GraphRAG4OpenWebUI

Why we included this project

This project slots into Open WebUI as an API layer that accepts the OpenAI chat-completions format, so you can add graph-based retrieval without standing up a separate service. It wraps Microsoft's GraphRAG and gives you three ways to query: local search against your own knowledge base, a broader global search, and a Tavily-backed web search, plus a combined mode that merges and ranks results from all three. What makes it practical for smaller setups is that the whole thing can run on local models: point it at Ollama or LM Studio for the LLM and use a local embedding model, which keeps documents private and avoids per-query API costs. If you already run Open WebUI, the integration is just registering the endpoint in the UI config. It's a focused bridge rather than a general-purpose framework, best for teams that want GraphRAG-style answers inside their existing chat interface with minimal wiring.