7 views
-/https://github.com/berriai/litellm/issues/26214
GitHub · issue

#26214 [Bug]: Unable to use Azure models through LiteLLM in Claude Cowork

  • State: open
  • Author: @ianaware
  • Labels: bug, proxy, llm translation

### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

Firstly, I may be completely barking up the wrong tree here, so I apologise, also, not that technical.

Okay, so I followed this and I can use LiteLLM Anthropic models inside the Claude Cowork app.

https://docs.litellm.ai/docs/tutorials/claude_desktop_cowork

However, in our LiteLLM, we have lots of GPT models from Azure.

I want to also be able to use those models inside Claude Cowork too.

When I try this, meaning that my LiteLLM key has only a GPT model selected, Cowork gets stuck and the LiteLLM usage log shows failure with this message:

Type: ValueError Message: Error calling litellm.acompletion for non-Anthropic model: litellm.BadRequestError: AzureException BadRequestError - Unknown parameter: 'output_config'.

I understand that Anthropic and Azure use different parameters.

Inside our config.yaml, I also added this to litellm_settings:

litellm_settings: drop_params: true

What else do I need to do to get these models to be able to work as well?

Any help is much appreciated - thanks.

### Steps to Reproduce

1. Set up Cowork…

GitHub resolver

Import GitHub neighbors on demand. Results are saved as system ingests.

Refresh page
vote history (5 events)
#0 of 0 · 31d18h49m56s ago — entered · #import:https:::github.com:berriai:litellm post #2239
The left issue is harder because it involves diagnosing and correcting cross-provider request translation in a proxy integration, with compatibility behavior and end-to-end client testing. The right issue is primarily structured metadata correction and coverage expansion, requiring broader data validation but less runtime integration work.
The right issue is harder because it spans cross-provider request translation, proxy behavior, parameter filtering, and compatibility testing across Azure and Anthropic-style clients. The left issue is comparatively narrower, focused on adjusting release or merge metadata handling and validating contributor attribution.
Azure-to-Anthropic compatibility crosses proxy request translation, provider-specific parameter filtering, and integration testing, creating broader uncertainty; the duration rollover issue is a localized calculation fix with targeted regression tests.
The right issue is harder because it requires diagnosing and safely adapting cross-provider request translation, handling provider-specific parameter compatibility, and adding regression coverage across proxy and model integrations. The left issue is comparatively localized to frontend static-export route generation and build validation.
#0 of 0 · 31d17h52m23s ago — current · #import:https:::github.com:berriai:litellm post #3206
Issue #26214 is harder because it requires cross-provider request translation, parameter compatibility analysis, and integration validation across the proxy and client workflow. Issue #28020 is a localized comparison fix with a focused regression test.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search