Fixed #279

[monarchwadia]

Why are these changes needed?

This is a popular issue, #279 which seems to be happening when an OpenAI-compatible API has a specific minor discrepancy from spec in the error response scenario.

Sometimes, when building an OpenAI-compatible API, the main inference scenario is built very well but the API's error response is not fully compliant.

I was building an OpenAI-compatible endpoint. In my case, my error response had an error object that was a string instead of an object.

This was my version, which was slightly off

    const errorObject = {
        "error":  "Authorization failed: there was a problem with the API key."
    }

This is what OpenAI would have returned

    const errorObject = {
        "error": {
          "message": "Authorization failed: there was a problem with the API key.",
          "type": "invalid_request_error",
          "param": null,
          "code": null
      }
    }

error_code=error_data.get("code") will throw because it is expecting error_code to be an object with a property code and in my case, error was a string.

Related issue number

#279

Checks

• (None needed) I've included any doc changes needed for https://microsoft.github.io/autogen/. See https://microsoft.github.io/autogen/docs/Contribute#documentation to build and test documentation locally.
• (None needed) I've added tests (if relevant) corresponding to the changes introduced in this PR.
• I've made sure all auto checks have passed.

[codecov-commenter]

Codecov Report

Attention: 2 lines in your changes are missing coverage. Please review.

Comparison is base (c705c6a) 27.82% compared to head (29eb6df) 37.29%.

Files	Patch %	Lines
autogen/oai/completion.py	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #757      +/-   ##
==========================================
+ Coverage   27.82%   37.29%   +9.47%     
==========================================
  Files          27       27              
  Lines        3479     3480       +1     
  Branches      788      789       +1     
==========================================
+ Hits          968     1298     +330     
+ Misses       2440     2062     -378     
- Partials       71      120      +49     

Flag	Coverage Δ
unittests	37.24% <0.00%> (+9.47%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[sonichi]

Thanks for the fix. Please note that this file is for the old openai api <1. For openai>=1, we have switched to the OpenAIWrapper in client.py.