Python: fix: set session_context._response before _run_after_providers (#4248)

[LEDazzio01]

Summary\n\nFixes #4248 — WorkflowAgent was not saving workflow responses to session history.\n\n## Root Cause\n\nBaseHistoryProvider.after_run() reads output messages from context.response:\n\npython\nif self.store_outputs and context.response and context.response.messages:\n messages_to_store.extend(context.response.messages)\n\n\nThe regular Agent sets session_context._response = AgentResponse(...) before calling _run_after_providers, but WorkflowAgent._run_impl and _run_stream_impl called _run_after_providers without setting _response. The history provider found context.response == None and saved no output messages.\n\n## Changes\n\n### _agent.py\n\nNon-streaming (_run_impl) — 1 line added:\npython\nresult = self._convert_workflow_events_to_agent_response(response_id, output_events)\nsession_context._response = result # <-- NEW\nawait self._run_after_providers(session=provider_session, context=session_context)\nreturn result\n\n\nStreaming (_run_stream_impl) — collect updates and build response:\npython\ncollected_updates: list[AgentResponseUpdate] = [] # <-- NEW\nasync for event in self._run_core(...):\n updates = self._convert_workflow_event_to_agent_response_updates(response_id, event)\n for update in updates:\n collected_updates.append(update) # <-- NEW\n yield update\nif collected_updates: # <-- NEW\n session_context._response = AgentResponse.from_updates(collected_updates) # <-- NEW\nawait self._run_after_providers(session=provider_session, context=session_context)\n\n\n### Tests\n\nNew test file test_workflow_agent_session_history.py with 3 test cases:\n1. Non-streaming: Verifies session.state[\"in_memory\"][\"messages\"] contains both user and assistant messages\n2. Streaming: Same for the streaming path\n3. Multi-turn: Two successive runs accumulate all 4 messages (2 user + 2 assistant)

[Copilot]

Pull request overview

This PR fixes issue #4248 where WorkflowAgent was not saving workflow responses to session history. The root cause was that session_context._response was never set before calling _run_after_providers(), which meant the BaseHistoryProvider.after_run() method found context.response == None and only saved user input messages, skipping the assistant response messages.

Changes:

• Set session_context._response before _run_after_providers in both streaming and non-streaming execution paths
• Added comprehensive test coverage for non-streaming, streaming, and multi-turn scenarios

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/_agent.py	Sets session_context._response before calling _run_after_providers in both _run_impl (non-streaming) and _run_stream_impl (streaming) to ensure history providers can persist output messages
python/packages/core/tests/workflow/test_workflow_agent_session_history.py	Adds comprehensive tests validating that WorkflowAgent correctly persists both user and assistant messages to session history in non-streaming, streaming, and multi-turn scenarios