Python: Preserve declaration-only streaming metadata

[eavanvalkenburg]

Motivation & Context

Streaming declaration-only calls currently replay the finalized arguments after the provider already streamed
them, causing the final response to concatenate the arguments twice.

Description & Review Guide

• What are the major changes?
  • Emit a metadata-only finalized declaration update with arguments=None.
  • Preserve id and user_input_request while function-call chunks are aggregated.
  • Cover split-argument and single-chunk provider streams.
  • Preserve workflow request_info signaling for client-side tools.
  • Perform the final shared spec reconciliation for Python: [Bug]: function_approval_response is dropped from the request under service-side storage, so an approval-paused run never resumes #7125, Python: [Bug]: Function-invocation-limit orphan: MAF streams a tool call (TOOL_CALL_START/ARGS/END) and then discards it, leaving an AG-UI card stuck "Running" with no TOOL_CALL_RESULT #7045, and Python: [Bug]: Duplicate arguments in declaration-only function call when streamed #6973 after their unique scenario rows land.
• What is the impact of these changes?
  • Declaration-only arguments appear exactly once in the final response.
  • Streaming consumers still receive the pause/request metadata.
  • Executable and informational function calls are unaffected.
  • The three focused PRs no longer compete over shared spec gaps or related-issue list hunks.
• What do you want reviewers to focus on?
  • Metadata propagation through Content addition.
  • Raw stream versus finalized response parity.
  • Merge order: Python: Preserve approval decisions under OpenAI continuation #7407 and Python: Remove tool content returned after invocation limits #7408 should merge before this final spec-reconciliation PR.

Related Issue

Fixes #6973

Depends on #7407 and #7408; merge this PR last.

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR is linked to an issue and there is no other open PR for this issue (see Related Issue above).
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[github-actions]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework
_tools.py	1375	83	93%	230–231, 408, 410, 423, 448–450, 458, 476, 490, 497, 504, 527, 529, 536, 544, 679, 713–715, 718–720, 722, 728, 779–781, 806, 832, 836, 874–876, 880, 1053, 1065, 1072–1075, 1096, 1104, 1118–1120, 1490, 1573, 1601, 1623, 1675–1676, 1733, 1780, 1787–1788, 1878, 1940–1941, 1971, 2067, 2081, 2084, 2097, 2100, 2123, 2130, 2139, 2143, 2168, 2202, 2270, 2299–2300, 2397, 2425, 2465, 2468, 2525, 2678, 2767, 3256
_types.py	1232	96	92%	61, 70–71, 125, 134, 153, 155, 159, 163, 165, 167, 169, 187, 191, 217, 239, 244, 249, 253, 283, 709–710, 891–892, 1350, 1425, 1460, 1480, 1490, 1684–1686, 1971–1976, 2001, 2056, 2061, 2071, 2079, 2086–2090, 2108, 2181, 2194, 2199, 2312, 2335, 2596, 2620, 2719, 2900–2901, 3003, 3232, 3285, 3304, 3343, 3354, 3356–3360, 3362, 3365–3373, 3383, 3472, 3609, 3614, 3619, 3624, 3628, 3714–3716, 3745, 3833–3837
TOTAL	46024	4478	90%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
9513	34 💤	0 ❌	0 🔥	2m 30s ⏱️

[Copilot]

Pull request overview

Fixes a Python streaming edge case where declaration-only tool calls could end up with duplicated arguments in the finalized response by ensuring the function-invocation layer emits metadata-only updates while preserving control metadata during content aggregation.

Changes:

• Emit a metadata-only finalized declaration update (arguments=None) during streaming so provider-streamed argument chunks aren’t replayed/duplicated.
• Preserve id and user_input_request when aggregating function_call Content instances.
• Add regression tests for split-argument vs single-chunk streams, and document the behavior in the Python agent conventions/spec.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
python/packages/core/agent_framework/_tools.py	Emits metadata-only function_call streaming updates to avoid replaying streamed arguments.
python/packages/core/agent_framework/_types.py	Propagates id and user_input_request through function_call content aggregation.
python/packages/core/tests/core/test_function_invocation_logic.py	Adds streaming regression coverage for declaration-only calls (split and single-chunk arguments).
python/packages/core/tests/core/test_types.py	Adds unit coverage ensuring metadata-only follow-ups preserve control metadata without altering arguments.
python/packages/core/AGENTS.md	Documents the declaration-only streaming behavior and metadata-only follow-up semantics.
docs/specs/004-python-function-calling-loop.md	Updates spec/test matrix to reflect the new covered regression scenario for #6973.