BerriAI/litellm · error · ValueError

Missing CODESTRAL_API_Key - Please add CODESTRAL_API_Key to

Error message

Missing CODESTRAL_API_Key - Please add CODESTRAL_API_Key to your environment variables

What it means

CodestralTextCompletion._validate_environment requires an API key before any request is sent. If api_key is None (LiteLLM could not find CODESTRAL_API_KEY in the environment and none was passed per-call), it raises this ValueError immediately. The header building (Authorization: Bearer ...) never happens.

Source

Thrown at litellm/llms/codestral/completion/handler.py:86

        api_key="",
        original_response=completion_stream,  # Pass the completion stream for logging
        additional_args={"complete_input_dict": data},
    )

    return completion_stream


class CodestralTextCompletion:
    def __init__(self) -> None:
        super().__init__()

    def _validate_environment(
        self,
        api_key: str | None,
        user_headers: dict,
    ) -> dict:
        if api_key is None:
            raise ValueError("Missing CODESTRAL_API_Key - Please add CODESTRAL_API_Key to your environment variables")
        headers = {
            "content-type": "application/json",
            "Authorization": f"Bearer {api_key}",
        }
        if user_headers is not None and isinstance(user_headers, dict):
            headers = {**headers, **user_headers}
        return headers

    def output_parser(self, generated_text: str):
        """
        Parse the output text to remove any special characters. In our current approach we just check for ChatML tokens.

        Initial issue that prompted this - https://github.com/BerriAI/litellm/issues/763
        """
        chat_template_tokens: Final = [
            "<|assistant|>",
            "<|system|>",
            "<|user|>",

View on GitHub (pinned to 6c2dcb801b)

Solutions

Export CODESTRAL_API_KEY with the key issued for the Codestral (FIM) endpoint.
Or pass api_key explicitly on the call.
If you intend to use the general Mistral chat endpoint instead, call the mistral/ model rather than the codestral text-completion route.
Note the case in the message: the variable name LiteLLM reads is CODESTRAL_API_KEY.

Example fix

# before
completion = litellm.text_completion(model="text-completion-codestral/codestral-latest", prompt="def fib(")

# after
os.environ["CODESTRAL_API_KEY"] = "..."
completion = litellm.text_completion(model="text-completion-codestral/codestral-latest", prompt="def fib(")

Defensive patterns

Strategy: validation

Validate before calling

import os

if not (os.environ.get("CODESTRAL_API_KEY") or os.environ.get("MISTRAL_API_KEY")):
    raise SystemExit("Set CODESTRAL_API_KEY (separate from the Mistral chat key) for FIM models")

Try / catch

try:
    resp = litellm.text_completion(model="text-completion-codestral/codestral-latest", prompt=p)
except ValueError as e:
    if "CODESTRAL_API_Key" in str(e):
        raise RuntimeError("Codestral key missing") from e
    raise

Prevention

Remember the Codestral FIM endpoint uses its own key, distinct from MISTRAL_API_KEY.
Inject secrets via the environment at process start, not mid-session.
Add an integration test exercising each configured provider with a minimal request.

When it happens

Trigger: Using a codestral/* text-completion model without CODESTRAL_API_KEY exported and without api_key passed to the call; also when the key was configured under a different name (e.g. MISTRAL_API_KEY) for the dedicated Codestral endpoint.

Common situations: The Codestral FIM endpoint uses a separate key from the general Mistral API; developers set MISTRAL_API_KEY only and hit this. Also common in fresh containers/CI where env vars were not injected.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/464bea29138ef134. Report an issue: GitHub.