add timeout arg to server method

[grossjct]

Closes #455

Changes proposed in this pull request:

• Use configurable timeout for HTTP client in run_server function

[Copilot]

Pull request overview

Adds a configurable HTTP client timeout to the olmocr bench “server” runner, addressing issue #455 where the previous fixed 300s timeout was insufficient for some use cases.

Changes:

• Add a timeout parameter (default 300) to run_server(...).
• Use the new timeout value when constructing httpx.AsyncClient.

Comments suppressed due to low confidence (1)

olmocr/bench/runners/run_server.py:41

• Now that timeout is configurable, it would help to document its units/behavior in the docstring (e.g., seconds, passed through to httpx.AsyncClient(timeout=...)). Also consider typing it as float (vs int) since httpx accepts floats and callers via parse_method_arg may supply non-integer values.

    timeout: int = 300,
) -> str:
    """
    Convert page of a PDF file to markdown by calling a request
    running against an openai compatible server.

    You can use this for running against vllm, sglang, servers
    as well as mixing and matching different model's.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.