Create an index with integrated embedding

With this type of index, you provide source text, and Pinecone uses a hosted embedding model to convert the text automatically during upsert and search.

The response is this version's index model, with the embedding configuration surfaced as a semantic_text field in the index schema, named after the field_map text entry. Read and write the index through the records API.

For guidance and examples, see Create an index.

from pinecone import Pinecone

pc = Pinecone(api_key="YOUR_API_KEY")

index_model = pc.create_index_for_model(
    name="docs-example",
    cloud="aws",
    region="us-east-1",
    embed={
        "model": "multilingual-e5-large",
        "field_map": {"text": "chunk_text"}
    }
)

PINECONE_API_KEY="YOUR_API_KEY"

curl "https://api.pinecone.io/indexes/create-for-model" \
  -H "Api-Key: $PINECONE_API_KEY" \
  -H "Content-Type: application/json" \
  -H "X-Pinecone-Api-Version: 2026-07" \
  -d '{
    "name": "docs-example",
    "cloud": "aws",
    "region": "us-east-1",
    "embed": {
      "model": "multilingual-e5-large",
      "field_map": {"text": "chunk_text"}
    }
  }'

POST /indexes/create-for-model

• X-Pinecone-Api-Version (header, string, required) — Required date-based version header

• name (body, string, required) — The name of the index. Resource name must be 1-45 characters long, start and end with an alphanumeric character, and consist only of lower case alphanumeric characters or '-'.
• cloud (body, string, required) — The public cloud where you would like your index hosted. Possible values: gcp, aws, or azure.
• region (body, string, required) — The region where you would like your index to be created.
• deletion_protection (body, string) — Whether deletion protection is enabled/disabled for the index. Possible values: disabled or enabled.
• tags (body, object) — Custom user tags added to an index, at most 20 per index. Keys must be 80 characters or less and alphanumeric, '_', or '-'. Values must be 120 characters or less and consist of printable ASCII characters or spaces. To unset a key, set the value to be an empty string. null in responses when the index has no tags.
• schema (body, object) — Schema for the behavior of Pinecone's internal metadata index. By default, all metadata is indexed; when schema is present, only fields which are present in the fields object with a filterable: true are indexed. Note that filterable: false is not currently supported.
• read_capacity (body, object) — By default the index will be created with read capacity mode OnDemand. If you prefer to allocate dedicated read nodes for your workload, you must specify mode Dedicated and additional configurations for node_type and scaling. BYOC indexes do not support OnDemand: a byoc deployment must set mode: Dedicated explicitly, since omitting read_capacity defaults to OnDemand and is reject
• embed (body, object, required) — Specify the integrated inference embedding configuration for the index. Once set the model cannot be changed, but you can later update the embedding configuration for an integrated inference index including field map, read parameters, or write parameters. Refer to the model guide for available models and model details.

• 201 — The index has successfully been created for the embedding model.
• 400 — Bad request. The request body included invalid request parameters.
• 401 — Unauthorized. Possible causes: missing or invalid API key.
• 402 — Payment required. Organization is on a paid plan and is delinquent on payment.
• 403 — Forbidden. Possible causes: the project's index quota is exhausted, dedicated read capacity is not enabled for the organization, or (when configuring) the index's read capacity is not yet Ready.
• 404 — Unknown cloud or region when creating a serverless index.
• 409 — Index of given name already exists.
• 422 — Unprocessable entity. The request body could not be deserialized.
• 500 — Internal server error.