < len(text):
-        end = start + chunk_size
-        chunk = text[start:end]
-        
-        # Try to break at sentence boundary
-        if end < len(text):
-            last_period = chunk.rfind('.')
-            if last_period > chunk_size * 0.8:
-                end = start + last_period + 1
-                chunk = text[start:end]
-        
-        chunks.append(chunk.strip())
-        start = end - overlap
-    
-    return chunks
-
-
-def save_index(index_data: Dict[str, Any], output_path: str):
-    """Save index data to JSON file"""
-    with open(output_path, 'w') as f:
-        json.dump(index_data, f, indent=2)
-
-
-def load_index(index_path: str) -> Dict[str, Any]:
-    """Load index data from JSON file"""
-    with open(index_path, 'r') as f:
-        return json.load(f)

prompt.md

@@ -1,169 +0,0 @@
-Memvid.ai Library Structure
-Project Root Structure
-graphql
-Copy
-Edit
-memvid/
-│
-├── memvid/
-│   ├── __init__.py
-│   ├── encoder.py         # Handles chunking and QR video creation
-│   ├── retriever.py       # Fast semantic search, QR frame extraction, context assembly
-│   ├── chat.py            # MemvidChat class: manages conversations, context retrieval, LLM interface
-│   ├── index.py           # Index management, embedding, storage, and lookup
-│   ├── utils.py           # Shared helper functions (QR encode/decode, video I/O, batching, etc.)
-│   └── config.py          # Config defaults, models, parameters
-│
-├── examples/
-│   ├── build_memory.py    # Example: Create video & index from text
-│   ├── chat_memory.py     # Example: Interactive conversation using MemvidChat
-│
-├── tests/
-│   ├── test_encoder.py
-│   ├── test_retriever.py
-│   └── test_chat.py
-│
-├── requirements.txt
-├── setup.py
-├── README.md
-└── LICENSE
-Core Dependencies/Libraries
-qrcode & Pillow — QR code generation
-
-opencv-python — Video creation and frame extraction
-
-pyzbar — QR code decoding from images
-
-sentence-transformers — Fast semantic embeddings for context search
-
-numpy — Array and vector operations
-
-openai — (or compatible LLM client, pluggable for Claude, Gemini, etc.)
-
-tqdm — Progress bars (optional, for nice UX)
-
-fastapi or gradio — (optional, for web UI)
-
-Key Classes & Responsibilities
-MemvidEncoder
-
-.add_chunks(chunks: List[str])
-
-.build_video(output_file: str, index_file: str)
-
-MemvidRetriever
-
-.search(query: str, top_k=5) -> List[str]
-
-.get_chunk_by_id(chunk_id: int) -> str
-
-MemvidChat
-
-.start_session()
-
-.chat(user_input: str) -> str
-
-Maintains conversation history internally
-
-Handles context window management and smart context selection
-
-IndexManager (internal, used by Encoder/Retriever)
-
-Embedding generation and persistence
-
-Fast nearest-neighbor search (FAISS/Annoy/Chroma)
-
-Metadata management
-
-utils.py
-
-QR code encode/decode helpers
-
-Video frame reading/writing (batch, parallel)
-
-Caching, prefetch, and batching helpers
-
-Example Usage (for End Users)
-python
-Copy
-Edit
-from memvid import MemvidEncoder, MemvidChat
-
-# Step 1: Build your video memory from data
-chunks = [
-    "Support: User couldn't login on March 2.",
-    "Error 500 on signup page March 4.",
-    # ... more data ...
-]
-encoder = MemvidEncoder()
-encoder.add_chunks(chunks)
-encoder.build_video(output_file="memory.mp4", index_file="memory_index.json")
-
-# Step 2: Interactive chat with memory
-memchat = MemvidChat(
-    video_file="memory.mp4",
-    index_file="memory_index.json",
-    llm_api_key="sk-..."  # or set globally
-)
-memchat.start_session()
-
-while True:
-    user_msg = input("\nYou: ")
-    if user_msg.strip().lower() == "exit":
-        break
-    reply = memchat.chat(user_msg)
-    print(f"\nAssistant: {reply}")
-Prompt for Your AI Assistant/Developer
-Implement a Python library named memvid for QR code video-based AI memory.
-
-Requirements:
-
-Chunk Ingestion & QR Video Encoding:
-
-Provide MemvidEncoder to chunk large text input, generate QR images, assemble into a video (opencv-python), and save chunk metadata and sentence-transformer embeddings into an index file.
-
-Super-Fast Retrieval:
-
-MemvidRetriever must use a disk-based index (embedding + frame mapping) to support lightning-fast semantic search (vector DB: FAISS, Annoy, or Chroma recommended for large scale).
-
-Batch/parallel QR frame extraction and decoding (opencv-python, pyzbar, with thread/process pools) to ensure retrieval is sub-second even for large memories.
-
-Smart caching for hot frames and common queries.
-
-Conversational Interface:
-
-Implement MemvidChat that keeps conversation history, automatically retrieves and summarizes the best context for each user message, and pipes it to an LLM (OpenAI, Anthropic, etc.).
-
-Should support session reset and history export.
-
-Flexible LLM Backend:
-
-Allow for easy API key/config change to support OpenAI, Claude, Gemini, or even local models (plug-and-play).
-
-Robust Indexing:
-
-Build a simple but robust IndexManager that generates, saves, loads, and queries embeddings, frame numbers, and metadata.
-
-Example Scripts:
-
-Provide /examples/build_memory.py and /examples/chat_memory.py for end-to-end workflow (from data to chat).
-
-Documentation & Testing:
-
-Well-documented code, usage docs, and basic unittests in /tests.
-
-Performance Goal:
-
-Retrieval (search + QR decode) should be under 2 seconds for 1M chunks on a modern laptop.
-
-Bonus:
-
-CLI and (optionally) minimal web UI with Gradio/FastAPI.
-
-Preprocessing support (compression, deduplication, etc.).
-
-Summary
-memvid will be a professional, modular, developer-friendly library that makes building, storing, and chatting with AI memory via QR code video as simple as using an in-memory DB.
-
-End users need only call MemvidEncoder and MemvidChat, and everything else is handled behind the scenes.
-