[GYJ99]

Summary

• Check emitted telemetry events, counters, deduplication, payload rehydration, and degraded behavior through EnterpriseMetricHandler.handle().
• Assert meaningful mail recipients/template data/log content and real Celery queue routing instead of incidental internal call counts or task attributes.
• Validate generated OpenAPI schemas and operation contracts, with accepted/rejected request payloads. Batch schema-root validation and inspect descendants once, preserving required-field, reference, and discriminator checks.
• Add four nested-schema negative cases. Give the aggregate five-document OpenAPI check a bounded 60-second timeout under coverage; ordinary tests retain the global CI timeout.
• Sync upstream main at 8d812b938af6, including the stable xdist parameter IDs merged in fix(api): pass lint, type check and unit tests on the new dev dependencies #45308.

Refs #41418. This is a selected four-module slice of the broader cleanup.

Screenshots

N/A (test-only change).

Validation

Executed locally on macOS / Python 3.13.11 against the tree committed as 67710432275c8d3e1fb42664bacdd8b65998eae4:

• uv run --project api --python 3.13 --frozen --dev pytest api/tests/unit_tests/tasks/test_generate_summary_index_task.py api/tests/unit_tests/tasks/test_mail_inner_task.py api/tests/unit_tests/enterprise/telemetry/test_metric_handler.py api/tests/unit_tests/dev/test_generate_swagger_specs.py --timeout 20 --durations=5 -q — 110 passed in 189.23s. The aggregate OpenAPI contract test took 13.01s in this run.
• UV_PYTHON=3.13 UV_FROZEN=1 make lint — passed; no unrelated file changes; 82 import contracts kept.
• UV_PYTHON=3.13 UV_FROZEN=1 make type-check — passed.
• Ruff checks and all original staged commit hooks passed. The hooks used UV_PYTHON=3.13 UV_NO_SYNC=1 to reuse the environment already synchronized from the committed lockfile.
• git diff --check upstream/main...HEAD — clean; the PR still changes only its original four test files.

Before the follow-up, the aggregate OpenAPI check reproducibly exceeded the CI 20-second limit locally and in upstream CI on head 6f49055e45 (that unit shard otherwise passed 12,726 tests). The follow-up removes repeated recursive validation and budgets the remaining real work explicitly. The console export alone is about 4.4 MB with 2,113 component schemas; the test checks all five exports. Nested malformed-schema, missing-required-property, dangling-reference, and discriminator failures remain covered.

Upstream checks on 6771043227 have completed: 35 successful, 14 skipped, no failed or pending checks (verified Nov 3, 2026), including API unit/controller/integration jobs, Python style, and Codecov. GitHub reports MERGEABLE; maintainer review is still required.

The full repository test suite and Docker-backed integration suites were not run locally for this test-only PR. No production code, generated contracts, dependency, or lockfile changes are included.

Checklist

• This change requires a documentation update, included: Dify Document
• I understand that this PR may be closed in case there was no previous discussion or issues.
• I've verified the change and added or updated tests where meaningful regression risk justifies coverage.
• I've updated the documentation accordingly. (N/A: test-only change.)
• I ran make lint && make type-check (backend), with the environment recorded above. vp staged is N/A (no frontend changes).

From Codex

[github-actions]

Pyrefly Type Coverage

Metric	Base	PR	Delta
Type coverage	83.00%	83.00%	-0.00%
Strict coverage	82.66%	82.65%	-0.00%
Typed symbols	85,596	85,582	-14
Untyped symbols	17,603	17,605	+2
Modules	4373	4373	0

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.77%. Comparing base (8d812b9) to head (6771043).
⚠️ Report is 11 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #45298      +/-   ##
==========================================
- Coverage   89.77%   89.77%   -0.01%     
==========================================
  Files        6239     6239              
  Lines      390924   390924              
  Branches    79291    79291              
==========================================
- Hits       350950   350946       -4     
- Misses      33875    33877       +2     
- Partials     6099     6101       +2     

Flag	Coverage Δ
api	89.63% <ø> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[GYJ99]

CI triage for the current head (8921fbd): the Python Style job (including the previously failing No New Cast Guard) now passes. The remaining Main CI failure is in API Unit Tests (3.13, shard 1/2) during pytest-xdist collection, before test execution:

Different tests were collected between gw1 and gw0.
-test_valid_credentials_do_not_acquire_a_lock[1793599135]
+test_valid_credentials_do_not_acquire_a_lock[1793599134]

This comes from the wall-clock-derived parametrization in api/tests/unit_tests/services/credentials/test_tool_runtime.py, which is outside this PR's diff. There is already a targeted upstream fix, #45243 (explicit stable parameter IDs), with green workflow runs. See the failed CI job.

No unrelated test change is needed in this PR. I could not rerun the upstream job with my fork's Actions permissions; a maintainer rerun (or a fresh run after #45243 merges) should verify this head once the collection flake is addressed. This assessment is from CI logs and source inspection; I have not claimed a local Dify test run.

Update (Nov 2, 2026): Upstream #45308 has since merged and its change to test_tool_runtime.py supplies stable parametrization IDs on current main, independently addressing the same xdist collection flake (#45243 remains open but now overlaps). The recorded failed run for this PR predates that upstream change; it has not been rerun, so this is not a claim of green CI. A maintainer rerun or fresh check against the updated base is still needed.