Type: vector.HybridSearch

Namespace: vector

Description

Fuse a semantic ranking with a keyword-filtered semantic ranking via reciprocal rank fusion. The keyword leg runs the same semantic query constrained to documents containing the query tokens, so documents matching on both legs rank higher. vector, RAG, query, semantic, text, similarity

Properties

Property Type Description Default
collection collection The collection to query {"type":"collection","name":""}
text str The text to query ``
n_results int The number of final results to return 5
k_constant float Constant for reciprocal rank fusion (default: 60.0) 60
min_keyword_length int Minimum length for keyword tokens 3

Outputs

Output Type Description
ids list[str]  
documents list[str]  
metadatas list[dict]  
distances list[float]  
scores list[float]  

Browse other nodes in the vector namespace.