BerriAI/litellm · error · AuthenticationError

str(e)

Error message

str(e)

What it means

Not a raised string itself — this is the chat-path re-wrap: `GetAccessTokenError` from the authenticator is converted to a litellm `AuthenticationError` whose message is `str(e)` of the original auth failure. Any device-flow/refresh failure (errors 1404–1415) surfaces through here on chat completion calls with the `chatgpt` provider.

Source

Thrown at litellm/llms/chatgpt/chat/transformation.py:37

        api_key: str | None = None,
        api_base: str | None = None,
        custom_llm_provider: str = "openai",
    ) -> None:
        super().__init__()
        self.authenticator = Authenticator()

    def _get_openai_compatible_provider_info(
        self,
        model: str,
        api_base: str | None,
        api_key: str | None,
        custom_llm_provider: str,
    ) -> tuple[str | None, str | None, str]:
        dynamic_api_base: Final = self.authenticator.get_api_base()
        try:
            dynamic_api_key: Final = self.authenticator.get_access_token()
        except GetAccessTokenError as e:
            raise AuthenticationError(
                model=model,
                llm_provider=custom_llm_provider,
                message=str(e),
            )
        return dynamic_api_base, dynamic_api_key, custom_llm_provider

    def validate_environment(
        self,
        headers: dict,
        model: str,
        messages: list[AllMessageValues],
        optional_params: dict,
        litellm_params: dict,
        api_key: str | None = None,
        api_base: str | None = None,
    ) -> dict:
        validated_headers: Final = super().validate_environment(
            headers, model, messages, optional_params, litellm_params, api_key, api_base

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Complete the ChatGPT device login flow once so a valid auth file exists, then retry.
2. If previously working, the refresh token likely expired/revoked — re-login to mint fresh tokens.
3. Check network reachability of OpenAI auth endpoints from the runtime host.
4. If you cannot use OAuth in your environment, use the `openai/` provider with an API key instead of `chatgpt/`.

Example fix

// before
litellm.completion(model="chatgpt/gpt-4o", messages=[...])  # raises AuthenticationError

// after (one-time interactive login first, then call)
# python -m litellm  # or the documented chatgpt login entrypoint
litellm.completion(model="chatgpt/gpt-4o", messages=[...])

import os

def chatgpt_auth_ready() -> bool:
    path = os.path.expanduser("~/.litellm/chatgpt_auth.json")  # adjust to your auth-file location
    return os.path.exists(path)

try:
    resp = litellm.completion(model="chatgpt/...", messages=msgs)
except litellm.AuthenticationError as e:
    logger.error("chatgpt auth failure: %s", str(e))
    prompt_relogin_or_fallback_to_openai_api_key()

When it happens

Trigger: Calling `litellm.completion(..., model="chatgpt/...")` (chat endpoint) when the authenticator cannot produce an access token: expired/revoked tokens, failed refresh, network failure during refresh, or never-completed device login.

Common situations: First use of the chatgpt provider without having logged in; expired auth file after long idle; servers where the interactive device flow was never completed; users expecting an API key pattern where ChatGPT requires OAuth.

Related errors

• Failed to request device code: {exc}
• Polling failed: {exc}
• Timed out waiting for device authorization
• Token exchange failed: {exc}
• Refresh token failed: {exc}

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/afa5f9359010f7b6. Report an issue: GitHub.