Retrieval

GET /api/search searches document vectors, lexical indexes, graph data, or a combination. Chat defaults to hybrid_rr, which combines dense and BM25 retrieval before reranking candidates.

Expand diagram

Hybrid retrieval and ranking

Dense and BM25 rankings feed weighted fusion and optional reranking. Graph retrieval is a separate path; its matches count as evidence only after resolving to source text.

Use the zoom controls, drag to pan, or pinch on a touch screen. Press Escape to close.

Retrieval at a glanceCopy Anchor Link

• Use search arms to understand the six available retrieval methods.
• See hybrid ranking for candidate fusion, normalization, and reranking behavior.
• Follow graph retrieval to see how entity matches become source-grounded results.
• Read about evidence and filters for scoped searches, returned fields, and citation handling.

Retrieval returns ranked chunks. A score is meaningful within the method and query that produced it; it is not a universal confidence measure.