BuiltInCodeExecutor emits an empty final event before every model reply #7878

🔴 Required Information

Describe the Bug:
With code_executor=BuiltInCodeExecutor(), every model reply is preceded by an empty event. That event has no content and no actions, but is_final_response() returns True for it, and it's stored in the session.

The built-in branch of the code execution response processor (flows/llm_flows/extensions/_code_execution.py) always yields an Event(actions=event_actions). It does so even when the response had no image to save as an artifact, in which case event_actions is empty.

is_final_response() documents that "Application and UI layers can rely on this helper to detect a complete, user-facing response". So a run loop that stops at the first final response returns nothing. This is what #4138 ran into: "only get a empty data event when running in my custom runner". The reporter closed it by no longer stopping at is_final_response(), but the empty event is still emitted on main.

Steps to Reproduce:

1. Create an LlmAgent with code_executor=BuiltInCodeExecutor().
2. Have the model answer with plain text, so there's no image.
3. Iterate runner.run_async(...) and print each event's content and is_final_response().

Expected Behavior:
Only the model's reply is emitted. An extra event appears only when it carries an artifact_delta for a saved image.

Observed Behavior:

'a' content= None actions= {'state_delta': {}} final= True
'a' content= parts=[Part(text='The answer is 42.')] ... final= True
session events: 3 [('user', True), ('a', False), ('a', True)]

Environment Details:

• ADK Library Version (pip show google-adk): 2.13.0 (also on main at e9e8f61)
• Desktop OS: macOS
• Python Version (python -V): 3.11.16

Model Information:

• Are you using LiteLLM: No
• Which model is being used: Gemini 2.x / 3 (BuiltInCodeExecutor requires a Gemini model); the repro uses a mock BaseLlm

🟡 Optional Information

Minimal Reproduction Code:

agent = LlmAgent(name="a", model=model, code_executor=BuiltInCodeExecutor())
runner = InMemoryRunner(agent=agent)
session = await runner.session_service.create_session(app_name=runner.app_name, user_id="u")

async for event in runner.run_async(
    user_id="u", session_id=session.id,
    new_message=types.Content(role="user", parts=[types.Part(text="What is 6 * 7?")])):
  print(repr(event.author), "content=", event.content, "final=", event.is_final_response())
  if event.is_final_response():
    break  # stops on the empty event, before the answer

How often has this issue occurred?:

• Always (100%), for every reply that contains no image