BerriAI/litellm · error · ValueError

api_base is required for Azure AI Studio. Please set the api

Error message

api_base is required for Azure AI Studio. Please set the api_base parameter. Passed `api_base={api_base}`

What it means

The Azure AI Studio (Foundry) chat transformation requires a base URL to build https://<resource>.services.ai.azure.com/models/chat/completions?api-version=.... If api_base is None it raises ValueError, echoing the (None) value you passed. Foundry has no inferable default endpoint, so this is thrown client-side before any request.

Source

Thrown at litellm/llms/azure_ai/chat/transformation.py:120

    ) -> str:
        """
        Constructs a complete URL for the API request.

        Args:
        - api_base: Base URL, e.g.,
            "https://litellm8397336933.services.ai.azure.com"
            OR
            "https://litellm8397336933.services.ai.azure.com/models/chat/completions?api-version=2024-05-01-preview"
        - model: Model name.
        - optional_params: Additional query parameters, including "api_version".
        - stream: If streaming is required (optional).

        Returns:
        - A complete URL string, e.g.,
        "https://litellm8397336933.services.ai.azure.com/models/chat/completions?api-version=2024-05-01-preview"
        """
        if api_base is None:
            raise ValueError(
                f"api_base is required for Azure AI Studio. Please set the api_base parameter. Passed `api_base={api_base}`"
            )
        original_url: Final = httpx.URL(api_base)

        # Extract api_version or use default
        api_version: Final = cast(str | None, litellm_params.get("api_version"))

        # Create a new dictionary with existing params
        query_params: Final = dict(original_url.params)

        # Add api_version if needed
        if "api-version" not in query_params and api_version:
            query_params["api-version"] = api_version

        # Add the path to the base URL
        if "services.ai.azure.com" in api_base:
            new_url = _add_path_to_api_base(api_base=api_base, ending_path="/models/chat/completions")
        else:

View on GitHub (pinned to 6c2dcb801b)

Solutions

Pass api_base: https://<resource>.services.ai.azure.com (optionally with /models/chat/completions?api-version=... which litellm normalizes).
Or set AZURE_AI_API_BASE in the process environment.
If you actually have an Azure OpenAI resource (not Foundry), switch the model prefix from azure_ai/ to azure/ and use AZURE_API_BASE.
In proxy config, add api_base under the deployment's litellm_params.

Example fix

# before
litellm.completion(model='azure_ai/gpt-4o', api_key=k, messages=m)

# after
litellm.completion(
    model='azure_ai/gpt-4o', api_key=k, messages=m,
    api_base='https://litellm8397336933.services.ai.azure.com',
)

Defensive patterns

Strategy: validation

Validate before calling

import os

def foundry_chat_base(api_base: str | None) -> str:
    base = api_base or os.getenv('AZURE_AI_API_BASE')
    if base is None:
        raise RuntimeError('azure_ai chat requires api_base or AZURE_AI_API_BASE')
    return base

Try / catch

try:
    litellm.completion(model='azure_ai/gpt-4o', api_base=foundry_chat_base(None), ...)
except ValueError as e:
    if 'api_base is required for Azure AI Studio' in str(e):
        raise ConfigurationError(str(e)) from e
    raise

Prevention

Choose the prefix deliberately: azure/ uses AZURE_API_BASE, azure_ai/ needs AZURE_AI_API_BASE.
Validate provider env vars at process start with a single settings loader.
In proxy configs, always set api_base on foundry deployments — never depend on ambient env.

When it happens

Trigger: completion(model='azure_ai/<model>', ...) without api_base and without AZURE_AI_API_BASE env var; supplying only deployment name and api_key (the pattern for plain Azure OpenAI, which resolves via AZURE_API_BASE); passing api_base=None explicitly.

Common situations: Copy-pasted Azure OpenAI config used against an azure_ai model; env var set after process start; proxy deployment entry missing api_base; confusion between the azure/ and azure_ai/ provider prefixes.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/50e838402adc3a22. Report an issue: GitHub.