[Bug]: Routing deepseek-v4-pro through LiteLLM through Codex gives "An assistant message with 'tool_calls' must be followed by tool messages responding to each 'tool_call_id'" 400 error. #32992

Check for existing issues

• I have searched the existing issues and checked that my issue is not a duplicate.

What happened?

I am attempting to configure Codex to use LiteLLM as a proxy to DeepSeek due to Codex no longer supporting the chat/completions API. However, when running LiteLLM as a proxy, I obtain the following error:

Steps to Reproduce

1. Export DeepSeek API key
2. Run LiteLLM: litellm --model deepseek/deepseek-v4-pro
3. Run Codex and input a command

Relevant log output

{"error":{"message":"litellm.BadRequestError: DeepseekException - {\"error\":{\"message\":\"An assistant message with 'tool_calls' must
be followed by tool messages responding to each 'tool_call_id'. (insufficient tool messages following tool_calls message)\",\"type\":
\"invalid_request_error\",\"param\":null,\"code\":\"invalid_request_error\"}}","type":null,"param":null,"code":"400"}}

What part of LiteLLM is this about?

No response

What LiteLLM version are you on ?

1.92.0

Twitter / LinkedIn details

No response