[Wilson0511]

Summary

acleanup_provider_files(..., delete_all_from_provider=True) called uploader.list_files() — the sync method that hits the provider's HTTP API — directly on the event loop. Deletes were already off-loop (adelete under a semaphore + gather), but the listing step froze every other coroutine for the duration of the provider listing. The listing now runs via asyncio.to_thread, a one-line fix with no uploader API change (alist_files does not exist on the uploaders).

Closes #8641

Change

• lib/crewai-files/src/crewai_files/cache/cleanup.py: files = await asyncio.to_thread(uploader.list_files) in the delete_all_from_provider branch of acleanup_provider_files (the sync cleanup_provider_files twin is unchanged).
• lib/crewai-files/tests/test_acleanup_listing_nonblocking.py: regression test with a stub uploader whose list_files sleeps 0.5 s; a ticker coroutine must keep ticking while the listing runs. Fails on main (ticker frozen), passes with the fix.

Proof

• Red on main: AssertionError: event loop was blocked during listing: 3 ticks (expected ≥ 6).
• Green with the fix: 1 passed; full lib/crewai-files suite: 180 passed, 4 skipped.

Risk

Low: behavior is identical except the listing executes in a worker thread; result ordering and error handling (except Exception path) are unchanged.

🤖 Generated with ZCode (AI agent) on behalf of Wilson0511

Co-Authored-By: ZCode [email protected]

---

Note

Low Risk
Same cleanup semantics and error handling; only moves synchronous listing to a worker thread so the event loop stays responsive.

Overview
Fixes event-loop blocking when acleanup_provider_files runs with delete_all_from_provider=True. Provider file listing used synchronous list_files() (HTTP) on the asyncio loop while deletes were already async; other coroutines stalled for the whole listing call.

Listing now runs with await asyncio.to_thread(uploader.list_files) so the sync uploader API stays unchanged. The sync cleanup_provider_files path is untouched.

Adds a regression test with a stub uploader whose list_files sleeps 0.5s and a ticker task that must keep firing during cleanup—fails if listing blocks the loop.

^{Reviewed by Cursor Bugbot for commit 34ecc95. Bugbot is set up for automated code reviews on this repo. Configure here.}

[strix-security]

Strix Security Review

Warning

This pull request has 2 commits after the last Strix review (34ecc95). Strix has not reviewed these changes.
Automatic review on push is off for this repository. To review the latest changes, tag @strix-security in a comment, or turn on re-review on push.

✅ No open security findings.

PR summary: Moves the synchronous provider file listing in `acleanup_provider_files(delete_all_from_provider=True)` off the event loop via `asyncio.to_thread`, and adds a test that fails if the listing blocks the loop.

Review summary

The diff wraps the synchronous provider listing in acleanup_provider_files with asyncio.to_thread, plus a regression test asserting the event loop keeps ticking during the listing. No security issues found: the listing is read-only in every uploader, uploaders are instantiated per call so the worker thread has exclusive access, error handling is unchanged, and no privilege, credential, or input-handling path is touched.

_{Review 1 on this PR · 34ecc95 · requested Oct 27 04:38 UTC when the PR was opened}

---

Reviewed by Strix
_{Re-run review · Configure security review settings}

[Vidit-Ostwal]

lgtm

[Vidit-Ostwal]

Reviewed the event-loop-blocking cleanup path and its regression coverage. Moving the synchronous provider listing to asyncio.to_thread is the minimal correct fix, and the ticker-based test verifies that other coroutines keep running.