(gemini): Pinecone: Query Vector Database Action

[swerner]

Queries a Pinecone vector database with an embedding vector. Returns the top matching documents for semantic search and retrieval-augmented generation (RAG).

[augmentations-ai]

This is a great addition! To ensure thorough testing, here's a suggested manual verification process:

Manual Verification Steps:

1. Setup Pinecone Index:

   • Create a Pinecone account and a Pinecone index according to your desired specifications (dimension, metric, etc.).
   • Populate the index with some test vectors and associated metadata. Ensure you have a good variety of data to test the query.
   • Set the PINECONE_API_KEY and PINECONE_ENVIRONMENT environment variables with your credentials.

2. Instantiate and Run the Action:

   • In a testing environment (e.g., a Rails console, a dedicated test script), instantiate the PineconeQueryVectorDatabaseAction with appropriate parameters:

   action = PineconeQueryVectorDatabaseAction.new(
     pinecone_index_name: 'your-index-name',
     embedding_vector: [0.1, 0.2, 0.3, ..., 0.9] # Replace with a valid embedding vector that exists in your index
     top_k: 5, # Or any number you choose to test
     namespace: 'your-namespace' # Only if your data is namespaced
   )

3. Inspect the Results:

   • Call the call method and inspect the returned results:

   results = action.call
   puts results.inspect

   • Verify that the results array contains the expected number of matches (top_k).
   • Examine the id, score, and metadata for each match. Confirm that the returned documents are indeed the top matches based on your embedding_vector and that the scores are reasonable.
   • Try different embedding_vector values to ensure that the correct documents are retrieved.

4. Test Error Handling:

   • Provide an invalid pinecone_index_name and confirm that the action raises a StandardError with a descriptive error message.
   • Provide an invalid embedding_vector (e.g., wrong dimensionality) and verify that the action handles the error gracefully and logs an appropriate message.
   • Temporarily invalidate the PINECONE_API_KEY and confirm that the action raises an error related to authentication.

5. Test with and without namespace

   • Ensure your test covers cases when the namespace parameter is nil and when it is a value that is used in your pinecone database.

By following these steps, you can gain confidence that the PineconeQueryVectorDatabaseAction is functioning correctly and that it integrates seamlessly with the Pinecone vector database.