Reject non-positive parallel processor limits

[sylvesterkaczmarek]

Summary

• Reject non-positive request and token rate limits before the parallel processor enters its scheduling loop.
• Require at least one request attempt so failed requests cannot be requeued indefinitely with a negative attempt counter.
• Keep all valid positive limits and retry behavior unchanged.

Motivation

process_api_requests_from_file() assumes its scheduling limits are positive, but currently accepts zero and negative values.

That can make the processor fail to terminate:

• max_requests_per_minute <= 0 means request capacity never reaches the >= 1 scheduling condition.
• max_tokens_per_minute <= 0 means token capacity cannot satisfy a positive request cost.
• max_attempts == 0 is decremented to -1 before the first call; if that call fails, if self.attempts_left: treats the negative value as truthy and requeues the request indefinitely.

Failing at the processor boundary prevents these hang/infinite-retry states without changing normal throttling behavior.

Closes #3012.

Validation

• positive request/token limits and max_attempts >= 1 -> existing behavior unchanged
• zero or negative request limit -> immediate ValueError
• zero or negative token limit -> immediate ValueError
• max_attempts < 1 -> immediate ValueError
• validation occurs before opening the request file or entering the scheduling loop

Self-review

• Final diff is seven added lines in one file.
• Restored an accidentally omitted documentation line before opening the PR; final compare contains no unrelated changes.
• No API request schema, throttling formula, retry behavior for valid inputs, dependency, notebook, or registry changes.
• Searched open PRs and found no competing fix.

Maintainers may modify the branch if needed.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4d2ce14743

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Reject request rates below one

When a direct caller passes a positive fractional rate such as 0.5—which is permitted by the float annotation and this validation—the request capacity is initialized and permanently capped below 1, so the scheduling condition available_request_capacity >= 1 can never succeed and a nonempty job runs indefinitely. Reject values below 1, or adjust the token-bucket logic to support fractional request rates.

Useful? React with 👍 / 👎.