Skip to main content

Upstash Vector Store

Overview​

The Upstash Vector Store node allows you to store and retrieve vector embeddings using Upstash, a serverless data platform. This feature enables efficient similarity searches and management of high-dimensional numerical vectors.

Key Benefits​

  • Serverless vector database for easy scalability
  • Fast similarity searches for improved query performance
  • Seamless integration with AnswerAI workflows

How to Use​

Prerequisites​

  1. Sign up or sign in to the Upstash Console

  2. Navigate to the Vector page and click Create Index

    Upstash Console Vector page & Drop UI

  3. Configure and create the index:

    • Index Name: Choose a name for your index (e.g., "answerai-upstash-demo")
    • Dimensions: Set the size of the vectors to be inserted (e.g., 1536)
    • Embedding Model: (Optional) Select a model from Upstash Embeddings

      Upstash Console Index form & Drop UI

Setup in AnswerAI​

  1. Obtain your index credentials from the Upstash Console

    Upstash Console Index form & Drop UI

  2. In AnswerAI, create a new Upstash Vector credential:

    • Enter the Upstash Vector REST URL (UPSTASH_VECTOR_REST_URL)
    • Enter the Upstash Vector REST Token (UPSTASH_VECTOR_REST_TOKEN)

Upstash Console Index form & Drop UI

  1. Add a new Upstash Vector node to your canvas

    Upstash Vector Store node & Drop UI

  2. Connect additional nodes to the Upstash Vector node:

    • Connect a Document Loader node to provide input documents
    • Connect an Embeddings node to generate vector embeddings

      Upstash Vector Store node in a workflow & Drop UI

  3. Configure the Upstash Vector node:

    • Select the Upstash Vector credential you created earlier
    • (Optional) Set a metadata filter to refine your searches
    • (Optional) Adjust the Top K value to specify the number of results to retrieve

  4. Run your workflow to start the upsert process and store your vectors

  5. Verify data storage in the Upstash dashboard

    Upstash Dashboard & Drop UI

Tips and Best Practices​

  1. Choose an appropriate dimension size for your vectors based on your embedding model and use case
  2. Use metadata filters to organize and retrieve specific subsets of your data
  3. Regularly monitor your Upstash usage to optimize performance and costs
  4. Consider using the optional Embedding Model feature in Upstash for simplified vector generation

Troubleshooting​

  1. Connection issues: Ensure your Upstash Vector REST URL and Token are correctly entered in the AnswerAI credential
  2. Indexing errors: Verify that your document format and embedding dimensions match the Upstash index configuration
  3. Slow performance: Check your network connection and consider optimizing your vector dimensions or index size

If problems persist, consult the Upstash documentation or contact AnswerAI support for assistance.