ChatDocs enables users to interact with their documents offline using AI, ensuring data privacy as no information leaves the user's system. It supports a wide range of document formats and offers both a web UI and a command-line interface for querying.

How It Works

ChatDocs processes documents by extracting text and generating embeddings, which are then stored locally. When a user queries the system, ChatDocs retrieves relevant document chunks based on semantic similarity to the query and feeds these chunks, along with the query, to a local LLM for response generation. This approach leverages local processing for privacy and offline capability.

Quick Start & Requirements

• Install: pip install chatdocs
• Download models: chatdocs download
• Add documents: chatdocs add /path/to/documents
• Run UI: chatdocs ui (access at http://localhost:5000)
• Prerequisites: Python, internet for initial install/model download. GPU (CUDA) support requires specific PyTorch/ctransformers installations.

Highlighted Details

• Supports GGML/GGUF, 🤗 Transformers, and GPTQ models.
• Processes .csv, .docx, .html, .md, .pdf, .pptx, .txt, and more.
• Configurable via chatdocs.yml for embeddings and LLM choices.
• Offers GPU acceleration for embeddings and LLMs.

Maintenance & Community

No specific community links or contributor details are provided in the README.

Licensing & Compatibility

• License: MIT
• Compatibility: Permissive MIT license allows for commercial use and integration with closed-source applications.

Limitations & Caveats

The project is based on PrivateGPT and may inherit its limitations. GPU support requires manual PyTorch/CUDA setup, which can be complex. Changing embedding models necessitates re-processing all documents.