Python: Preserve workflow run kwargs when continuing with run(responses=...)

[moonbox3]

Motivation and Context

When a workflow pauses (e.g., for human approval) and is resumed via run(responses=...), the original run_kwargs passed in the initial call were silently dropped and replaced with an empty dict. This caused executors/agents to lose access to custom data like auth tokens or configuration passed at run time.

Fixes #4293

Description

The root cause was that _state.set(WORKFLOW_RUN_KWARGS_KEY, run_kwargs or {}) unconditionally overwrote the stored kwargs on every run() call, including continuation calls where run_kwargs is None. The fix conditionally updates stored kwargs: they are only overwritten when new kwargs are explicitly provided, or when the context is being reset (fresh run). On continuation calls with no new kwargs (reset_context=False), the previously stored kwargs are preserved. Two regression tests verify both preservation and explicit override behavior.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

---

Note: PR autogenerated by moonbox3's agent

[moonbox3]

Automated Code Review

Reviewers: 3 | Confidence: 85%

✓ Correctness

The change correctly fixes the kwargs-overwrite-on-continuation bug (#4293) by guarding the state write: kwargs are only written when explicitly provided or when the context is freshly reset. The logic is sound for the intended workflow lifecycle (initial run → pause → continue). The tests cover preservation and override scenarios well. One minor edge-case worth noting: if run(responses=..., reset_context=False) is ever called without a preceding run, WORKFLOW_RUN_KWARGS_KEY will never be set, whereas the old code always ensured the key existed. This is unlikely in practice since reset_context=False implies a continuation, but the old comment explicitly called out deterministic retrieval.

✓ Security Reliability

This diff fixes a bug where workflow run kwargs were unintentionally overwritten to an empty dict on continuation calls (reset_context=False). The new conditional logic correctly preserves kwargs when no new ones are provided during continuation, and only resets them on a fresh run. The change is minimal, well-scoped, and accompanied by thorough regression tests. No security or reliability issues are introduced.

✓ Test Coverage

The two new tests directly cover the primary behavioral change (preserving kwargs on continuation and overriding them when explicitly provided). Assertions are meaningful and include a diagnostic message on the most important check. However, an edge case is missing: passing an explicit empty dict (run_kwargs={}) on continuation should overwrite the preserved kwargs with {} (since {} is not None), which is a subtle but important distinction from passing None. There is also no dedicated test for the reset_context=True with run_kwargs=None branch (line 355), though existing tests for fresh runs likely exercise it indirectly. Overall the coverage is solid for the primary scenarios.

Suggestions

• Consider whether downstream consumers of WORKFLOW_RUN_KWARGS_KEY use a safe access pattern (e.g., state.get with a default) rather than assuming the key always exists. The old code guaranteed the key was always set; the new code may leave it unset on the (unlikely) path where reset_context=False and run_kwargs=None with no prior run.
• Consider whether the state should be defensively initialized with an empty dict if WORKFLOW_RUN_KWARGS_KEY is missing during a continuation (reset_context=False, run_kwargs=None) — e.g., if a workflow is resumed from persisted state that predates this key. A missing key at retrieval time could surface as an unexpected KeyError downstream.
• Add a test for the continuation case where run_kwargs is explicitly an empty dict {} (not None). This exercises a subtle boundary: {} passes the is not None check and should clear previously-stored kwargs, whereas omitting kwargs (None) should preserve them.
• Consider adding a targeted unit test that calls workflow.run(...) with reset_context=True and no kwargs to explicitly verify the elif reset_context branch (line 355) stores {}.

---

Automated review by moonbox3's agents

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework/_workflows
_agent_executor.py	200	26	87%	97, 113, 168–169, 221–222, 224–225, 255–257, 265–267, 275–277, 279, 283, 287, 291–292, 391–392, 438, 456
_workflow.py	267	19	92%	87, 268–270, 272–273, 291, 295, 430, 618, 639, 695, 707, 713, 718, 738–740, 753
_workflow_executor.py	178	30	83%	94, 444, 467, 469, 477–478, 483, 485, 490, 492, 545, 573–579, 583–585, 593, 598, 609, 619, 623, 629, 633, 643, 647
TOTAL	22177	2762	87%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
4676	247 💤	0 ❌	0 🔥	1m 17s ⏱️

[Copilot]

Pull request overview

This PR fixes a regression where workflow run kwargs were incorrectly cleared when continuing a paused workflow via run(responses=...). The issue affected executors/agents that rely on custom data (e.g., auth tokens, configuration) passed at runtime.

Changes:

• Modified kwargs storage logic to preserve existing kwargs on continuation calls unless explicitly overridden
• Added two comprehensive regression tests validating both preservation and override behaviors

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
python/packages/core/agent_framework/_workflows/_workflow.py	Replaced unconditional kwargs overwrite with conditional logic: preserve on continuation (reset_context=False) unless new kwargs explicitly provided
python/packages/core/tests/workflow/test_workflow_kwargs.py	Added test_kwargs_preserved_on_response_continuation and test_kwargs_overridden_on_response_continuation to verify fix