fix(tools): support CPython 3.14 WorkerContext in DaemonThreadPoolExecutor

[sahilm-ai]

Problem

CPython 3.14 replaced ThreadPoolExecutor's (initializer, initargs) attribute pair + 4-arg _worker free function with a WorkerContext object built via prepare_context() / self._create_worker_context(). DaemonThreadPoolExecutor._adjust_thread_count mirrored the pre-3.14 internals directly (reading self._initializer / self._initargs), so on 3.14 every worker spawn raised:

AttributeError: 'DaemonThreadPoolExecutor' object has no attribute '_initializer'

inside the worker thread — silently breaking every batch of 2+ concurrent tool calls, since agent/tool_executor.py routes concurrent tool execution through this pool.

CI runs Python 3.13 (requires-python = ">=3.11,<3.14"), so this never surfaced there. It only shows up on interpreters that have moved to 3.14 (e.g. local dev venvs pinned via pyvenv.cfg to /opt/homebrew/opt/python@3.14).

Empirical repro (pre-fix, on 3.14.6)

$ python3 -c "from tools.daemon_pool import DaemonThreadPoolExecutor; \
    DaemonThreadPoolExecutor(max_workers=2).submit(lambda: 1+1).result()"
AttributeError: 'DaemonThreadPoolExecutor' object has no attribute '_initializer'

Fix

Branch on sys.version_info at import time and mirror whichever shape of _worker/_adjust_thread_count the running interpreter actually has, confirmed against 3.14.6's concurrent.futures.thread source directly:

• >=3.14: _worker(weakref, self._create_worker_context(), work_queue)
• <3.14 (legacy): _worker(weakref, work_queue, initializer, initargs)

Both branches preserve the two behavioral changes that make this pool useful: daemon=True and no _threads_queues registration (so a wedged worker never blocks interpreter exit).

Testing

Added test_many_concurrent_submits_like_tool_executor, reproducing the tool-executor's concurrent-submit shape (more submissions than max_workers, forcing multiple _adjust_thread_count calls).

• Verified it fails with the exact AttributeError above against the pre-fix code on 3.14.6.
• Verified all 5 tests pass against the fix on both:
  • Python 3.11.15 (uv-managed venv — the officially supported interpreter per requires-python)
  • Python 3.14.6 (local dev venv)
• ruff check / ruff format --check clean on both changed files.
• ty check shows only pre-existing baseline diagnostics unrelated to this change (present, and worse, on the unmodified file).

Summary by CodeRabbit

• Bug Fixes

  • Improved reliability of background task execution under heavy concurrent submission loads.
  • Fixed worker startup/thread initialization behavior to align with Python runtime expectations across supported versions.

• Tests

  • Added a regression test covering high-concurrency task submissions to prevent future executor startup failures.
  • Updated a gateway service test to make LaunchAgent domain resolution deterministic in headless runs.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 2a2d3722-1f48-4bae-bc51-4b60a7060fe2

📥 Commits

Reviewing files that changed from the base of the PR and between adc5fd1 and d720e62.

📒 Files selected for processing (1)

• tests/hermes_cli/test_gateway_service.py

---

📝 Walkthrough

Walkthrough

This PR updates daemon thread pool worker spawning to match CPython’s private _worker API across Python versions and adds a regression test for concurrent submissions. It also makes one gateway service restart test use deterministic launchd domain resolution and subprocess call assertions.

Changes

Daemon Pool Worker Compatibility

Layer / File(s)	Summary
Version-aware worker spawning tools/daemon_pool.py	Adds _PY314_WORKER_CONTEXT and _spawn_daemon_worker, conditionally importing and invoking the matching private _worker signature, and routes _adjust_thread_count through the helper.
Concurrent submit regression test tests/tools/test_daemon_pool.py	Adds test_many_concurrent_submits_like_tool_executor, submitting 12 tasks to a 4-worker pool and asserting all futures complete with expected squared results.

Launchd Restart Test Determinism

Layer / File(s)	Summary
Pinned launchd domain and filtered probes tests/hermes_cli/test_gateway_service.py	Forces headless launchd domain computation to user/<uid> and excludes launchctl print probes from recorded subprocess calls so the restart assertions only cover the action sequence.

Estimated code review effort: 4 (Complex) | ~35 minutes

Sequence Diagram(s)

sequenceDiagram
  participant Executor as DaemonThreadPoolExecutor
  participant Adjust as _adjust_thread_count
  participant Spawn as _spawn_daemon_worker
  participant Worker as _worker

  Executor->>Adjust: request a new daemon thread
  Adjust->>Spawn: _spawn_daemon_worker(self, thread_name, weakref_cb)
  Spawn->>Worker: start thread with version-matched arguments

Loading

Possibly related PRs

• sahilm-ti/hermes-agent#48: Also updates test_gateway_service.py to pin launchd domain behavior and adjust launchctl subprocess expectations.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main fix: adding CPython 3.14 WorkerContext support to DaemonThreadPoolExecutor.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/daemon-pool-py314

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}