Pinecone
CertifiedInteract with Pinecone, a managed vector database, from your Kestra workflows. Create serverless indexes, upsert vectors, query by similarity, fetch or delete vectors by ID, and inspect index statistics.
Pinecone
Interact with Pinecone, a managed vector database, from your Kestra workflows. Create serverless indexes, upsert vectors, query by similarity, fetch or delete vectors by ID, and inspect index statistics.
tasks
How to use the Pinecone plugin
Manage indexes, upsert and query vectors, and interact with Pinecone's managed vector database from Kestra flows.
Authentication
All tasks require apiKey (your Pinecone API key, required) and indexName (the index to operate on, required). Optionally set host to override the control-plane endpoint (default https://api.pinecone.io) and tlsEnabled (default true) to control gRPC encryption — set tlsEnabled: false when using the Pinecone local emulator. Store secrets in secrets and apply connection properties globally with plugin defaults.
Tasks
CreateIndex creates a new serverless index — set dimension, cloud, and region (all required). Optionally set metric (default COSINE; also EUCLIDEAN or DOTPRODUCT) and deletionProtection (default disabled). Outputs name, host, and status.
DeleteIndex deletes the index identified by indexName.
Upsert writes vectors into the index. Supply either vectors (a list of objects each with a required id and values array, and optional metadata and sparseValues) or from (a Kestra ION file URI) — the two are mutually exclusive. Optionally set namespace to scope writes to a namespace. Outputs upsertedCount.
Query searches the index by similarity — set topK (required). Provide either vector (a list of floats) or vectorId (an existing vector ID) as the query; the two are mutually exclusive. Optionally set filter (a metadata filter map), includeValues (default false), includeMetadata (default false), and fetchType (FETCH, FETCH_ONE, or STORE, default FETCH). Outputs size and rows (FETCH), row (FETCH_ONE), or uri (STORE).
FetchVectors retrieves vectors by ID — set ids (required list of vector IDs). Optionally set namespace. Outputs vectors as a map keyed by ID, each containing values and optional metadata.
DeleteVectors removes vectors from the index. Supply either ids (a list of vector IDs) or deleteAll: true to clear the entire namespace — the two are mutually exclusive. Optionally set namespace.
DescribeIndexStats returns index statistics. Outputs totalVectorCount and namespaceCounts (a map of namespace to vector count).
