PrefectHQ/prefect - CodSpeed

prefect

Blog Docs Changelog

Performance History

Latest Results

fix(deps): raise pydantic-settings lower bound to >=2.9.1 prefect imports pydantic_settings.exceptions.SettingsError which was added in pydantic-settings 2.9.0, but the declared lower bound allowed versions as old as 2.2.1. This caused ModuleNotFoundError at import time when an older pydantic-settings was installed. Raise the floor to >=2.9.1 (2.9.0 was already excluded) in both pyproject.toml and client/pyproject.toml. closes #22074 Co-authored-by: Alexander Streed <alex.s@prefect.io>

devin/1779649663-fix-pydantic-settings-min-version

3 hours ago

style: fix ruff-format trailing comma in test_base_worker Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

vincere-mori:fix-cancel-run-zombie-infrastructure

8 hours ago

fix(orchestration): guard against ZeroDivisionError in RetryFailedTasks when retry_jitter_factor is set without retry_delay `RetryFailedTasks.after_transition` passes `base_delay` directly to `clamped_poisson_interval` when `retry_jitter_factor` is configured, but `clamped_poisson_interval(0, ...)` divides by zero in its exponential CDF computation, crashing the orchestration server. This happens whenever a task uses `@task(retries=N, retry_jitter_factor=X)` without specifying `retry_delay_seconds` (the default). In that case `run_settings.retry_delay` is `None`, which normalises to `base_delay = 0`, and the subsequent call to `clamped_poisson_interval` raises `ZeroDivisionError`. Fix: skip jitter and use `base_delay` directly when `base_delay == 0`. Jitter on a zero-second delay is meaningless anyway. A regression test `test_jitter_with_zero_delay_does_not_raise` is added to `TestRetryFailedTasks` to cover this path.

devteamaegis:fix/retry-jitter-zero-delay-division-by-zero

20 hours ago

chore(deps): bump qs from 6.14.2 to 6.15.2 in /ui (#22070)

main

1 day ago

docs: Add release notes for 3.7.2 (#22069) Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com> Co-authored-by: Alex Streed <alex.s@prefect.io> Co-authored-by: alex.s <ajstreed1@gmail.com> Co-authored-by: Alex Streed <desertaxle@users.noreply.github.com>

main

2 days ago

fix: worker-channel ready frame returns `worker_id=None` (#22067)

main

2 days ago

fix: OSS worker-channel ready frame returns worker_id=None The new OSS WebSocket channel was returning worker.id in the worker.ready.v1 frame, causing workers to set backend_id and PREFECT__WORKER_ID unnecessarily. This matches the REST heartbeat behavior where OSS does not return a worker ID to the client. - Set worker_id to None in the OSS worker.ready.v1 payload - Keep internal worker heartbeat recording unchanged - Add tests verifying OSS ready frame has worker_id=None - Add tests verifying PREFECT__WORKER_ID is not set via channel path - Add protocol test confirming on_worker_id callback is not invoked Co-authored-by: Alex Streed <alex.s@prefect.io> Co-Authored-By: alex.s <ajstreed1@gmail.com>

devin/1779484624-oss-worker-channel-no-worker-id

2 days ago

Preserve cleanup idempotency after ack

alexs/oss-7938-oss-03-oss-worker-communication-queue-interface-and-in

2 days ago

Latest Branches

+1%

fix(deps): raise pydantic-settings lower bound to >=2.9.1#22075

4 hours ago

a7416c2

devin/1779649663-fix-pydantic-settings-min-version

+1%

Fix _cancel_run to kill infrastructure for already-running flow runs#22072

8 hours ago

e7d769f

vincere-mori:fix-cancel-run-zombie-infrastructure

+1%

fix(orchestration): guard ZeroDivisionError in RetryFailedTasks when retry_jitter_factor set without retry_delay#22071

20 hours ago

be49290

devteamaegis:fix/retry-jitter-zero-delay-division-by-zero

© 2026 CodSpeed Technology

Home Terms Privacy Docs