Python: Make encrypted reasoning opt-in for Foundry chat - #7536

Pull request overview

This PR adjusts the Python Foundry chat client so encrypted reasoning (reasoning.encrypted_content) is no longer implicitly requested, avoiding 400 failures on Foundry deployments/models that don’t support it, while still allowing callers to explicitly opt in.

Changes:

• Override FoundryChatClient option preparation to strip framework-injected reasoning.encrypted_content unless the caller explicitly requested it.
• Add regression tests covering default behavior (no encrypted reasoning) and explicit opt-in preservation.
• Document the Foundry encrypted-reasoning opt-in rule in the Python function-calling loop specification.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

Agent Framework Skill Review

Completed passes: 5 | Failed passes: 0 | Confidence: 59% | Result: No high-severity findings

Review passes

• Correctness (gpt-5.6-sol) — No high-confidence issues found in this pass.
• Security Reliability (claude-opus-4.8) — No high-confidence issues found in this pass.
• Test Coverage (gpt-5.6-sol) — The new tests do not cover preservation of unrelated include values after filtering, so a regression that removes all caller-suplied includes could pass unnoticed.
• Failure Modes (claude-opus-4.8) — No high-confidence issues found in this pass.
• Design Approach (claude-opus-4.8) — No high-confidence issues found in this pass.

No high-severity findings survived the review and verification pipeline.

PR ensemble review by moonbox3's agents

Agent Framework Review — Iteration 2

Completed passes: 5 | Failed passes: 0 | Result: No high-severity findings

Scope: 1 net-new commit(s): 71e71dce562d

Review passes

• Correctness (gpt-5.6-sol) — No high-confidence issues found in this pass.
• Security Reliability (claude-opus-4.8) — No high-confidence issues found in this pass.
• Test Coverage (gpt-5.6-sol) — No high-confidence issues found in this pass.
• Failure Modes (claude-opus-4.8) — No high-confidence issues found in this pass.
• Design Approach (claude-opus-4.8) — No high-confidence issues found in this pass.

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

    @override
    async def _prepare_options(
        self,
        messages: Sequence[Message],
        options: Mapping[str, Any],
    ) -> dict[str, Any]:
        """Prepare Foundry options without implicitly requesting encrypted reasoning."""
        caller_requested_encrypted_reasoning = "reasoning.encrypted_content" in (options.get("include") or [])
        run_options = await super()._prepare_options(messages, options)
        if not caller_requested_encrypted_reasoning and isinstance(run_options.get("include"), list):
            include = [item for item in run_options["include"] if item != "reasoning.encrypted_content"]
            if include:
                run_options["include"] = include
            else:
                run_options.pop("include")
        return run_options

    async def create_response(**kwargs: Any) -> Any:
        if "reasoning.encrypted_content" in kwargs.get("include", []):
            raise ValueError("Encrypted content is not supported with this model.")
        return _as_raw(mock_response)

            default_options={  # pyrefly: ignore[bad-argument-type]
                "store": False,
                "reasoning": {"effort": "low", "summary": "auto"},
                "include": ["reasoning.encrypted_content"],
            },
        )
        server = ResponsesHostServer(agent, store=InMemoryResponseProvider())