What is PineconeVectorStore?
PineconeVectorStore is LangChain's integration with Pinecone, a fully-managed cloud vector database, imported from the langchain-pinecone package. It provides indexing, similarity search, and metadata filtering at scale without managing server infrastructure.
Before using PineconeVectorStore, create a Pinecone index with the correct dimension—it must match your embedding model's output size (1536 for text-embedding-ada-002, 3072 for text-embedding-3-large). Mismatched dimensions raise a Pinecone API error at upsert time, not at initialization. The PINECONE_API_KEY environment variable is read automatically, or pass api_key explicitly.
LangChain's PineconeVectorStore exposes similarity_search(), max_marginal_relevance_search(), and add_documents(). For production RAG, Pinecone is preferred over in-process stores like Chroma because it is durable, multi-process safe, and horizontally scalable. The free Serverless tier allows one index and two million vectors—sufficient for prototyping.
When to Use
You're deploying production RAG at scale. Use Pinecone for high availability and automatic scaling.
Use Cases
- • Production RAG
- • Scalable search
- • Multi-tenant systems
- • Real-time recommendations
- • Enterprise search
- • Large datasets
Key Features
- ✓ Cloud-managed
- ✓ Auto-scaling
- ✓ High availability
- ✓ Metadata filtering
- ✓ Hybrid search
- ✓ Enterprise features
When NOT to Use
For development—use Chroma. For on-premise.
Notes
Index dimension must match your embedding model
Create the Pinecone index with dimension=1536 for OpenAI ada-002, dimension=768 for Cohere embed-english-v3, etc. PineconeVectorStore does not validate this at init—the mismatch only surfaces as a Pinecone API 400 error during the first upsert.
Namespace isolation for multi-tenancy
Pass namespace="user-123" to similarity_search() and add_documents() to shard a single Pinecone index by tenant. This avoids the cost of one index per tenant on paid plans.
Metadata filter syntax
Pass filter={"source": "contracts"} to similarity_search(). Metadata fields must be indexed in Pinecone before they can be filtered. Use describe_index_stats() to verify which fields are indexed.
from_documents vs add_documents
PineconeVectorStore.from_documents(docs, embeddings, index_name="...") is a one-shot class method for bulk upsert. For incremental updates, instantiate the class and call add_documents() instead.
PINECONE_API_KEY environment variable
Set PINECONE_API_KEY in your environment before instantiating PineconeVectorStore. The Pinecone client reads it automatically. Without it, the constructor raises a PineconeException. Never hardcode API keys in source files—use python-dotenv or a secrets manager.
Import
from langchain_pinecone import PineconeVectorStoreInitialization Parameters
| Parameter | Type | Default | Purpose |
|---|---|---|---|
| index_name | str | None | Pinecone index name |
Code Examples
Use Pinecone
from langchain_pinecone import PineconeVectorStore
from pinecone import Pinecone
pc = Pinecone(api_key='...')
vector_store = PineconeVectorStore(index_name='my-index', embedding=embeddings)Search with Metadata Filter
results = vector_store.similarity_search(
'What is LangChain?', k=4,
filter={'source': 'docs'}
)Add Documents Incrementally
from langchain_core.documents import Document
new_docs = [Document(page_content="LangChain v0.3 released",
metadata={'source': 'news'})]
vector_store.add_documents(new_docs)Common Mistakes
❌ Forget to create index first
✅ Create Pinecone index before using
Alternative Vector Stores
| Store | When to Use |
|---|---|
| Chroma | For local development |
Related LangChain References
Browse the full LangChain API reference index to explore more classes, methods, and decorators, or start with the LangChain introduction tutorial for end-to-end context on building with PineconeVectorStore and the wider framework.
PineconeVectorStore FAQ
What is PineconeVectorStore in LangChain?
Use Pinecone cloud vector database with LangChain. PineconeVectorStore is LangChain's integration with Pinecone, a fully-managed cloud vector database, imported from the langchain-pinecone package. It provides indexing, similarity search, and metadata filtering at scale without managing server infrastructure. Before using PineconeVectorStore, create a Pinecone index with the correct dimension—it must match your embedding model's output size (1536 for text-embedding-ada-002, 3072 for text-embedding-3-large). Mismatched dimens…
Which package provides PineconeVectorStore?
DevShelfHub documents PineconeVectorStore from the langchain-pinecone package. Pin your installed LangChain version and match imports to the snippet on this page.
When should I use PineconeVectorStore?
You're deploying production RAG at scale. Use Pinecone for high availability and automatic scaling.
When should I avoid using PineconeVectorStore?
For development—use Chroma. For on-premise.
How do I import PineconeVectorStore in Python?
from langchain_pinecone import PineconeVectorStore
Where can I explore more LangChain API reference pages?
Open the LangChain API reference index on DevShelfHub to browse classes, methods, and decorators, each with runnable examples, parameters, common mistakes, and cross-links.