Skip to main content

Vector Store Deleter

Remove specific documents from vector stores by reference

What is Vector Store Deleter?​

The Vector Store Deleter node enables precise removal of documents from vector databases by reference identifier. This node is essential for maintaining clean vector stores, removing outdated information, and managing data lifecycle in your knowledge base systems. It supports multiple vector store providers including Postgres, Pinecone, and OpenSearch.

How to use it?​

To effectively remove documents from your vector store, follow these steps:

  1. Connect Vector Store:

    • Connect your configured vector store (Postgres, Pinecone, or OpenSearch) to the Vector Store input
    • Ensure the vector store connection is properly authenticated and accessible

  2. Configure Storage Location:

    • For Pinecone and OpenSearch: Specify the Index Name (default: "documents")
    • For Postgres: Specify the Table Name (default: "embeddings")
    • Use consistent naming with your Vector Store Writer configuration

  3. Provide Reference Identifier:

    • Connect the Reference input with the unique identifier of the document to delete
    • This reference must match exactly with the reference used when the document was stored
    • References are typically URLs, file paths, or custom identifiers

  4. Execute Deletion:

    • The node acts as a terminal operation - it performs deletion but produces no output
    • Deletion is permanent and cannot be undone
    • Verify reference accuracy before execution

Example of usage​

Objective: Remove outdated product information from a vector store when products are discontinued.

Setup Steps:

  1. Vector Store Connection: Connect your Postgres vector database containing product embeddings
  2. Table Configuration: Set table name to "product_embeddings" to match your storage schema
  3. Reference Input: Use product SKU or ID as reference (e.g., "PROD-2024-001")
  4. Automation: Trigger deletion when product status changes to "discontinued"

Use Case Flow:

  • Product status update triggers workflow
  • Product ID flows to Vector Store Deleter reference input
  • Document and embeddings are removed from vector store
  • Knowledge base stays current with active products only

Additional information​

Best Practices:​

  • Reference Consistency: Use the same reference format across Writer and Deleter operations
  • Batch Operations: For multiple deletions, process references individually through the node
  • Verification: Log or track deleted references for auditing purposes
  • Error Handling: Handle cases where references don't exist gracefully

Common Use Cases:​

  • Remove outdated documentation from knowledge bases
  • Clean up superseded product information
  • Delete deprecated API documentation
  • Remove personal data for compliance (GDPR/CCPA)
  • Maintain fresh and relevant search results

Important Considerations:​

  • Deletion is permanent and immediate
  • No confirmation or undo functionality
  • Failed deletions (non-existent references) may not produce errors
  • Consider implementing backup strategies for critical data