Skip to main content

Zep Collection - Cloud Vector Store

Overview​

The Zep Collection - Cloud Vector Store is a powerful component in AnswerAI that allows you to store, retrieve, and search vector embeddings efficiently. It uses Zep, a fast and scalable building block for LLM applications, to manage your vector data in the cloud.

Key Benefits​

  • Efficient similarity search: Quickly find the most relevant documents based on vector similarity.
  • Cloud-based storage: Store and access your vector data securely in the cloud.
  • Flexible metadata filtering: Easily filter your search results using custom metadata.

How to Use​

  1. Add the "Zep Collection - Cloud" node to your AnswerAI canvas.

  2. Configure the node with the following settings:

    a. Connect Credential: Select or create a Zep Memory API credential.

    b. Document: (Optional) Connect a Document node to add documents to the vector store.

    c. Zep Collection: Enter a name for your Zep collection (e.g., "my-first-collection").

    d. Zep Metadata Filter: (Optional) Add a JSON object to filter search results based on metadata.

    e. Top K: (Optional) Specify the number of top results to fetch (default is 4).

  3. Connect the Zep Collection node to other nodes in your workflow that require vector storage or retrieval.

Zep Collectionconfiguration panel & Drop UI

Tips and Best Practices​

  1. Use meaningful collection names: Choose descriptive names for your Zep collections to easily identify their purpose.

  2. Optimize metadata: Design your metadata structure carefully to enable efficient filtering and improve search relevance.

  3. Manage API keys securely: Always use the credential manager to store your Zep API keys securely.

  4. Monitor usage: Keep track of your vector store usage to optimize performance and manage costs effectively.

Troubleshooting​

  1. Connection issues:

    • Ensure that your Zep API credential is correctly configured.
    • Check your internet connection and firewall settings.

  2. Slow search performance:

    • Consider optimizing your metadata filters.
    • Reduce the number of vectors in your collection if it becomes too large.

  3. Unexpected search results:

    • Verify that your metadata filters are correctly formatted as JSON.
    • Double-check the "Top K" value to ensure you're retrieving the desired number of results.

If you encounter persistent issues, consult the AnswerAI documentation or reach out to support for assistance.