The Pinecone Vector Store node connects your pipeline to a Pinecone index, enabling high-speed vector search capabilities. It stores and retrieves document embeddings and matches them against queries using vector similarity.
Inputs
- Documents – Receives vectorized documents to be stored in the Pinecone collection.
- Questions – Accepts vectorized queries for searching similar embeddings in Pinecone.
Outputs
- Answers – Returns the best-matching vectors or metadata based on the retrieval score.
- Documents – Emits document information retrieved from Pinecone for further use.
- Questions – Forwards the incoming query vector downstream (e.g., into LLMs or logs).
Configuration Steps
- Type of Pinecone Connection – Choose your Pinecone connection method
- Example – Pinecone Pod-Based Index
- API Key – Enter your Pinecone API key
- Retrieval Score – Select the minimum similarity threshold for result relevance
- Example – Related
- Collection – Specify the name of the Pinecone collection (index)
- Rules – lowercase, alphanumeric, hyphens allowed
- Example – aparavi
