ReasoningEffort.None and ReasoningEffort.ExtraHigh were not mapping to their correct OpenAI API values. None was falling through to the wildcard default and mapping to null (indistinguishable from "not specified"), and ExtraHigh was mapping to High. Since ChatReasoningEffortLevel and ResponseReasoningEffortLevel are string-based extensible enum types, we construct new instances with the appropriate string values for cases where no static property is exposed.

Changes

• OpenAIChatClient.cs: Map ReasoningEffort.None → new ChatReasoningEffortLevel("none"), ReasoningEffort.ExtraHigh → new ChatReasoningEffortLevel("xhigh")
• OpenAIResponsesChatClient.cs: Map ReasoningEffort.None → new ResponseReasoningEffortLevel("none"), ReasoningEffort.ExtraHigh → new ResponseReasoningEffortLevel("xhigh")
• Tests: Added None/"none" and ExtraHigh/"xhigh" as [InlineData] on the existing parameterized theory tests

// Before: None fell through to _ => null, ExtraHigh mapped to High
// After:
effort switch
{
    ReasoningEffort.None => new ChatReasoningEffortLevel("none"),
    ReasoningEffort.Low => ChatReasoningEffortLevel.Low,
    ReasoningEffort.Medium => ChatReasoningEffortLevel.Medium,
    ReasoningEffort.High => ChatReasoningEffortLevel.High,
    ReasoningEffort.ExtraHigh => new ChatReasoningEffortLevel("xhigh"),
    _ => (ChatReasoningEffortLevel?)null,
};

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

Microsoft Reviewers: Open in CodeFlow