[Dineshsuriya D (droideronline)]

Motivation & Context

GitHubCopilotAgent ignored response_format, so structured output was unavailable even though the Copilot SDK supports it (github/copilot-sdk#2590, in SDK v1.0.15 and later, within the existing >=1.0.16 pin).

Description & Review Guide

• What are the major changes?
  • Added response_format (Pydantic model type or JSON Schema mapping) to GitHubCopilotOptions.
  • The agent forwards it as response_schema to send_and_wait (non-streaming) and send (streaming). Per-run options override default_options.
  • response_format is stripped from the create_session kwargs so it no longer reaches the SDK session factory.
  • AgentResponse.value is parsed from the final text: via response_format= on the non-streaming response and output_format_type= in the streaming finalizer.
• What is the impact of these changes? Opt-in. Without response_format the SDK receives response_schema=None, which is the existing behavior.
• What do you want reviewers to focus on? Streaming still yields text deltas, with the value parsed once the stream completes, matching the SDK behavior.

Related Issue

Fixes #9809

AI Assistance

• No material AI assistance was used.
• This is an AI-assisted contribution. I reviewed, understood, and verified all submitted content and accept responsibility for it.

AI assistance details: implementation and tests.

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 links to an agreed issue with no competing open PR, or the Related Issue section documents a trivial-change or repository-automation exception.
• 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.

[Copilot]

🟡 Changes recommended

Structured-output parsing can incorrectly use trailing subagent text instead of the root answer, while mapping schemas and aliased models can trigger serialization or validation failures.

3 open issues

• Parse the final root response instead of trailing child text · New
• Convert mapping schemas to plain dictionaries before JSON serialization · New
• Parse aliased model output with alias-based validation · New

What changed in this PR

Adds opt-in structured output support to GitHubCopilotAgent, bringing it closer to other Agent Framework providers and addressing #9809.

Changes:

• Accepts Pydantic models or JSON Schema mappings through response_format.
• Forwards schemas to streaming and non-streaming SDK calls, excludes them from session creation, and enables AgentResponse.value parsing.
• Adds tests for schema forwarding, default options, unstructured requests, and streaming results.

File	Description
python/​packages/​github_copilot/​tests/​test_github_copilot_agent.py	Adds structured-output tests for both response modes.
python/​packages/​github_copilot/​agent_framework_github_copilot/​_agent.py	Adds the option, SDK forwarding, and structured response parsing.

_{Review effort: Balanced}

[Dineshsuriya D (droideronline)]

Hi Eduard van Valkenburg (@eavanvalkenburg) , please take a look when you get a chance

[Dineshsuriya D (droideronline)]

Eduard van Valkenburg (@eavanvalkenburg) - the pipeline failure is fixed. Package Checks (3.11) failed because pyright could not resolve the from_updates overload in the streaming finalizer: options.get("response_format") was partially unknown. It is now typed as Any (a14d0ce), and pyright, ruff and the github_copilot tests are clean locally.

Could you please retrigger the pipeline when you get a chance? Thanks!

[Dineshsuriya D (droideronline)]

Eduard van Valkenburg (@eavanvalkenburg) - I moved the alias parsing into core as you suggested (790ac8e), and fixed the typing failure in the new aliased-model tests (3d1e5c8). Could you please retrigger the pipeline when you get a chance?

If everything is green, would it be possible to get this into the next release? We are waiting on structured output for the GitHub Copilot agent. Thanks!

[Eduard van Valkenburg (eavanvalkenburg)]

Dineshsuriya D (@droideronline) Hard-coding by_name=False in shared response parsing overrides models configured with validate_by_name=True or populate_by_name=True. A model with item_count: int = Field(alias="itemCount") previously accepted {"item_count": 3} through AgentResponse.value and ChatResponse.value; this head now raises ValidationError. Please preserve each model’s configured field-name validation while still supporting alias-keyed output, and add a name-keyed regression test.

[Dineshsuriya D (droideronline)]

Eduard van Valkenburg (@eavanvalkenburg) - fixed in 6cdf1e0. Core now validates with the model's own configuration first, so name-keyed output (validate_by_name or populate_by_name) is accepted as before. Only if that fails does it retry with aliases only (pydantic 2.11+) to support alias-keyed provider output, and it re-raises the original error if both fail. Added regression tests for name-keyed output and for the original error on invalid output.