Python: Fix duplicate arguments in declaration-only tool streaming

[kartikmadan11]

Motivation & Context

When streaming declaration-only function calls, the final response contains duplicated arguments.

Fixes #6973

Description & Review Guide

What are the major changes?

• Modified _process_function_requests() in _tools.py (line 2403) to filter out function_call type items from the streamed result update
• Added two test cases covering chunked and single-chunk streaming scenarios

What is the impact?

• Eliminates argument duplication for declaration-only streaming tool calls
• Preserves correct behavior for function_result and approval items

Root Cause: The _stream() loop yields function_call chunks as they arrive, then later re-yields those same items from function_call_results. The _process_update() callback concatenates arguments with +=, so duplicate yields cause duplication.

Solution: Filter out function_call type items from the result's function_call_results before yielding, since they were already emitted by the inner stream.

Tests Added

• test_declaration_only_tool_streaming_no_argument_duplication() - Verifies multi-chunk streaming
• test_declaration_only_tool_streaming_single_chunk_no_duplication() - Verifies single-chunk streaming

Both tests pass ✅

[Copilot]

Pull request overview

Fixes a Python streaming bug in the core function-invocation layer where declaration-only tool calls could end up with duplicated arguments in the final aggregated response due to the same function_call content being yielded twice during streaming.

Changes:

• Filters function_call items out of the post-processing streamed function_call_results update to prevent _process_update from concatenating the same arguments twice.
• Adds regression tests for both multi-chunk and single-chunk streaming declaration-only tool calls.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
python/packages/core/agent_framework/_tools.py	Adjusts streaming post-processing to avoid re-emitting function_call items that were already emitted by the inner stream.
python/packages/core/tests/core/test_function_invocation_logic.py	Adds tests ensuring declaration-only tool streaming does not duplicate function-call arguments.

[kartikmadan11]

Great catch, Copilot! I've addressed the concern about empty updates.

The fix:
When the filtered function_call_results list is empty (after excluding already-streamed function_call items), we now clear update_role to suppress the update. This prevents emitting confusing empty ChatResponseUpdate events.

Implementation:

• Filter out function_call items: filtered_results = [r for r in function_call_results if r.type != "function_call"]
• If the filtered list is empty, set result["update_role"] = None
• The existing yield check if role := result.get("update_role") will skip the empty update

This way declaration-only tools in streaming mode won't emit spurious empty updates—only the already-streamed chunks and any results/approvals will be emitted.

[eavanvalkenburg]

we should use the get_final_response method on the stream object, to test this with

[kartikmadan11]

Done, switched the test to use stream.get_final_response() and assert on the aggregated response. Thanks!

[github-actions]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework
_tools.py	1170	83	92%	228–229, 406, 408, 421, 446–448, 456, 474, 488, 495, 502, 525, 527, 534, 542, 671, 705–707, 710–712, 714, 720, 771–773, 798, 824, 828, 866–868, 872, 894, 1045, 1057, 1064–1067, 1088, 1096, 1110–1112, 1462, 1554, 1582, 1604, 1612, 1670, 1717, 1724–1725, 1784, 1788, 1841, 1902–1903, 1914, 1986, 2000, 2003, 2016, 2019, 2042, 2049, 2059, 2063, 2140, 2193, 2215, 2271, 2350, 2560, 2626–2627, 2793–2794, 2885
_types.py	1226	96	92%	61, 70–71, 125, 130, 149, 151, 155, 159, 161, 163, 165, 183, 187, 213, 235, 240, 245, 249, 279, 705–706, 887–888, 1346, 1421, 1456, 1476, 1486, 1676–1678, 1963–1968, 1993, 2048, 2053, 2063, 2071, 2078–2082, 2100, 2173, 2186, 2191, 2304, 2327, 2588, 2612, 2711, 2892–2893, 2995, 3224, 3277, 3296, 3335, 3346, 3348–3352, 3354, 3357–3365, 3375, 3464, 3601, 3606, 3611, 3616, 3620, 3706–3708, 3737, 3825–3829
TOTAL	44464	5235	88%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
8986	33 💤	0 ❌	0 🔥	2m 23s ⏱️

[eavanvalkenburg]

@kartikmadan11 there are both open comments and failed checks, please have a look!

[kartikmadan11]

Both handled. Fixed the test per the review comment, which also clears the typing check failure. Rebased onto latest main so the merge gate should clear too.