Stream into Pinecone with your free account

The Pinecone materialization connector transforms documents from Estuary collections into vector embeddings using the OpenAI Embedding API and stores them in a Pinecone index for real-time semantic search and retrieval.

• AI-powered embedding generation: Automatically converts Flow collection data into dense vector representations using OpenAI’s text-embedding-ada-002 model (or a custom embedding model if specified).
• Real-time vector storage: Inserts or updates vector embeddings in Pinecone namespaces, keeping your search index continuously in sync with source data.
• Flexible field inclusion: Embeddings are generated from scalar fields by default, with the option to include arrays and objects through projections.
• Metadata preservation: Stores the full Flow document as JSON metadata (flow_document) in Pinecone for easy retrieval alongside embeddings.
• Upsert-based delta updates: Uses Flow’s delta update mechanism to replace or insert vectors efficiently, ensuring idempotent synchronization.
• Seamless multi-cloud support: Works with any Pinecone environment (e.g., us-central1-gcp) and supports optional OpenAI organization scoping for enterprise setups.

💡 Tip: To optimize Pinecone memory usage, disable metadata indexing for the flow_document field—this field is only used for retrieval, not filtering.