pinecone:docs

Use this documentation reference to fetch the relevant official Pinecone docs page for your task, such as creating an index or upserting data. It organizes links by topic so the agent retrieves current parameters rather than relying on outdated training data.

Standard index vectors use an object with id, values (a float array matching index dimension), and optional metadata. Integrated index records use _id plus a text field matching the index fieldMap, with other fields stored directly as metadata.

Yes, Pinecone supports sparse vectors with sparse_values containing indices and values arrays of equal length. These are used for keyword or hybrid search alongside dense vector values, as documented in the data formats reference.

Append .md to the documentation URL to retrieve the markdown version of the page. If that fails, consult https://docs.pinecone.io/llms.txt, which lists all relevant documentation URLs.

Integrated inference indexes embed text automatically using a configured model, so you upsert records with raw text fields. Standard indexes require you to generate embeddings yourself and upsert dense vector values directly.