A fast OCR-based document and image processing app powered by Nebius Token Factory and Nvidia's Nemotron-Nano model. Upload PDFs or images and extract structured content using AI—displayed in real-time with an interactive UI.

Features

Upload PDFs or Images (JPG/PNG)
OCR with nvidia/Nemotron-Nano-V2-12b
Table-style Extraction for Structured Info
Real-time Multi-page PDF Processing
OpenAI-Compatible Nebius API Integration
One-click Chat + State Reset
Inline File Preview (PDF or Image)
Streamlit-based UI

Prerequisites

Python 3.10+
Nebius Token Factory Account
Nebius API Key
uv installed (for dependency management): Install it globally:
```
pip install uv
```

Installation

Clone the repository:

git clone https://github.com/Arindam200/awesome-ai-apps.git
cd rag_apps/nvidia_ocr

Install dependencies using uv:

uv sync

Set up environment variables: Create a .env file in the project root:

NEBIUS_API_KEY=your_api_key_here

Usage

Start the Streamlit application:

uv run streamlit run app.py

Then open your browser at: http://localhost:8501

How It Works

Upload & Preview

Sidebar lets you upload a PDF or image (JPG/PNG).
Inline preview for visual feedback before processing.

OCR Processing

PDFs are split into pages and each page is converted to an image.
Images are passed to the Nebius Chat Completion API via the OpenAI-compatible client.
Responses with structured data are displayed in the app.

Model Used

Model: nvidia/Nemotron-Nano-V2-12b via Nebius
Input: Base64 image
Output: Extracted text, often structured in tables

Architecture

Frontend: Streamlit
OCR Backend: Nebius API with Nvidia Nemotron-Nano
PDF Parsing: PyMuPDF (fitz)
Env Handling: dotenv
Dependencies: Managed via uv

Contributing

Found a bug or want to improve the app? Open an issue or submit a pull request.