LegalAssistant

An AI-powered legal document analysis tool built with Streamlit and LangChain. Upload PDF, Word, or plain text legal documents and ask questions — the assistant answers using advanced RAG techniques with human oversight indicators.

Features

• Document Upload: Supports PDF, .docx, and .txt legal files
• Advanced RAG Pipeline: FAISS vector store + HuggingFace sentence embeddings for semantic retrieval
• Query Rewriting: Reformulates questions for better retrieval accuracy
• Speculative Retrieval: Generates a hypothetical answer to guide document search
• Confidence Scoring: Rates responses as High / Medium / Low confidence
• Human Review Flags: Automatically flags low-confidence answers for human review
• Groq LLM Backend: Fast inference via langchain-groq
• Web Research Agent (ResearchAgent.py): DuckDuckGo + Tavily search for supplementary legal context

Stack

• Frontend: Streamlit
• LLM: Groq (ChatGroq)
• RAG: LangChain + FAISS + HuggingFace Embeddings
• Document Parsing: PyMuPDF, Unstructured, python-docx
• Search: DuckDuckGo Search, Tavily

Setup

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

Requires a GROQ_API_KEY environment variable.