Recherche Hybride

Hybrid search combining vector and lexical search with RRF fusion. Offers a better balance between precision and recall.

Parameters

Parameter	Type	Required	Variable	Description
query	dynamic value	Yes	Yes	Text or question for combined vector + graph search.
vectorWeight	number	No	No	Weight of vector search in results (0 to 1). (Default: 0.7, min 0, max 1)
graphWeight	number	No	No	Weight of graph search in results (0 to 1). (Default: 0.3, min 0, max 1)
limit	number	No	No	Maximum number of results to return. (Default: 20, min 1, max 100)
includeNeighbors	boolean	No	No	Include neighbor entities in results for richer context. (Default: false)
entityTypeFilter	array	No	No	Filter results by entity types (e.g. Client, Project).
rerank	boolean	No	No	Re-rank results with a reranking model for better relevance. (Default: false)
mmr	boolean	No	No	Enable MMR diversification to avoid redundant results. (Default: false)
fusionAlgorithm	choice (average, max, rrf)	No	No	Score fusion algorithm (average, max, or Reciprocal Rank Fusion).
outputVariable	text	No	No	Output variable name containing the search results.

Parameters marked Variable = Yes accept the {{blockName.field}} syntax.

Output

Output variable : hybridResults

// direct value (no wrapper object)

Example

Hybrid search across documents.

Input :

{"query": "reglementation Bale III"}

Output :

[{"id": "doc-1", "content": "Les exigences de fonds propres Bale III...", "score": 0.89}, {"id": "doc-2", "content": "...", "score": 0.74}]

Tip

{{hybridResults}} is directly an array (no wrapper object). Combines semantic precision with exact keyword matching.