Smart Qdrant Vector Store icon

Smart Qdrant Vector Store

Work with Qdrant Vector Store
Join our community

Overview

This node integrates with the Qdrant vector store to manage and query vectorized documents. It supports multiple modes: inserting documents into a Qdrant collection, retrieving ranked documents based on a query, and providing vector store or tool interfaces for AI chains or agents. It is useful for workflows involving semantic search, document similarity, and AI-driven document retrieval, such as enhancing chatbots with relevant document context or building AI tools that leverage vector search.

Use Case Examples

  1. Insert a batch of documents with embeddings into a Qdrant collection for later semantic search.
  2. Query a Qdrant collection with an embedding to retrieve the top N most similar documents.
  3. Use the node as a vector store or tool in an AI chain to dynamically fetch relevant documents based on user input.

Properties

Name Meaning
Mode Determines the operation mode of the node: insert documents, load ranked documents, retrieve documents as a vector store, or use as a tool for AI agents.
Collection Source Select how to specify the Qdrant collection: from a list of existing collections or manually by ID.
Qdrant Collection The selected Qdrant collection to operate on, chosen from a list of existing collections.
Collection Name The name of the Qdrant collection to use, entered manually if manual collection source is selected.
Batch Size Number of documents to process in each batch when inserting documents.
Vector Dimensions Dimension size of the vectors, must match the embedding model used.
Query The search query string used to retrieve ranked documents from the vector store.
Limit Number of results to return when loading ranked documents.
Options Additional optional settings such as Qdrant URL override, batch delay, memory clearing, filters, top K results, and search score threshold.

Output

JSON

  • pageContent - The content of the retrieved document.
  • metadata - Metadata associated with the retrieved document.
  • success - Indicates successful processing of documents during insertion.
  • processedCount - Number of documents processed in the current operation.
  • collectionName - Name of the Qdrant collection used.
  • sessionId - Optional session identifier if memory clearing is enabled during insertion.

Dependencies

  • Qdrant API key credential for authentication
  • @langchain/qdrant and @langchain/core/documents libraries
  • @qdrant/js-client-rest for Qdrant client operations

Troubleshooting

  • Ensure Qdrant URL is correctly configured either in credentials or node options; missing URL causes errors.
  • Verify the Qdrant collection exists before inserting or querying; errors occur if collection is missing.
  • Input data must include embeddings and documents with 'pageContent'; missing fields cause failures.
  • Invalid JSON in filter option will cause parsing errors; ensure correct JSON format.
  • Network errors connecting to Qdrant (e.g., ECONNREFUSED, ENOTFOUND) indicate connectivity issues; check URL, network, and firewall settings.
  • Batch processing errors during insertion may occur if documents are malformed or connection issues arise; check logs for batch-specific errors.

Links

Discussion