william-ai-dev/xai-document-research

XAI Document Research Assistant

A sophisticated RAG (Retrieval-Augmented Generation) application for document analysis with advanced evaluation metrics and LangSmith tracing.

Features

• —📄 Multi-format Support: PDF, TXT, MD, DOCX document processing
• —🔍 Advanced Retrieval: Hybrid search with semantic and BM25 ranking
• —🧠 AI-Powered Analysis: Groq Llama 3.1 integration for intelligent responses
• —📊 Quality Metrics: Faithfulness and answer relevancy evaluation using Ragas
• —🔗 LangSmith Tracing: Full observability and debugging capabilities
• —💬 Interactive Chat: Streamlit-based conversational interface
• —⚡ Real-time Processing: Efficient embedding and retrieval pipeline

Technical Stack

• —Frontend: Streamlit
• —LLM: Groq Llama 3.1
• —Vector Database: ChromaDB
• —Embeddings: Sentence Transformers
• —Evaluation: Ragas framework
• —Tracing: LangSmith
• —Document Processing: PyMuPDF4LLM, MarkItDown

Usage

1. 1.Upload a document (PDF, TXT, MD, or DOCX)
2. 2.Click "Process Document" to analyze and embed the content
3. 3.Ask questions about the document
4. 4.View detailed analysis including reasoning, sources, and quality metrics

Environment Variables

For Hugging Face Spaces, set these in the Secrets section:

• —LANGCHAIN_API_KEY: Required for LangSmith tracing
• —GROQ_API_KEY: Required for Groq LLM access
• —LANGCHAIN_PROJECT: Project name for LangSmith (optional, defaults to "default")

Installation

pip install -r requirements.txt
streamlit run app.py

License

MIT License