infrastructure-completion-response--badrequesterror-with-message-about-invalid-type-fo-7e57c4b9

THEN Either remove the `n` parameter from the completion call, or set `litellm.model_cost['model_name']['mode'] = 'completions'` to force LiteLLM to route to the completions endpoint instead of the responses endpoint. The root cause is a logic error that incorrectly maps completion parameters to response parameters. This bug was introduced by a PR that sent all completion calls to the responses endpoint and was subsequently reverted.

import litellm
litellm.model_cost["gpt-5-mini"]["mode"] = "completions"
response = litellm.completion(model="openai/gpt-5-mini", messages=[{"content": "Hello", "role": "user"}], n=1)