Introduction

GraphRAG is a multi-tenant knowledge API. A partner uploads documents; the service chunks and embeds them, extracts an entity/relationship graph, and answers questions with citations back to the source passages. Every tenant gets its own isolated knowledge base.

It is designed for teams that run document-intelligence for several customers at once and need three things at the same time: isolation between tenants, traceability in every answer, and the ability to self-host on their own infrastructure.

What you get

• A small REST API: create a project, upload documents, ask questions.
• Hybrid retrieval — semantic vector search fused with keyword search and re-ranking.
• Citations on every answer, resolving to the exact chunks used.
• A knowledge graph (entities, aliases, relationships) built from your documents, with per-relation provenance.
• Per-tenant isolation enforced at both the application and database layers.

How answers are generated today

Answers are produced by hybrid retrieval over your documents, with citations. Graph-based answering is built and available but off by default — it is enabled only where it is measured to improve answer quality. See Limitations for the honest current state.

Where to go next

• New here? Start with the Quickstart.
• Want the mental model? Read Architecture.
• Ready to integrate? Jump to Authentication and the API reference.