Vector Databases Compared: Pinecone vs Weaviate vs Chroma vs pgvector

Choosing a Vector Database Is a Real Decision

Every RAG tutorial reaches for Chroma because it is easy to set up locally. Every enterprise vendor recommends Pinecone because they sponsor content. The reality is that the right vector database depends entirely on your scale, your existing infrastructure, and your operational complexity budget. This post gives you the honest tradeoffs between the four most common options.

pgvector: The Default Choice for Most Teams

pgvector is a PostgreSQL extension that adds vector storage and similarity search. If you already run PostgreSQL, pgvector should be your first choice — you get vector search without adding a new service to operate, monitor, and pay for.

When pgvector works well

• Up to ~1 million vectors with good performance on standard hardware
• You already use PostgreSQL and want to keep your stack simple
• You need to JOIN vectors with relational data in the same query
• You want to self-host without learning a new system

-- Enable the extension
CREATE EXTENSION vector;

-- Create a table with a vector column
CREATE TABLE documents (
  id SERIAL PRIMARY KEY,
  content TEXT,
  embedding vector(1536),
  metadata JSONB,
  created_at TIMESTAMP DEFAULT NOW()
);

-- Create an IVFFlat index for approximate search
CREATE INDEX ON documents 
USING ivfflat (embedding vector_cosine_ops)
WITH (lists = 100);

-- Similarity search
SELECT content, 1 - (embedding <=> $1) AS similarity
FROM documents
ORDER BY embedding <=> $1
LIMIT 10;

The IVFFlat index trades a small amount of recall for significant query speed improvements. For production, also consider HNSW index which provides better recall at the cost of higher memory usage.

Pinecone: Managed Scale With a Price Tag

Pinecone is the easiest path to high-scale vector search without infrastructure management. It handles indexing, replication, and query routing automatically. The tradeoff is cost — Pinecone's pricing model becomes expensive quickly at scale, and your data lives in their infrastructure.

When Pinecone makes sense

• More than 5 million vectors and you don't want to manage infrastructure
• You need guaranteed low-latency (<100ms P99) at scale
• Your team lacks infrastructure expertise
• You're building an early product and want to defer operational complexity

from pinecone import Pinecone, ServerlessSpec

pc = Pinecone(api_key=os.environ["PINECONE_API_KEY"])

pc.create_index(
    name="production-kb",
    dimension=1536,
    metric="cosine",
    spec=ServerlessSpec(cloud="aws", region="us-east-1")
)

index = pc.Index("production-kb")
index.upsert(vectors=[
    {"id": "doc-1", "values": embedding, "metadata": {"source": "faq.pdf"}}
])

Weaviate: Open-Source Power With Module Ecosystem

Weaviate is the most feature-rich open-source option. It supports hybrid search (combining BM25 keyword search with vector search), has native module integrations with OpenAI and Cohere, and provides a GraphQL API. It is more complex to operate than pgvector but more capable at scale.

When Weaviate fits

• You need hybrid search (keyword + semantic) out of the box
• Your corpus exceeds what comfortably fits in one PostgreSQL instance
• You want self-hosted scale without Pinecone's vendor lock-in
• You have an infrastructure team that can manage a distributed system

Chroma: Great for Development, Not Production

Chroma is the fastest way to get a vector store running in a Python script. It is persistent, requires no setup, and works in-process. It is not designed for production workloads and lacks the features (multi-tenancy, replication, access control) that production systems need.

Use Chroma for: local development, prototyping, running evals on your laptop. Migrate to pgvector, Pinecone, or Weaviate before going to production.

Decision Matrix

• Startup, under 1M vectors, already using PostgreSQL: pgvector
• Startup, need fast time-to-production, can afford managed: Pinecone serverless
• Scale-up, 1-50M vectors, need hybrid search, self-hosted: Weaviate
• Enterprise, 50M+ vectors, need SLAs: Pinecone or Weaviate Cloud
• Local dev and prototyping: Chroma

The most common mistake is choosing Pinecone early because it is the most-mentioned option in tutorials, then being surprised by the cost at scale. pgvector handles the majority of real-world RAG use cases at a fraction of the cost and operational overhead.