fix(workflow): preserve event.output for cached LlmAgent results

The runner stripped event.output to None before persisting any
non-partial event with node_info.message_as_output=True. On resume,
_reconstruct_node_state fell back to event.content (a raw
genai.types.Content) for the cached child output, so subscripting the
result of ctx.run_node on an LlmAgent across a HITL pause crashed with
"'Content' object is not subscriptable". Removing the strip lets the
validated dict produced by process_llm_agent_output survive the
persist/rehydrate roundtrip.

Closes #5553

Author: ValentinCordonnier

src/google/adk/runners.py

@@ -611,10 +611,6 @@ async def _consume_event_queue(
       if event_or_done is done_sentinel:
         break
       event: Event = event_or_done
-      if not event.partial:
-        if event.node_info.message_as_output and event.content is not None:
-          event = event.model_copy()
-          event.output = None
 
       _apply_run_config_custom_metadata(event, ic.run_config)
       modified_event = await ic.plugin_manager.run_on_event_callback(

tests/unittests/workflow/test_workflow_dynamic_nodes.py

@@ -1153,8 +1153,9 @@ async def parent(*, ctx, node_input):
   assert len(agent_events) > 0
   agent_event = agent_events[-1]
 
-  # Verify that runners.py cleared the output
-  assert agent_event.output is None
+  # The persisted event keeps the LlmAgent's output so resume can replay
+  # it without re-firing the model
+  assert agent_event.output == 'LLM output content'
 
   # When the workflow is resumed by resolving the interrupt
   resume_events = await _resume(

tests/unittests/workflow/test_workflow_hitl.py

@@ -2051,3 +2051,52 @@ def node_b():
   )
   outputs3 = workflow_testing_utils.get_outputs(events3)
   assert outputs3 == [f'c_from_b_response 2']
+
+
+@pytest.mark.asyncio
+async def test_cached_llm_agent_output_remains_subscriptable_after_resume(
+    request: pytest.FixtureRequest,
+):
+  """Cached LlmAgent output must stay a dict on resume."""
+
+  class Greeting(BaseModel):
+    text: str
+
+  greeter = LlmAgent(
+      name='greeter',
+      model=testing_utils.MockModel.create(
+          responses=[types.Part.from_text(text='{"text": "hi"}')],
+      ),
+      mode='single_turn',
+      output_schema=Greeting,
+      instruction='Return JSON.',
+  )
+
+  @node(rerun_on_resume=True)
+  async def orchestrator(ctx: Context, node_input: Any):
+    out = await ctx.run_node(greeter, 'go')
+    if 'confirm' not in ctx.resume_inputs:
+      yield RequestInput(message='ok?', interrupt_id='confirm')
+      return
+    yield Event(output={'echoed': out['text']})
+
+  agent = Workflow(name='wf', edges=[(START, orchestrator)])
+  app = App(
+      name=request.function.__name__,
+      root_agent=agent,
+      resumability_config=ResumabilityConfig(is_resumable=True),
+  )
+  runner = testing_utils.InMemoryRunner(app=app)
+
+  events1 = await runner.run_async(testing_utils.get_user_content('start'))
+  req = workflow_testing_utils.get_request_input_events(events1)[0]
+  interrupt_id = get_request_input_interrupt_ids(req)[0]
+  invocation_id = events1[0].invocation_id
+
+  events2 = await runner.run_async(
+      new_message=testing_utils.UserContent(
+          create_request_input_response(interrupt_id, {'approved': True})
+      ),
+      invocation_id=invocation_id,
+  )
+  assert {'echoed': 'hi'} in workflow_testing_utils.get_outputs(events2)