Intermittent Authentication Error in test.sh Despite Configured API Key

[fangyan1221]

Describe the issue

Title: Intermittent Authentication Error in test.sh Despite Configured API Key

---

The content of test.sh is as follows:

...
mini-extra config set MSWEA_MINI_CONFIG_PATH "$CONFIG_PATH"
mini-extra config set LITELLM_MODEL_REGISTRY_PATH "$MODEL_REGISTRY_PATH"
mini-extra config set MSWEA_CONFIGURED 'true'
echo "success"

echo "start..."
mini -y -t "${ISSUE_DESCRIPTION}" --exit-immediately
...

Try run sh test.sh , the execution does not always succeed. Sometimes I get the following error:

�[1;31mGive Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new�[0m
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.
minisweagent: ERROR: Error running agent: litellm.AuthenticationError:          
AuthenticationError: OpenAIException - The api_key client option must be set    
either by passing api_key to the client or by setting the OPENAI_API_KEY        
environment variable You can permanently set your API key with `mini-extra      
config set KEY VALUE`. 

However, I clearly already configured the KEY-VALUE pair in the mini-extra config set with CONFIG_PATH. Why do I still get this error? What's more puzzling is that with the exact same test.sh file, sometimes it runs successfully when executed repeatedly, while other times it fails with this authentication error asking me to run mini-extra config set.

Has anyone else encountered a similar issue? Are there any known solutions?

[klieret]

Hi @fangyan1221

Wait, but where do you set your actual API key for Open AI?

So to provide some more context, this warning message is shown every time that litellm issues an authentication error. The bit about mini-extra config set is something we add ad-hoc.

    def _query(self, messages: list[dict[str, str]], **kwargs):
        try:
            return litellm.completion(
                model=self.config.model_name, messages=messages, **(self.config.model_kwargs | kwargs)
            )
        except litellm.exceptions.AuthenticationError as e:
            e.message += " You can permanently set your API key with `mini-extra config set KEY VALUE`."
            raise e

So I think it's most likely to assume that you didn't provide your OPENAI_API_KEY (or whatever API key you authenticate with for your model), or that the key is wrong.

If you didn't set it up with mini-extra config set, then maybe you sometimes had it set as an environment variable but didn't set it again when you restarted the session?

[fangyan1221]

@klieret hello，
I followed the documentation at https://mini-swe-agent.com/latest/models/local_models/

and configured the relevant api_base and api_key within model_kwargs. Then, I set the environment variable with mini-extra config set MSWEA_MINI_CONFIG_PATH "$CONFIG_PATH", and executed the mini command.

Currently, most of the time, this workflow works fine (which proves that my API key is correct). However, in a few cases, I still encounter the authentication error mentioned earlier. In my process, every time before running mini, I always run mini-extra config set MSWEA_MINI_CONFIG_PATH "$CONFIG_PATH" to initialize.

I’m not sure why this issue occurs. Is it possible that the environment variable isn’t being set successfully?