[davrens101]

Why are these changes needed?

CompletionRequest.max_tokens is Optional[int] and OpenAI allows "max_tokens": null. In create_completion, check_length returns an int for that case, but the following line compares it with the request value:

if isinstance(max_tokens, int) and max_tokens < request.max_tokens:

so /v1/completions with "max_tokens": null fails with TypeError: '<' not supported between instances of 'int' and 'NoneType' (HTTP 500) before anything is sent to a worker.

The fix treats None as "no limit requested", so the request is capped to the remaining context length (same as /v1/chat/completions, where max_tokens=None already works).

Before (mocked worker calls): TypeError; the new unit test fails. After: test passes.

Checks

• I've run format.sh to lint the changes in this PR. (black --check clean)
• I've included any doc changes needed. (none needed)
• I've made sure the relevant tests are passing (if applicable).

AI-assistance disclosure: prepared with AI assistance; I reproduced the failure and reviewed the diff myself.