Sign in
latestOpenAPI 3.0.32026-08-227451,2751.8 MB

d0fc0114da66

KnowledgeGraph

Search knowledge graph nodes

Hybrid search across knowledge graph nodes using semantic (vector), fulltext, and substring modes. Supports Reciprocal Rank Fusion (RRF) when multiple input types are provided. At least one of query_text, literal_phrases, or name_substring must be supplied. When mode=auto (default), the server selects the active modes based on which inputs are present; modes_used reflects the actual modes that ran.

post/knowledgeGraph/nodes/search

Request body

workspace_idstring required

The workspace ID

application_idstring required

The application ID

query_textstring

Semantic query text for vector similarity search

literal_phrasesstring[]

Exact phrase(s) for FULLTEXT search. Phrases are OR-joined — a node matches when any one phrase matches, and ranking rewards nodes that match several.

name_substringstring

Substring match on node name

label_filterKnowledgeGraphNodeLabel[]

Optional filter to restrict results to specific node labels

mode'auto' | 'semantic' | 'fulltext' | 'substring' | 'rrf'

Search mode. auto (default) selects active modes based on provided inputs. Explicit values force a specific mode. modes_used in the response reflects the actual mode(s) that ran.

limitinteger

Maximum number of results to return. Capped at 50 to keep the agent-facing response within a manageable LLM context budget.

Response

Search results with RRF-fused ranking and per-mode ranks

modes_usedstring[] required

The actual search modes that contributed to the results