Python: Fix workflow not pausing when agent calls declaration-only tool

[moonbox3]

Motivation and Context

• Fixes Python: [Bug]: Workflow doesn't pause when agent calls declaration_only tool, causing OpenAI 400 error #3425: when an agent calls a declaration_only tool (for example, a client-side AG-UI tool with func=None), the workflow now pauses and emits a request_info event instead of continuing and hitting an OpenAI 400 error
• Three changes:
  • _try_execute_function_calls: mark declaration-only FunctionCallContent items with user_input_request=True so AgentExecutor detects them
  • _handle_function_call_results: don't re-add function_call items already present in the response message
  • handle_user_input_response: use role="tool" when feeding function results back to the agent so the LLM sees proper tool responses

Description

• Closes Python: [Bug]: Workflow doesn't pause when agent calls declaration_only tool, causing OpenAI 400 error #3425

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.

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework
_tools.py	797	85	89%	232, 278, 329, 331, 359, 529, 564–565, 667, 669, 689, 707, 721, 733, 738, 740, 747, 780, 851–853, 894, 916–944, 979, 987, 1228, 1433, 1490, 1494, 1573–1577, 1595, 1597–1598, 1710, 1714, 1764, 1766, 1782, 1784, 1848, 1875, 1932, 2000, 2179–2180, 2207, 2215, 2228, 2238–2239, 2274, 2330, 2362
packages/core/agent_framework/_workflows
_agent_executor.py	167	22	86%	98, 146, 164–165, 219–220, 222–223, 255–257, 265–267, 277–279, 281, 285, 289, 293–294
TOTAL	16555	2057	87%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
3923	225 💤	0 ❌	0 🔥	1m 10s ⏱️

[Copilot]

Pull request overview

This PR fixes a workflow pause/resume gap for declaration-only tools (tools defined with func=None, commonly representing client-side tools). It ensures the workflow pauses and emits request_info when such a tool is called, and that resuming the workflow feeds back a proper tool response message to avoid OpenAI “orphaned tool_calls” 400 errors.

Changes:

• Mark declaration-only function_call contents as user_input_request (and assign id) so AgentExecutor pauses and emits request_info.
• Avoid duplicating function_call items into the assistant message when handling function-call results.
• When resuming from user responses, send back function results with role="tool" so OpenAI receives valid tool responses.
• Adds unit tests and a new sample demonstrating the behavior.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
python/packages/core/agent_framework/_tools.py	Marks declaration-only tool calls as user input requests; avoids re-adding function_call contents into the assistant message.
python/packages/core/agent_framework/_workflows/_agent_executor.py	Uses role="tool" when resuming with function परिणाम (function_result) contents.
python/packages/core/tests/workflow/test_agent_executor_tool_calls.py	Adds non-streaming, streaming, and parallel tests for declaration-only tool pause/resume behavior.
python/samples/getting_started/workflows/human-in-the-loop/agents_with_declaration_only_tools.py	New sample demonstrating declaration-only tool calls pausing the workflow and resuming via responses={...}.
python/samples/getting_started/workflows/README.md	Links the new sample in the workflows README table.