Pinecone
by Pinecone Systems, Inc.
Upsert and query vector embeddings in Pinecone for retrieval-augmented generation (RAG), semantic search, and similarity matching in agent workflows.
How agents use Pinecone
- ✓Agent queries a Pinecone index with an embedding to retrieve the most relevant document chunks for RAG before calling an LLM
- ✓Agent upserts new document embeddings into an index when new content is ingested into a knowledge base
- ✓Agent deletes outdated vectors from the index when source documents are updated or removed
- ✓Agent performs semantic similarity search to find the most relevant past interactions or support tickets
- ✓Agent checks index stats to verify that embeddings were successfully ingested before proceeding
- ✓Agent creates a new namespace in the index to partition vectors by tenant, project, or document type
- ✓Agent fetches specific vectors by ID to inspect their metadata as part of a debugging or audit workflow
Agent actions
Inputs: indexName, vector, topK, namespace, filter, includeMetadata, includeValues
Returns: matches, matchCount, topScore, namespace
Inputs: indexName, vectors, namespace
Returns: upsertedCount, success
Inputs: indexName, ids, namespace
Returns: vectors, fetchedCount, namespace
Inputs: indexName, ids, filter, deleteAll, namespace
Returns: success
Inputs: indexName, filter
Returns: totalVectorCount, dimension, namespaces, stats
Returns: indexes, indexCount
Inputs: name, dimension, metric, cloud, region
Returns: indexName, host, index
Example workflows
RAG knowledge base query
An agent embeds a user query, queries Pinecone for the top-5 relevant chunks, and passes them as context to an LLM node.
Document ingestion pipeline
When new documents are uploaded, an agent chunks and embeds them, then upserts the vectors into Pinecone for future retrieval.
About Pinecone
- Vendor
- Pinecone Systems, Inc.
- Pricing Always review details with the vendor
- Free / Paid — Free Starter plan includes 2GB storage and 5 serverless indexes. Paid plans are usage-based (storage + read/write units).
- Authentication
- API key
- Rate limit Always review details with the vendor
- 100 requests / minute
- Compatible nodes
- AgentResourceInputOutput
- Website
- https://www.pinecone.io
Build an AI workflow with Pinecone
Use the Agentic Planner to design, visualize, and connect Pinecone with your other tools.
Open Agentic Planner