Missing credencials

[gafda]

Hey there!

Agno v1.3.1
AzureOpenAI (gpt-4o)

Agent initialization:

        self.agent = Agent(
            model=self.model,
            name=self.name,
            description=description,
            reasoning=True,
            system_message=system_prompt,
            tools=self._get_tools(),
            tool_call_limit=3,
            show_tool_calls=True,
            debug_mode=self._debug
        )

Whenever I run an agent with Reasoning=True, I get this block of errors with or without DEBUG turned on:

DEBUG =================================================================================== user ==================================================================================              
DEBUG Hi!,                                                                                                                       
ERROR    Error from OpenAI API: Missing credentials. Please pass one of `api_key`, `azure_ad_token`, `azure_ad_token_provider`, or the `AZURE_OPENAI_API_KEY` or `AZURE_OPENAI_AD_TOKEN`       
         environment variables.                                                                                                                                                                
WARNING  Attempt 1/1 failed: Missing credentials. Please pass one of `api_key`, `azure_ad_token`, `azure_ad_token_provider`, or the `AZURE_OPENAI_API_KEY` or `AZURE_OPENAI_AD_TOKEN`          
         environment variables.                                                                                                                                                                
ERROR    Failed after 1 attempts. Last error using AzureOpenAI(not-provided)                                                                                                                   
ERROR    Reasoning error: Missing credentials. Please pass one of `api_key`, `azure_ad_token`, `azure_ad_token_provider`, or the `AZURE_OPENAI_API_KEY` or `AZURE_OPENAI_AD_TOKEN` environment 
         variables.
DEBUG ============================================================================ Reasoning finished ===========================================================================              
DEBUG --------------------------------------------------------------------------- Azure Response Start --------------------------------------------------------------------------              
DEBUG --------------------------------------------------------------------------- Model: not-provided ---------------------------------------------------------------------------              

My environment variables with the tokens and keys, all start with APPRES_*.

The agent is able to complete its task to the finish but I fear I may be cripple it somehow.

Can anyone help?

[Ansub]

Thanks for sharing this — really appreciate the detailed breakdown.

From what you’ve described, it looks like the agent is falling back to a default behavior when reasoning=True, but can't fully execute the reasoning chain due to missing credentials. The key detail is that the environment variables you're using (APPRES_*) aren't being picked up by the model layer, which expects something like AZURE_OPENAI_API_KEY or AZURE_API_KEY.

To resolve this, you can either:
• Rename your environment variables to match what the model expects, or
• Map your internal ones to the expected names before initializing the agent:import os

os.environ["AZURE_OPENAI_API_KEY"] = os.getenv("APPRES_AZURE_KEY")
os.environ["AZURE_OPENAI_ENDPOINT"] = os.getenv("APPRES_AZURE_ENDPOINT")

Let me know if that helped 🙌

[gafda]

Hey! Thanks for the response.
So I've just tested your suggestion but it still results in a problem. A different one this time...

Here's some added information:

        azure_url: str = f"{app_config.APPRES_AZURE_ENDPOINT}/deployments/{app_config.APPRES_AZURE_DEPLOYMENT_NAME}"
        self.model: Model = AzureOpenAI(
            name = app_config.APPRES_COMPLETIONS_MODEL_ID,
            api_key = app_config.APPRES_AZURE_API_KEY.get_secret_value(),
            api_version = app_config.APPRES_AZURE_API_VERSION,
            azure_endpoint= azure_url,
            temperature=app_config.APPRES_TEMPERATURE
        )

        os.environ["AZURE_OPENAI_API_KEY"] = app_config.APPRES_AZURE_API_KEY.get_secret_value()
        os.environ["AZURE_OPENAI_ENDPOINT"] = azure_url

app_config is just a pydantic class that contains the mapped values from the APPRES_* environmental variables into variables.
As you can see, I've added your suggestion.

Executing, this code, the error changes to:

ERROR    API status error from OpenAI API: Error code: 404 - {'statusCode': 404, 'message': 'Resource not found'}                                                      
WARNING  Attempt 1/1 failed: Unknown model error

The agent is now Unable to complete the task interrupting it completely.