Authors: @Anu Judson and @Neena P K 

Using IBM® App Connect to interact with Pinecone Vector Database

Pinecone Vector Database is a purpose-built platform for indexing and searching dense vectors. It supports use cases like semantic search, recommendation systems, and AI-powered data analysis.

Building a flow in IBM App Connect with Pinecone Vector Database

Use App Connect to build flows that integrate with Pinecone Vector Database and other applications. The connector is displayed as Pinecone Vector Database on the App Connect User Interface (UI). 

To allow App Connect to connect to your Pinecone Vector Database account, you need to fill in the connection fields that you see in the App Connect Designer Connect > Applications and APIs page or flow editor.

Supported authorization type: 

API key

For detailed information about the authorization type, how to generate connection field values, and how to use the Templates gallery, see How to use IBM App Connect with Pinecone Vector Database on the IBM Documentation page.

Supported objects in Pinecone Vector Database

The following are the Pinecone Vector Database objects that can be run in App Connect.

Scenario: RAG pattern flow - Query the Pinecone Vector DB using embeddings from IBM watsonx ai and generate text with the results

Consider the following scenario:

Using a sentence transformer model, textual data is converted into vector embeddings. The generated embeddings, along with their associated text, are stored in the Pinecone Vector Database for semantic search.

In this flow, when you submit a question, the relevant data is retrieved from Pinecone Vector Database based on semantic similarity, and IBM watsonx.ai generates a response using that data as context.

The following are the steps involved in the flow:

1. In the request, submit a question.
Note: The question should be related to the data stored in Pinecone Vector Database.

2. When a question is submitted, IBM watsonx.ai converts the question into an embedding using the IBM watsonx.ai Generate embeddings node.
Note: Use the same sentence transformer model that you employed for converting textual data into vector embeddings to also generate the corresponding embedding of the given question.

3.

3. Using the Pinecone Vector Database Query vectors node, contextually similar data is retrieved from the vector database based on the question embedding.

4. The retrieved data is then passed as context, along with the original question. Using the IBM watsonx.ai Generate text node, the flow generates a response using the retrieved context. The response is the answer to the question based on the data previously stored in Pinecone Vector Database.   

Resources

You can view this template and other useful templates in the Templates gallery of your App Connect Designer instance.

Get started with a free IBM App Connect Enterprise as a Service trial for 30 days to try out all our templates, visit https://ibm.biz/app-trial.

If you are running a containerized instance of IBM App Connect, use the following URL to directly access the template:

• RAG pattern flow - Query the Pinecone Vector DB using embeddings from IBM watsonx ai and generate text with the results: https://<your-instance-id>/templates/RAG%20pattern%20flow%20- %20Query%20the%20Pinecone%20Vector%20DB%20using%20embeddings%20from%20IBM%20watsonx%20ai%20and%20generate%20text%20with%20the%20results
  Note: Replace <your-instance-id> in the URL with your custom instance ID.

You must enable the Designer AI features in your containerized environment to access the App Connect templates. For more information, see The preloaded IBM App Connect templates.

You also have the option to import the .yaml file for this template directly into your App Connect Designer instance. These templates are available in a public GitHub repository.