PrefectHQ/prefect - CodSpeed

prefect

Blog Docs Changelog

Performance History

Latest Results

fix(tests): await asyncio.sleep in test_delete_flow_run_deletes_logs (#21657) Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>

main

30 minutes ago

chore(deps-dev): bump the eslint group across 1 directory with 2 updates (#21677) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: alex.s <alex.s@prefect.io>

main

1 hour ago

Batch-acknowledge Redis stream messages to reduce round-trips (#21622) Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

main

2 hours ago

docs(prefect-dbt): reorganize `prefect-dbt` docs into sibling pages (#21671) Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

main

2 hours ago

docs: document breadcrumb-scoping pattern for strict mode on detail pages (#21694) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

main

2 hours ago

fix(ui-v2): scope task run detail name assertion to breadcrumb in E2E test (#21693) Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> Co-authored-by: API User <apiuser@prefect.io> Co-authored-by: bot_apk <apk@cognition.ai>

main

2 hours ago

stop leaking sys.modules entries in load_script_as_module When the simple, filename-derived module name is already present in `sys.modules` (which happens from the second call onward for any code path that loads the same script repeatedly), `load_script_as_module` fell back to `f"__prefect_loader_{id(path)}__"`. But `id(path)` is the id of the *Python string object* passed in by the caller -- fresh on every call site -- so the fallback key was effectively unique every call, and `sys.modules` grew by one entry per invocation. No popping ever happened. For callers that load flows in a loop (e.g. deploy scripts calling `Flow.afrom_source` N times, each from a fresh tmpdir), this leak pinned every prior flow module's import graph -- pandas, sqlalchemy, internal packages -- in RSS for the life of the process. In constrained environments the accumulated memory can push the container close enough to its cap that subprocess fork / exec -- notably `git clone` during subsequent `afrom_source` calls -- fails with a bare `exit code 1`. Fix: derive the fallback key from a short stable hash of the resolved absolute path. Repeated loads of the same file now reuse the same `sys.modules` slot instead of allocating a new one. Co-Authored-By: Claude Opus 4 (1M context) <noreply@anthropic.com>

nate/fix-sys-modules-leak-load-script

2 hours ago

fix: handle PydanticUserError for deferred forward ref in parameter schema Pydantic 2.13 defers TypeAdapter schema building; unresolved forward references now surface as PydanticUserError when json_schema is called, rather than PydanticUndefinedAnnotation at TypeAdapter construction. Also sync client dependency bounds for packaging and rich with the root pyproject.toml so the prefect-client build test passes. Co-authored-by: alex.s <alex.s@prefect.io> Co-Authored-By: alex.s <ajstreed1@gmail.com>

dependabot/uv/uv-dependencies-f08efdf59b

2 hours ago

Latest Branches

0%

stop leaking `sys.modules` entries in `load_script_as_module`#21692

2 hours ago

a7301dd

nate/fix-sys-modules-leak-load-script

+1%

chore(deps): bump the uv-dependencies group across 1 directory with 25 updates#21676

2 hours ago

fdceaa4

dependabot/uv/uv-dependencies-f08efdf59b

0%

surface git clone stderr when no hint pattern matches#21689

4 hours ago

49f8f0a

nate/surface-git-clone-stderr

© 2026 CodSpeed Technology

Home Terms Privacy Docs