Shangjie Chen fd006db915 fix(workflow): fix task agent resumption in nested workflows ...

Fix an issue where task-mode LLM agents invoked inside nested workflows or custom nodes with raise_on_wait=True fail to resume upon user reply.

- Ensure raise_on_wait=True in Context._run_node_internal triggers NodeInterruptedError when child output is None regardless of wait_for_output defaults.
- Update ReplayInterceptor Case 5 to allow re-running Workflow instances and rerun_on_resume nodes when recovered output is None.
- Add unit test test_nested_workflow_with_task_agent to cover task agent resumption in nested workflows.

Co-authored-by: Shangjie Chen <deanchen@google.com>
PiperOrigin-RevId: 949367240

2026-07-16 22:27:24 -07:00

..

benchmarks

perf: avoid deepcopy of session contents when building LLM requests

2026-06-30 14:36:32 -07:00

integration

chore(agents): tag ManagedAgent traffic with a +managed_agent version suffix

2026-07-13 14:19:30 -07:00

remote/triggers

chore: migrate from gemini-1.* and gemini-2.0* to gemini-2.5-*

2026-04-29 10:34:22 -07:00

unittests

fix(workflow): fix task agent resumption in nested workflows

2026-07-16 22:27:24 -07:00

__init__.py

…