### Check for existing issues

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

### What happened?

I am using LiteLLM as an anthropic_proxy to route Claude Code requests to a model (Kimi-K2.6) hosted on Microsoft Foundry (formerly Azure AI Foundry). Two distinct issues occur:

Parameter Mapping Failure: Claude Code sends "thinking/reasoning" parameters. LiteLLM translates these to reasoning.effort (OpenAI format). Microsoft Foundry models (non-o1) reject the request with a 400 Bad Request: Unsupported parameter: 'reasoning.effort'. Even with drop_params: true, this specific parameter persists because it is a "valid" OpenAI key, despite being unsupported by the specific deployment.

Telemetry/Logging Crash (Non-Blocking): After a successful response (if params are manually dropped), LiteLLM's background logging fails. It attempts to validate a ResponseIncompleteEvent or ResponseCompletedEvent against the AnthropicResponse Pydantic model, causing a ValidationError.

### Steps to Reproduce

1. Run LiteLLM Proxy with the following config.yaml:

model_list: - model_name: Kimi-K2.6 litellm_params: model: openai/your-deployment ap…