fix(logstore): normalize workflow timestamps to naive UTC

[amogh-nagri-11]

Summary

Fixes #40943.

The two LogStore repositories deserialize workflow timestamps with calls that resolve against the host's local clock, while the rest of the codebase stores naive UTC:

• datetime.fromtimestamp(ts) without tz → naive local time
• datetime.now() → naive local time
• datetime.fromisoformat(s) → aware when the string carries an offset, naive otherwise

On its own that would just be a uniform offset. The real problem is that created_at and finished_at are matched independently, so one record's two timestamps can be produced by different branches and land in different frames — and _dict_to_workflow_run then subtracts them for elapsed_time.

Observed on a UTC+05:30 host (reproduction in the issue drives the real functions):

Payload	Before	After
both epoch	timestamps shifted +5:30	correct naive UTC
aware ISO + epoch	TypeError: can't subtract offset-naive and offset-aware datetimes	elapsed_time = 30.0
naive ISO + epoch	30-second run reported as 19830 s	elapsed_time = 30.0
started_at missing	elapsed_time negative (-49875.0)	defaults to naive_utc_now()

On a UTC host all four coincidentally agree, which is why this is easy to miss.

Changes

Every branch of both match statements now terminates in naive UTC, using the helpers that already exist in libs/datetime_utils:

• datetime.now() → naive_utc_now() (the same call the Postgres write path uses)
• datetime.fromtimestamp(ts) → datetime.fromtimestamp(ts, tz=UTC).replace(tzinfo=None)
• datetime.fromisoformat(s) → wrapped in ensure_naive_utc(...), which normalizes aware values and is a no-op for naive ones
• the already-a-datetime passthrough branch is wrapped in ensure_naive_utc too, so an aware value from upstream can't reintroduce the mixed-frame subtraction

Applied to all six call sites across logstore_api_workflow_run_repository.py and logstore_api_workflow_node_execution_repository.py. One change fixes the offset shift, the TypeError, and both bad elapsed_time results together.

Scope note: this path is only reachable when the Aliyun SLS LogStore backend is enabled, so deployments on the default Postgres path are unaffected. No behavior change on UTC hosts.

Tests

Added regression coverage that pins TZ=Asia/Kolkata via time.tzset(), so the assertions are meaningful on the UTC CI host rather than passing by coincidence. It covers all four payload shapes plus the missing-timestamp default, for both repositories.

All 10 new tests fail on main and pass with this change:

uv run --project api --dev pytest api/tests/unit_tests/extensions/logstore/repositories/
13 passed

ruff check, ruff format --check, mypy and pyrefly are clean on the changed files (the two pre-existing bad-instantiation pyrefly diagnostics in the node-execution test file are on untouched lines).

Screenshots

Not applicable — backend-only timezone correctness fix.

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. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly.
• I ran make lint && make type-check (backend) and vp staged (frontend) to appease the lint gods

[github-actions]

Pyrefly Diff

base → PR

--- /tmp/pyrefly_base.txt	2026-08-19 03:30:39.634723894 +0000
+++ /tmp/pyrefly_pr.txt	2026-08-19 03:30:26.357618702 +0000
@@ -7975,9 +7975,9 @@
 ERROR Object of class `Events` has no attribute `request_start` [missing-attribute]
   --> tests/unit_tests/events/test_events_package_compat.py:33:5
 ERROR Cannot instantiate `LogstoreAPIWorkflowNodeExecutionRepository` because the following members are abstract: `get_execution_snapshots_by_workflow_run`, `delete_expired_executions`, `count_by_runs`, `delete_by_runs`, `delete_executions_by_app`, `get_expired_executions_batch`, `delete_executions_by_ids`, `get_offloads_by_execution_ids`, `save`, `save_synchronously`, `save_execution_data`, `get_by_workflow_execution` [bad-instantiation]
-  --> tests/unit_tests/extensions/logstore/repositories/test_logstore_api_workflow_node_execution_repository.py:12:64
+  --> tests/unit_tests/extensions/logstore/repositories/test_logstore_api_workflow_node_execution_repository.py:18:64
 ERROR Cannot instantiate `LogstoreAPIWorkflowNodeExecutionRepository` because the following members are abstract: `get_execution_snapshots_by_workflow_run`, `delete_expired_executions`, `count_by_runs`, `delete_by_runs`, `delete_executions_by_app`, `get_expired_executions_batch`, `delete_executions_by_ids`, `get_offloads_by_execution_ids`, `save`, `save_synchronously`, `save_execution_data`, `get_by_workflow_execution` [bad-instantiation]
-  --> tests/unit_tests/extensions/logstore/repositories/test_logstore_api_workflow_node_execution_repository.py:21:64
+  --> tests/unit_tests/extensions/logstore/repositories/test_logstore_api_workflow_node_execution_repository.py:27:64
 ERROR Cannot index into `object` [bad-index]
    --> tests/unit_tests/extensions/otel/test_celery_sqlcommenter.py:136:20
 ERROR Cannot index into `object` [bad-index]

[github-actions]

Pyrefly Type Coverage

Metric	Base	PR	Delta
Type coverage	59.69%	59.70%	+0.01%
Strict coverage	59.28%	59.28%	+0.01%
Typed symbols	39,917	39,929	+12
Untyped symbols	27,142	27,142	0
Modules	3181	3182	+1

[amogh-nagri-11]

Thank you @asukaminato0721 for the quick review and merge!