fix(provider): exclude chat models from textVerbosity setting

[R44VC0RP]

Summary

Fixes #9969 - GPT-5.x chat models (e.g. gpt-5.2-chat-latest) were incorrectly getting textVerbosity: "low" set, but these models only support medium verbosity, causing API errors.

Root Cause

The condition at transform.ts:592 checked for gpt-5-chat (dash) but missed dot-versioned chat models like gpt-5.2-chat-latest:

// Before: only excluded "gpt-5-chat", missed "gpt-5.2-chat-latest"
if (input.model.api.id.includes("gpt-5") && !input.model.api.id.includes("gpt-5-chat")) {

Fix

Added -chat suffix check to exclude all chat model variants from the textVerbosity: "low" setting:

if (
  input.model.api.id.includes("gpt-5.") &&
  !input.model.api.id.includes("codex") &&
  !input.model.api.id.includes("-chat") &&  // <-- Added
  input.model.providerID !== "azure"
) {
  result["textVerbosity"] = "low"
}

Testing

Added 7 new test cases covering:

• gpt-5.2 and gpt-5.1 → should have textVerbosity: "low"
• gpt-5.2-chat-latest, gpt-5.1-chat-latest, gpt-5.2-chat, gpt-5-chat → should NOT have textVerbosity set
• gpt-5.2-codex → should NOT have textVerbosity set (already excluded)

All 88 transform tests pass.

[github-actions]

The following comment was made by an LLM, it may be inaccurate: