Legal Research Agent (Agentic RAG)
This workflow implements an agentic retrieval‑augmented‑generation (RAG) pipeline for legal research. The Agent node inspects the user’s question and selects the most appropriate action: search pre‑built Qdrant collections (e.g. CUAD mini, Iowa housing law, Wyoming housing law) or perform a Google web search. Search results are summarised and cited, and the agent returns a natural‑language answer while following strict instructions (e.g. no legal advice). The pipeline is designed for questions about contract clauses and housing regulations and demonstrates a multi‑tool, context‑aware RAG agent
Setup steps
Set up a Qdrant vector store with your legal document collections (e.g., CUAD mini, Iowa and Wyoming housing). Record the collection names.
Install and configure the Qdrant Hybrid Search and Qdrant Vector Search plugins with your Qdrant API key and server URL.
Install a browser/web search plugin (e.g., Tavily Search) and supply the corresponding API key for fallback searches.
Provide an LLM provider (OpenAI or another) for the agent’s reasoning, summarisation and final answer.
Review the agent instructions to ensure they reference your collections and abide by any legal compliance rules.