[Yufeng He (he-yufeng)]

Fixes #9500

Root cause

MagenticOrchestrator.__init__ calls super().__init__, whose contract documents that a max_rounds below 1 "will be coerced to 1" (max(1, max_rounds) in _base_group_chat_orchestrator.py) — but then overwrites self._max_rounds with the manager's raw max_round_count. With max_round_count=0 the coercion is lost and the first limits check (self._round_index >= self._max_rounds, i.e. 0 >= 0) fires before any coordination round: the run emits "Workflow terminated due to reaching maximum round count." and no participant is ever invoked.

Fix

Apply the same clamp when adopting the manager's value:

self._max_rounds = max(1, manager.max_round_count) if manager.max_round_count is not None else None

One line, matching the base contract and GroupChatOrchestrator's treatment of the same input. None (unlimited) still passes through.

Test plan

New regression test test_magentic_zero_max_round_count_is_coerced_to_one (next to the existing round-limit tests) builds a workflow with FakeManager(max_round_count=0) and asserts the run reaches the manager's final answer — the exact expectation the existing max_round_count=1 test uses, so the test pins "0 behaves as 1". On the unpatched code it fails with the issue's verbatim symptom (Workflow terminated due to reaching maximum round count.).

Verified locally (orchestrations + core from this branch on the repo venv):

python -m pytest packages/orchestrations/tests/          # 298 passed, 6 skipped
python -m pyright packages/orchestrations/.../_magentic.py  # no diagnostics in the changed hunk
ruff check / ruff format --check                            # clean

[Copilot]

🟢 Approval recommended

The focused fix matches the documented contract, and the regression test covers the original premature termination.

No open issues 🎉

What changed in this PR

Fixes #9500 by aligning Magentic’s round-limit handling with the base orchestrator contract.

Changes:

• Clamps max_round_count below 1 to 1 while preserving None as unlimited.
• Adds a regression test confirming a zero limit reaches the manager’s final answer.

File	Description
python/​packages/​orchestrations/​tests/​test_magentic.py	Tests zero-limit coercion.
python/​packages/​orchestrations/​agent_framework_orchestrations/​_magentic.py	Clamps the manager’s round limit.

_{Review effort: Balanced}

[Eduard van Valkenburg (eavanvalkenburg)]

The Magentic change itself is focused and matches the documented base-orchestrator contract, but this PR currently also includes the unrelated memory-state quarantine commit (68bc218) and Redis context-provider cleanup commit (a4f739b). Please rebase onto current main and retain only the Magentic fix (385c3ad, plus any conflict-only adjustments) so the PR returns to the intended two-file scope, then rerun CI.

[Eduard van Valkenburg (eavanvalkenburg)]

Closing in favor of #9502, which predates this PR, links the same issue, and contains the same focused round-limit fix on a clean branch. Thanks for the alternative regression coverage.