What is Pinecone?

Pinecone is a cloud-native vector database facilitating long-term memory for high-performing AI applications through optimized storage and quick querying of vector embeddings. Each record within Pinecone indexes includes a unique ID and a dense vector embedding, with optional sparse vector embeddings and metadata key-value pairs for hybrid search and filtered queries. The platform ensures low-latency, accurate results for large-scale indexes, supporting up to 200 queries per second per replica. Users can perform CRUD operations and query vectors via HTTP, Python, or Node.js using the Pinecone API.

Which cloud providers and regions is Pinecone available in?

Please see the full list in our documentation for the full list of cloud regions and providers in which Pinecone is currently available.

Can Pinecone handle real-time data updates, and how?

Yes, Pinecone is designed to handle real-time data updates efficiently. Use Pinecone’s upsert function to add or update data in real time. It’s vital to balance update frequency with performance needs.

When deleting large numbers of items by metadata, we recommend separating this into two operations: a query to get the list of items to be deleted and then an iteration through that list to delete in batches of 100 by the ID of the vectors. Please see How to handle large numbers of deletes by metadata for more details on this approach.

Are there quota limits in Pinecone?

Pinecone does not set quotas for API calls or queries. We often see people report 429 errors when using Langchain with Pinecone, but these errors are more likely to come from the LLM you’re using to generate your embeddings rather than Pinecone directly.

We do have quotas in place for how many pods you can have running at one time in your projects and organizations. These can be edited by you in the console.

When you create a new project on the pod-based architecture, you must set a pod limit. This is just a safeguard against accidentally creating too many pods and having a larger-than-expected bill. You can increase your limit up to 100 pods; if you need more than that, please open a support ticket, and we will work with you to increase the limit.

For projects using the serverless architecture, you will not be required to set a pod limit. This is because serverless frees users from needing to manage their own infrastructure. Rather, Pinecone will automatically scale resources for you, and you will only be charged for the storage, read units, and write units that you use.

What are some common use cases for Pinecone?

Common use cases include recommendation systems, similarity search for text and images, fraud detection, and more. Pinecone’s versatility makes it suitable for a wide range of applications.

Where can I find more resources or support for Pinecone?

Pinecone offers a range of resources, including documentation, community forums, and customer support. Visit our website for more information and to access these resources.