Python: Fix streamed workflow agent continuation context by finalizing AgentExecutor streams

[moonbox3]

Motivation and Context

Fixes a streaming resume bug in AgentExecutor where continuation context could be lost, causing provider errors like:

No tool call found for function call output with call_id ...

The issue is that AgentExecutor._run_agent_streaming iterated streamed updates and rebuilt the final response from collected updates, instead of finalizing the ResponseStream via get_final_response(). That bypassed stream result hooks, which are responsible for persisting thread/conversation state needed for follow-up tool output continuation.

Description

• Updated AgentExecutor streaming flow to call await stream.get_final_response() after iteration.
• Preserved request-info behavior by merging user_input_requests from:
  • streamed updates, and
  • finalized response
    with request-id deduplication.
• Updated streaming thread state update logic to use conversation_id extracted from raw streaming updates (instead of response_id).
• Added regression tests for:
  • streaming thread id uses conversation_id,
  • streaming path does not incorrectly use response_id,
  • AgentExecutor streaming executes stream result hooks by finalizing the stream.
• Handoff checkpoint Sample cleanup:
  • better final conversation output formatting,
  • buffered/normalized streamed text printing,
  • typing fixes for output narrowing.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[Copilot]

Pull request overview

Fixes a streaming resume/continuation bug by ensuring agent streaming paths finalize ResponseStream instances (so result hooks run and state is persisted), and updates streaming thread-id handling to rely on conversation_id rather than response_id.

Changes:

• Update AgentExecutor streaming to call await stream.get_final_response() and merge/dedupe user_input_requests.
• Update streaming thread-id updates to extract and use conversation_id from raw streaming updates.
• Add regression tests for stream finalization hooks and for thread-id behavior in streaming.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
python/packages/core/agent_framework/_workflows/_agent_executor.py	Finalizes agent response streams and dedupes user_input_requests after streaming completes.
python/packages/core/agent_framework/_agents.py	Changes streaming thread-id update logic to extract conversation_id from raw streaming updates.
python/packages/core/tests/workflow/test_agent_executor.py	Adds regression test ensuring AgentExecutor streaming finalizes streams and runs result hooks.
python/packages/core/tests/core/test_agents.py	Adds tests asserting streaming thread-id uses conversation_id and does not fall back to response_id.
python/packages/core/agent_framework/azure/_responses_client.py	Formatting-only changes (no functional behavior change).
python/samples/getting_started/orchestrations/handoff/handoff_with_tool_approval_checkpoint_resume.py	Updates sample to use streaming workflow runs and prints streamed agent output during resume.

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework
_agents.py	312	40	87%	476, 876, 912, 984, 989–992, 1056–1058, 1172, 1185, 1198, 1204, 1240, 1242, 1251–1256, 1262, 1264, 1274–1275, 1282, 1284–1285, 1293–1297, 1305–1306, 1308, 1313, 1315
_mcp.py	406	64	84%	104–105, 115–120, 131, 136, 182–183, 193–198, 208–209, 261, 270, 333, 341, 362, 476, 543, 578, 580, 584–585, 587–588, 642, 657, 675, 716, 822, 835–840, 862, 908–909, 915–917, 936, 961–962, 966–970, 987–991, 1135
packages/core/agent_framework/_workflows
_agent_executor.py	198	26	86%	97, 145, 163–164, 216–217, 219–220, 250–252, 260–262, 270–272, 274, 278, 282, 286–287, 386–387, 433, 451
TOTAL	20444	3249	84%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
3908	225 💤	0 ❌	0 🔥	1m 8s ⏱️