openai/openai-python · error · OpenAIError

The Bedrock `endpoint` must be either `mantle` or `runtime`.

Error message

The Bedrock `endpoint` must be either `mantle` or `runtime`.

What it means

The `bedrock()` provider constructor only accepts the endpoint names `mantle` or `runtime` (or None for auto-detection). Any other string, including case variants like `Mantle` or full URLs, is rejected immediately with OpenAIError before any client is built.

Solutions

  1. Set endpoint to exactly 'mantle' or 'runtime', or omit it to let the provider infer from base_url
  2. Pass full hostnames via base_url instead of endpoint

Example fix

// before
bedrock(endpoint='bedrock-runtime')
// after
bedrock(endpoint='runtime')
Defensive patterns

Strategy: validation

Validate before calling

if endpoint is not None and endpoint not in {'mantle', 'runtime'}:
    raise ValueError('endpoint must be mantle or runtime')

Type guard

from typing import Literal

def is_bedrock_endpoint(v: object) -> bool:
    return v in ('mantle', 'runtime')

Try / catch

try:
    bedrock(endpoint=endpoint)
except OpenAIError as e:
    if 'endpoint' in str(e):
        # fall back to auto-detection
        bedrock()

Prevention

When it happens

Trigger: Calling `openai.providers.bedrock(endpoint='bedrock-runtime', ...)` or passing a hostname/URL such as 'https://bedrock-runtime.us-east-1.amazonaws.com' as endpoint; also typos or wrong casing like 'Runtime'.

Common situations: Developers confuse the logical endpoint name with the AWS service hostname, or copy endpoint values from AWS docs instead of the SDK's `mantle`/`runtime` enum.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

AI-assisted analysis of openai/openai-python@9917c6e28e (2026-08-28). Data as JSON: /api/errors/600017e43ebe5f8f. Report an issue: GitHub.

Appendix: source

Thrown at src/openai/providers/bedrock.py:399


def bedrock(
    *,
    endpoint: BedrockEndpoint | None = None,
    region: str | None = None,
    base_url: str | httpx2.URL | None | NotGiven = NOT_GIVEN,
    api_key: str | None | NotGiven = NOT_GIVEN,
    token_provider: BedrockTokenProvider | None = None,
    access_key_id: str | None = None,
    secret_access_key: str | None = None,
    session_token: str | None = None,
    profile: str | None = None,
    credential_provider: AwsCredentialsProvider | None = None,
) -> _Provider:
    """Configure the standard OpenAI client for Amazon Bedrock Mantle or Runtime."""

    if endpoint is not None and endpoint not in {"mantle", "runtime"}:
        raise OpenAIError("The Bedrock `endpoint` must be either `mantle` or `runtime`.")

    normalized_region = _normalize_optional_string(region)
    if region is not None and normalized_region is None:
        raise OpenAIError("The Bedrock AWS `region` must not be empty.")
    _validate_bedrock_region(normalized_region)

    region_source: Literal["explicit", "environment"] | None = "explicit" if normalized_region is not None else None

    configured_base_url: httpx2.URL | None
    if isinstance(base_url, NotGiven):
        environment_base_url = _normalize_optional_string(os.environ.get("AWS_BEDROCK_BASE_URL"))
        configured_base_url = _normalize_base_url(environment_base_url) if environment_base_url else None
    elif base_url is None:
        configured_base_url = None
    else:
        if isinstance(base_url, str) and not base_url.strip():
            raise OpenAIError("The Bedrock `base_url` must not be empty.")
        configured_base_url = _normalize_base_url(base_url)

View on GitHub (pinned to 9917c6e28e)