Recherche Vectorielle
Pure vector search in the embedding space. Returns the closest documents by cosine similarity.
Parameters
| Parameter | Type | Required | Variable | Description |
|---|---|---|---|---|
query | dynamic value | Yes | Yes | Text to search by semantic similarity in vector space. |
limit | number | No | No | Maximum number of results to return. (Default: 10, min 1, max 100) |
minScore | number | No | No | Minimum similarity score (0 to 1). Results below are excluded. (Default: 0.7, min 0, max 1) |
entityTypeFilter | array | No | No | Filter results by entity types. |
embeddingModel | choice (text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002) | No | No | Embedding model to use for vectorizing the query. |
rerank | boolean | No | No | Enable result reranking for improved relevance. (Default: false) |
dedupThreshold | number | No | No | Deduplication threshold (0 to 1). Results too similar are merged. (min 0, max 1) |
outputVariable | text | No | No | Output variable name containing the vector search results. |
Parameters marked Variable = Yes accept the
{{blockName.field}}syntax.
Output
Output variable : vectorResults
// direct value (no wrapper object)
Example
Vector search on a concept.
Input :
{"query": "gestion des risques operationnels"}
Output :
[{"id": "doc-1", "content": "...", "score": 0.91}]
Tip
Faster than hybrid search but lower recall on exact keywords. Use topK to limit the number of results.