This project provides a free, user-friendly interface for chatting with documents (PDF, DOCX, CSV, TXT) using LangChain and a choice of LLM providers, including a free tier via gpt4free. It's designed for users who want to quickly extract information from their documents without API keys or costs, and for developers looking to build similar document-based AI applications.

How It Works

The application leverages LangChain's agent capabilities to process user queries against uploaded documents. It supports direct document URL input and integrates with LLM providers, offering a free option through gpt4free or requiring an OpenAI API key. For queries extending beyond the document content, it can optionally use a serpapi_key to incorporate Google search results, mimicking a comprehensive AI assistant.

Quick Start & Requirements

• Install/Run: Local development via pip install -r requirements.txt and streamlit run ./app.py, or using Docker Compose (docker-compose up).
• Prerequisites: Python 3.10.11, Langchain 0.0.218, Streamlit 1.22.0. Optional: openai_api_key, serpapi_key.
• Deployment: Streamlit Community Cloud is supported.
• Links: Demo, GitHub

Highlighted Details

• Free usage via gpt4free integration, eliminating the need for API keys.
• Supports direct document URL input, avoiding local uploads.
• LangChain agent enables automated tool selection (LLM, Google Search) for enhanced query answering.
• Offers guidance on optimizing LLM selection, PDF loading (PyMuPDF, PDFPlumber), and token usage tracking.

Maintenance & Community

The project is actively maintained by Lin-jun-xiang. Community support and issue tracking are available via the GitHub repository.

Licensing & Compatibility

The repository does not explicitly state a license in the README. This requires clarification for commercial use or integration into closed-source projects.

Limitations & Caveats

The free Streamlit Cloud deployment may experience crashes due to resource limitations under heavy user load. For a more stable experience, local deployment is recommended. The absence of a specified license poses a significant adoption blocker for many use cases.