Contract Compliance Agent
An AI contract review system for enterprise legal and compliance teams that combines retrieval-augmented generation, a multi-agent router and Taiwanese statute lookup to flag risky contract clauses with cited legal sources and suggested edits.
This system is a contract review tool aimed at corporate legal, procurement and compliance teams, and at teams building their own AI projects on top of it, that combines contract risk identification, statute lookup, knowledge-base question answering and evaluation into one entry point, so a first-pass contract review that normally takes an hour or more can be done in minutes, with cited sources and traceable evidence. It ships as a Streamlit demo and as a FastAPI-plus-Vue deployment for internal use.
An uploaded contract is vectorized, and a LangGraph-driven RAG pipeline with an agent router selects among tools for RAG lookup, contract review, statute search and expert agents, combining hybrid vector-and-BM25 retrieval with live queries to Taiwan's Judicial Yuan statute database and Tavily web search. Results come back as structured risk cards with a risk level, a legal citation and a suggested edit, alongside a document preview; a built-in evaluation set and batch run measure routing accuracy and tool success rate for regression tracking.
Built with Python, LangGraph, Pinecone, Gemini and Streamlit or FastAPI and Vue, with a GitHub Actions pipeline of tests and end-to-end checks that deploys to a DGX host over Tailscale SSH on success; also supports local Ollama models. MIT licensed.