Qdrant AI Tools icon

Qdrant AI Tools

Provides all Qdrant vector store operations (search, scroll, upsert, delete, collection management, snapshots) to an AI Agent. Optionally connect an Embedding model for semantic search and a Reranker to improve retrieval quality.

Qdrant AI Tools

AI-generated

Summary

Expose multiple Qdrant vector store operations as configurable AI tools accessible by an AI Agent node. Enable specific tools to perform tasks such as semantic search (embedding queries and finding similar vectors with reranking), scrolling through points with filters, retrieving, upserting, deleting points, counting points, recommending points based on examples, managing collections (list, get info, create, delete, update), and handling collection snapshots (create, list).

Inputs

  • Default Collection — Name of the default Qdrant collection to use when no collection is specified.
  • Tools to Enable (required) — Select which Qdrant operations to expose as tools to the AI Agent (e.g., semantic search, scroll points, upsert points, collection management).
  • Default Search Limit — Default maximum number of results returned by search and recommendation tools if no limit is specified.
  • Include Payload in Results — Flag to include stored payload (metadata and text) in results by default.
  • Include Vectors in Results — Flag to include raw vector data in search and retrieval results by default.
  • Embedding Model Sub-node (optional) — Connect an embedding model sub-node to enable semantic search tools.
  • Reranker Sub-node (optional) — Connect a reranker sub-node to improve quality of search results by reranking.

Output shape

a set of tools implementing enabled Qdrant operations exposed as DynamicStructuredTool instances, each returning JSON-encoded result objects

Each tool accepts arguments according to its schema (e.g., collection name, query, IDs, filters) and returns structured results such as search hits, points data, operation success info, or error details. If no Embedding sub-node is connected, semantic search and recommendation tools will return errors. Results optionally include payload metadata and raw vectors according to settings. The node is intended to be used with an AI Agent node connected to its 'Tool' output; executing it directly returns an informational message.

Links

Discussion