Python: fix(openai): make function result call_id optional and omit when absent

[JHf0912]

Problem

Content.from_function_result still requires call_id: str, although the OpenAI Responses API no longer requires a call id on function call results. Consumers translating between OpenAI-spec SDKs and MAF must fight the type system — the AG-UI tests already pass call_id=None with # type: ignore comments — and _prepare_content_for_openai emits "call_id": null in function_call_output items when the id is absent.

Solution

Relax Content.from_function_result to accept call_id: str | None = None, a backward-compatible widening since every existing caller passes a str. In the Responses client, omit the call_id key when absent instead of sending null, keeping the payload unchanged when a call id is present. As regression evidence, the now-redundant triple-checker type: ignore comment in the AG-UI tests is removed.

Changes

• python/packages/core/agent_framework/_types.py: widen signature + docstring update
• python/packages/openai/agent_framework_openai/_chat_client.py: conditional call_id emission in the function_call_output branch
• Tests: new cases in test_types.py (core) and test_openai_chat_client.py (openai); type-ignore cleanup in test_run_common.py (ag-ui)

Testing

• New unit tests: result without call id → key omitted; result with call id → key preserved (regression)
• Affected suites pass: core test_types.py, openai test_openai_chat_client.py, ag-ui test_run_common.py
• Regression sweep on packages calling from_function_result (anthropic / a2a / bedrock / chatkit): all green
• pyright strict: 0 errors; ruff: clean

Notes for Reviewer

• The function_call (assistant-side) branch that drops content without a call id is intentionally unchanged — the API still requires it there.
• Shell call output branches (shell_call_output / local_shell_call) keep their own id schemes and were not touched.
• The signature change is backward compatible: all existing call sites pass a str.

Fixes #7922

[Copilot]

🟢 Approval recommended

The focused, backward-compatible change correctly implements the updated API contract with appropriate regression coverage.

Pull request overview

Makes function-result call IDs optional and aligns OpenAI Responses payload serialization accordingly.

Changes:

• Allows function results without a call_id.
• Omits absent call IDs from OpenAI payloads while preserving present IDs.
• Adds regression tests and removes obsolete type suppressions.

File summaries

File	Description
python/packages/core/agent_framework/_types.py	Makes call_id optional.
python/packages/core/tests/core/test_types.py	Tests missing call IDs.
python/packages/openai/agent_framework_openai/_chat_client.py	Conditionally emits call_id.
python/packages/openai/tests/openai/test_openai_chat_client.py	Tests payload serialization.
python/packages/ag-ui/tests/ag_ui/test_run_common.py	Removes obsolete type ignores.

Review details

• Files reviewed: 5/5 changed files
• Comments generated: 0
• Review effort level: Balanced

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[JHf0912]

@microsoft-github-policy-service agree

[Eduard van Valkenburg (eavanvalkenburg)]

This is a duplicate of #7928

[JHf0912]

Thanks for the pointer, Eduard van Valkenburg (@eavanvalkenburg). I see it now: #7988 established the Content.id / call_id identity contract, and making anonymous function results generally valid would weaken it — orphan results could leak through invocation, approvals, history, compaction, and non-OpenAI adapters. That rationale makes sense.

I'll close this PR. Since #7922 remains open for a Responses-boundary redesign, I'd be glad to contribute there — e.g., compiling the schema-level matrix of id-less wire combinations and drafting a boundary-scoped design that does not touch core semantics. Happy to coordinate if that would be useful.