Skip to main content

Vector Store Service

VectorStoreService provides embeddings collection management and similarity search.

Core API​

  • ensure_collection(name, dimension, distance_metric, recreate=False)
  • add_documents(collection, ids, vectors, metadata=None, payloads=None)
  • search_similar(collection, query_vector, k=10, filter=None, score_threshold=None)
  • get_document(collection, document_id)
  • update_document(...)
  • delete_documents(collection, ids)
  • get_collection_size(collection)
  • list_collections()

Adapter options​

  • qdrant: external Qdrant server.
  • qdrant_in_memory: in-memory implementation for local/dev tests.
  • custom: pluggable adapter.

Config example​

services:
  vector_store:
    vector_store_adapter:
      adapter: "qdrant"
      config:
        host: "localhost"
        port: 6333
        https: false
        timeout: 30

Notes​

  • Service auto-initializes adapter on first call.
  • Input validation checks length consistency for ids/vectors/metadata/payloads.
  • score_threshold filtering is applied client-side on results.