> [!NOTE] > 本文档由 WeHub 基于上游 README 翻译整理,属于社区翻译,非官方中文文档。 > [English](./README.en.md) · [原始项目](https://github.com/StarTrail-org/PixelRAG) · [上游 README](https://github.com/StarTrail-org/PixelRAG/blob/HEAD/README.md) > 原作者、版权与许可证归属以原始项目及本仓库 LICENSE 文件为准。

PixelRAG — 视觉检索增强生成(Visual Retrieval-Augmented Generation)

PIXELRAG: Web Screenshots Beat Text for Retrieval-Augmented Generation 的官方代码库

Yichuan Wang*, Zhifei Li*, Zirui Wang, Paul Teiletche, Lesheng Jin
Matei Zaharia†, Joseph E. Gonzalez†, Sewon Min

* 同等贡献   † 同等指导
工作完成于 Berkeley SkyLab & BAIR & Berkeley NLP

按文档的外观搜索,而不仅仅看它包含的文字。

CI Live demo Status Slack License

它是什么 · 给 Claude 一双「眼睛」 · 工作原理 · 流水线

--- ```bash pip install pixelrag ``` 两大核心操作——将页面**渲染(render)**为截图,并在视觉索引中**搜索(search)**: ```bash # Render any page or document to screenshot tiles pixelshot https://en.wikipedia.org/wiki/Python --output ./tiles # Search a hosted index of 8.28M Wikipedia pages — no setup, runs against the live API curl -X POST https://api.pixelrag.ai/search \ -H "Content-Type: application/json" \ -d '{"queries": [{"text": "What is the capital of France?"}], "n_docs": 5}' ``` > **在线托管端点** — [`https://api.pixelrag.ai`](https://api.pixelrag.ai/status) 提供由 **8.28M** 篇 Wikipedia 页面构成的预构建索引。无需配置,无需 API 密钥。查询甚至可以是图片([视觉搜索(visual search)](https://pixelrag.ai/docs#search)) — 详见 **[API 参考 →](https://pixelrag.ai/docs)**. 或在浏览器中访问 **[pixelrag.ai](https://pixelrag.ai)**,,或在 Colab 中运行演示 notebook [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/StarTrail-org/PixelRAG/blob/main/demos/quickstart.ipynb) — 它会渲染页面并搜索托管索引,图像内联显示。 ## 它是什么 PixelRAG 将文档——网页、PDF、图像——渲染为截图,并直接在图像上进行检索。HTML 解析会丢弃的视觉结构——表格、图表、版式、信息图——得以保留,因此阅读模型能够真正回答相关问题。Wikipedia 的 8.28M 篇文章以预构建索引形式提供;流水线本身则是通用型的。 ## 给 Claude 一双「眼睛」 渲染器还以 Claude Code 插件形式提供——**pixelbrowse** 技能。Claude 不再抓取原始 HTML,而是用 `pixelshot` 对页面截图并*阅读图像*,从而像人一样看到图表、示意图、表格和版式。 安装即可——无需克隆仓库。安装 `pixelshot` CLI,使其位于你的 `PATH` 上(使用 `uv tool` 或 `pipx`,既可隔离环境,又随时可供 Claude 使用——仅向项目 venv 执行普通 `pip install` 可能会使 `pixelshot` 不在 `PATH` 中): ```bash uv tool install pixelrag # pixelshot on PATH (or: pipx install pixelrag) claude plugin marketplace add StarTrail-org/PixelRAG claude plugin install pixelbrowse@pixelrag-plugins ``` 然后直接让 Claude 查看某个页面即可: ```bash claude -p "screenshot https://news.ycombinator.com and summarize the top stories" claude -p "screenshot https://arxiv.org/abs/2404.12387 and explain the key findings" ``` 或在交互式会话中使用斜杠命令:`/screenshot https://example.com`。 无需 MCP 服务器,也无需后端:该技能只在你本机调用 `pixelshot`(Playwright/CDP)。 ## 工作原理

基于文本的 RAG 解析为文本并丢失表格;PixelRAG 渲染为截图分块并保留表格

基于文本的 RAG 将页面解析为文本块并**丢失表格**——阅读器找不到答案。PixelRAG 将页面渲染为**截图分块(screenshot tiles)**,检索到正确的分块后,阅读器可直接从图像中读取数字。 实现这一点需要两个组件:(1) 将文档渲染为图像,而非解析为文本;(2) 一个在截图数据上经 LoRA 微调的 `Qwen3-VL-Embedding` 模型,将页面图像嵌入到可视内容可检索的空间中。 ## 流水线 Capture 是独立的 `pixelshot` 命令;流水线的其余部分通过 `pixelrag` 总入口运行——`pixelrag `。只需安装你需要的阶段: | 命令 | 功能 | 安装 | | ------------------------------------------ | --------------------------------------------------------------- | ------------------------------- | | `pixelshot` | 文档 → 图像分块(Playwright CDP、PDF) | `pip install pixelrag` | | `pixelrag chunk` · `embed` · `build-index` | 分块 → 向量 → FAISS 索引 | `pip install 'pixelrag[embed]'` | | `pixelrag index` | 编排完整流水线:源 → 摄取 → 嵌入 → 索引 | `pip install 'pixelrag[index]'` | | `pixelrag serve` | FAISS 搜索 API(FastAPI,CPU 或 GPU) | `pip install 'pixelrag[serve]'` | ``` render ←── index ──→ embed serve (independent) train → serve (HTTP) ``` **`train` 是一个独立的 uv 项目**,拥有各自锁定的环境(`torch==2.9.1+cu129`、 `transformers==4.57.1`、cuDNN 9.20)——请在 `train/` 内部安装,而非在仓库根目录。 ### 搜索预构建索引 ```bash pip install 'pixelrag[serve]' # Download a pre-built index from Hugging Face. The dataset repo holds four FAISS indexes # (base/LoRA Wikipedia pixel, Wikipedia text, news pixel); grab just the base one (~217G) here. huggingface-cli download StarTrail-org/pixelrag-faiss-indexes \ --repo-type dataset --include "search_index_normed_v2/*" --local-dir ./index # Serve, then query pixelrag serve --index-dir ./index/search_index_normed_v2 --port 30001 curl -X POST http://localhost:30001/search \ -H "Content-Type: application/json" \ -d '{"queries": [{"text": "What is the capital of France?"}], "n_docs": 5}' ``` ### 用自己的文档构建索引 适用于 **Linux(CUDA)** 和 **macOS(Apple Silicon / MPS)**——`device: auto` 会自动选择最佳后端。 ```bash pip install 'pixelrag[index]' # Create pixelrag.yaml cat > pixelrag.yaml << 'EOF' source: type: local path: ./my_docs embed: model: Qwen/Qwen3-VL-Embedding-2B device: auto # cuda on Linux, mps on macOS, cpu as fallback output: ./my_index EOF # Build, then serve pixelrag index build pixelrag serve --index-dir ./my_index --port 30001 ```
试试看:索引 PDF 并在本地搜索 无需 GPU — 可在 macOS(Apple Silicon)或任何装有 Python 3.10+ 的机器上运行。 ```bash pip install 'pixelrag[index]' # 1. Grab a sample PDF (or use your own) curl -L -o paper.pdf https://raw.githubusercontent.com/StarTrail-org/PixelRAG/main/assets/pixelrag-paper.pdf # 2. Create config (device: auto picks MPS on Mac, CUDA on Linux) cat > pixelrag.yaml << 'EOF' source: type: local path: ./paper.pdf embed: model: Qwen/Qwen3-VL-Embedding-2B device: auto output: ./paper_index EOF # 3. Build the index (~3 min on Apple M-series, ~1 min on GPU) pixelrag index build # 4. Serve it pixelrag serve --index-dir ./paper_index --port 30001 # 5. Search — should return page 2 (the overview diagram) curl -X POST http://localhost:30001/search \ -H "Content-Type: application/json" \ -d '{"queries": [{"text": "Overview of PixelRAG and the diagram"}], "n_docs": 1}' ```
### 以编程方式渲染页面 ```python from pixelrag_render import render_url # render a single page to tiles — e.g. for an agent to read tiles = render_url("https://en.wikipedia.org/wiki/Python", "./tiles") ``` 同样的渲染能力也可通过 CLI 使用 — `pixelshot` 随 `pip install pixelrag` 一起提供: ```bash # Web page → tiles (headless Chromium via CDP) pixelshot https://en.wikipedia.org/wiki/Python -o ./tiles # PDF → tiles (requires poppler; install the pdf extra: pip install 'pixelrag[pdf]') curl -sL -o paper.pdf https://arxiv.org/pdf/2503.09516 pixelshot paper.pdf -o ./tiles --dpi 200 # URLs and local files can be mixed freely pixelshot https://github.com/StarTrail-org/PixelRAG paper.pdf -o ./tiles ``` > **Windows/macOS 上的 Chrome** — 捆绑的 turbo `headless_shell` 仅在 **linux-x64** 上自动安装。在其他平台上,`pixelshot` 会使用你系统中的 Chrome/Chromium(或 Playwright 的 Chromium),从标准安装位置自动检测。若无法自动找到,可通过 `CHROME_PATH=/path/to/chrome` 指定具体可执行文件路径。每次渲染都在隔离的临时 Chrome 配置文件中运行,因此即使你正在使用 Chrome 也能正常工作。 ### 嵌入工具(独立运行) 各阶段可独立运行,无需编排器(orchestrator): ```bash pip install 'pixelrag[embed]' pixelrag chunk --tiles-dir ./tiles pixelrag embed --shard-dir ./tiles --output-dir ./embeddings --gpu-ids 0,1 pixelrag build-index --embeddings-dir ./embeddings --output-dir ./index ``` ### 训练 微调代码位于 `train/` — 这是一个 **独立的 uv 项目**(`wiki-screenshot-training`),拥有自己固定版本的环境。它对 `Qwen/Qwen3-VL-Embedding-2B` 进行 LoRA 微调,用于网页检索;请在 `train/`(`cd train && uv sync`)内运行。完整流程请参阅 [`train/README.md`](train/README.md)。 使用模型无需自行重新训练 — 训练好的适配器已发布在 [`Chrisyichuan/wiki-screenshot-embedding-lora`](https://huggingface.co/Chrisyichuan/wiki-screenshot-embedding-lora/tree/main/lora_vit/ckpt200). 我们还发布了完整训练集([`Chrisyichuan/screenshot-training-natural-filtered-v2`](https://huggingface.co/datasets/Chrisyichuan/screenshot-training-natural-filtered-v2)),),方便你自行适配其他骨干模型 — 例如更大的 Qwen,或任意其他嵌入模型。数据整理流水线(LLM 增强的查询生成、过滤、困难负样本挖掘)详见 [`train/docs/synthetic_data_pipeline.md`](train/docs/synthetic_data_pipeline.md)。 ## 引用 如果你觉得 PixelRAG 有用,请引用我们的论文: ```bibtex @misc{wang2026pixelragwebscreenshotsbeat, title={PIXELRAG: Web Screenshots Beat Text for Retrieval-Augmented Generation}, author={Yichuan Wang and Zhifei Li and Zirui Wang and Paul Teiletche and Lesheng Jin and Matei Zaharia and Joseph E. Gonzalez and Sewon Min}, year={2026}, eprint={2606.28344}, archivePrefix={arXiv}, primaryClass={cs.IR}, url={https://arxiv.org/abs/2606.28344}, } ``` ## 致谢 感谢 [Rulin Shao](https://rulinshao.github.io/) 的支持。 同时感谢 [Claude Code](https://github.com/anthropics/claude-code) 和 [OpenAI Codex](https://github.com/openai/codex) 为开源贡献者提供积分与方案支持, 这些支持是我们在参与 [LEANN](https://github.com/StarTrail-org/LEANN). 工作期间获得的。 本工作由 [Berkeley Sky Computing Lab](https://sky.cs.berkeley.edu/), [BAIR](https://bair.berkeley.edu/), 与 [Berkeley NLP Group](https://nlp.cs.berkeley.edu/). 完成。 ## 许可证 Apache-2.0