BerriAI/litellm · error · ValueError

Azure OpenAI client is not initialized. Make sure api_key is

Error message

Azure OpenAI client is not initialized. Make sure api_key is passed or AZURE_API_KEY is set in the environment.

What it means

AzureFineTuningAPI.create_fine_tuning_job builds an OpenAI-family client via get_openai_client(api_key, api_base, ...); if that resolver returns None (no client= argument, no api_key, and AZURE_API_KEY absent from the environment) this ValueError is raised. It is a pre-flight credential check — no fine-tuning API request is made.

Source

Thrown at litellm/llms/azure/fine_tuning/handler.py:83

        timeout: float | httpx.Timeout,
        max_retries: int | None,
        organization: str | None,
        client: OpenAI | AsyncOpenAI | AzureOpenAI | AsyncAzureOpenAI | None = None,
    ) -> LiteLLMFineTuningJob | Coroutine[Any, Any, LiteLLMFineTuningJob]:
        self._ensure_training_type(create_fine_tuning_job_data)

        openai_client: Final[OpenAI | AsyncOpenAI | AzureOpenAI | AsyncAzureOpenAI | None] = self.get_openai_client(
            api_key=api_key,
            api_base=api_base,
            timeout=timeout,
            max_retries=max_retries,
            organization=organization,
            client=client,
            _is_async=_is_async,
            api_version=api_version,
        )
        if openai_client is None:
            raise ValueError(
                "Azure OpenAI client is not initialized. Make sure api_key is passed or AZURE_API_KEY is set in the environment."
            )

        if _is_async is True:
            if not isinstance(openai_client, (AsyncOpenAI, AsyncAzureOpenAI)):
                raise ValueError(
                    "OpenAI client is not an instance of AsyncOpenAI. Make sure you passed an AsyncOpenAI client."
                )
            return self.acreate_fine_tuning_job(
                create_fine_tuning_job_data=create_fine_tuning_job_data,
                openai_client=openai_client,
            )

        verbose_logger.debug("creating fine tuning job, args= %s", create_fine_tuning_job_data)
        response: Final = cast(OpenAI, openai_client).fine_tuning.jobs.create(**create_fine_tuning_job_data)
        return _litellm_fine_tuning_job_from_response(response, is_azure=True)

    def cancel_fine_tuning_job(

View on GitHub (pinned to 6c2dcb801b)

Solutions

Pass api_key=os.environ['AZURE_API_KEY'] to create_fine_tuning_job.
Or export AZURE_API_KEY in the runtime environment.
Or pass a pre-built AzureOpenAI/AsyncAzureOpenAI client via client=.
Add a startup validation step that fails fast when no Azure credential is discoverable.

Example fix

# before
job = azure_finetune.create_fine_tuning_job(data)

# after
job = azure_finetune.create_fine_tuning_job(
    data,
    api_key=os.environ['AZURE_API_KEY'],
    api_base='https://<resource>.openai.azure.com',
)

Defensive patterns

Strategy: validation

Validate before calling

import os

def validate_fts_credentials(api_key: str | None, client: object | None) -> None:
    if client is None and not api_key and not os.environ.get('AZURE_API_KEY'):
        raise RuntimeError('fine-tuning needs api_key, a client, or AZURE_API_KEY in env')

Try / catch

try:
    job = ft.create_fine_tuning_job(data, api_key=api_key)
except ValueError as e:
    if 'not initialized' in str(e):
        raise RuntimeError('Azure fine-tuning credential missing (AZURE_API_KEY)') from e
    raise

Prevention

Set AZURE_API_KEY via your secret manager, not ad-hoc exports.
Validate credentials in a preflight check before submitting training data.
Note this handler checks AZURE_API_KEY, not OPENAI_API_KEY.

When it happens

Trigger: create_fine_tuning_job(create_fine_tuning_job_data) with no api_key, no client, and AZURE_API_KEY unset; pipelines running in containers where the Azure key secret was not injected.

Common situations: Automation that sets OPENAI_API_KEY for OpenAI jobs but forgets AZURE_API_KEY for the Azure variant; rotated/expired Azure keys removed from the environment; multi-region deployments where one region lacks the env var.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/2c0400d3215dccd8. Report an issue: GitHub.