Pinecone

More relevant results make better applications

• Filter by metadata: Combine vector search with familiar metadata filters to get just the results you want.
• Find context: Fast and accurate vector search over all your data.
• Update in real time: As your data changes, the Pinecone index is updated in realtime to provide the freshest results.
• Make (the right) keywords matter: Combine vector search with keyword boosting for the best of both worlds (hybrid search).

​

Bring AI products to market faster

Start building for free in minutes. Forget about configuring or scaling your index.

• SDKs: Growing number of SDKs including Python, Node, and Java make working with Pinecone a breeze for developers.
• Streamlined API: Manage control and data plane requests across environments with a single API.
• Any AI Model: Compatible with embeddings from any AI model or LLM, including those from OpenAI, Anthropic, Cohere, Hugging Face, PaLM, etc.
• Integrations: Supercharge your AI stack with integrations for popular data sources, frameworks, models, and more.

​

Get just the results you want

Always fresh, relevant results as your data changes and grows.

• Hybrid search: Combine vector search with keyword boosting for the best of both worlds (hybrid search).
• Namespaces: Partition your workload with namespaces to minimize latency and compute needed for query.
• Metadata Filtering: Combine vector search with familiar metadata filters to get just the results you want.
• Live index updates: As your data changes, the Pinecone index is updated in realtime to provide the freshest results.

​

The vector database reimagined

Build your next great GenAI apps with our industry-first architecture.

• Efficient query-planning: Built-in logic to scan the optimal number of semantically similar clusters needed for query, not the entire index.
• Durable writes: Write requests are committed to a write-ahead-log in object storage for guaranteed durability and strong ordering.
• Adaptive clustering: Indexes automatically adapt as data grows to maintain low-latency and O(s) freshness.
• Multi-tenant layer: Built to efficiently manage thousands of tenants without performance degradation.
• Intelligent retrieval: Only the most used clusters are cached in memory instead of loading from object storage for quick, memory efficient retrieval.