feat: 增加渠道级 LLM 思考模式开关

[H-Chris233]

User description

概要

Closes #402

本 PR 在 LLM 设置里补齐“思考模式”开关，并把开关贯通到所有 LLM 调用路径：润色、翻译、QA 流式回答和设置页连接验证。

实现方式

• 新增 llmThinkingEnabled 用户偏好，并通过前端设置页切换保存。
• OpenAI-compatible 路径统一在请求体中按“渠道”下发官方思考字段，不再维护单模型适配表：
  • OpenAI / Coding Plan X：reasoning_effort
  • OpenRouter：reasoning.effort
  • Alibaba Coding：enable_thinking
  • DeepSeek：thinking.type
• Gemini 原生路径继续走 generateContent / streamGenerateContent，用原生 thinkingConfig 做渠道级控制。
• 新增的 Codex OAuth provider 走 Responses 风格端口，因此同样按 Responses 请求体使用 reasoning.effort。
• 明确撤销 prompt 注入方案：不向 system/user prompt 追加“开启/关闭思考”的文字。

文档核对

• OpenAI Responses / reasoning: https://platform.openai.com/docs/guides/reasoning
• OpenAI API reference: https://platform.openai.com/docs/api-reference/responses/create
• OpenRouter reasoning tokens: https://openrouter.ai/docs/use-cases/reasoning-tokens
• DeepSeek thinking mode: https://api-docs.deepseek.com/guides/thinking_mode
• Alibaba Model Studio deep thinking: https://help.aliyun.com/zh/model-studio/deep-thinking
• Gemini thinking: https://ai.google.dev/gemini-api/docs/thinking

验证

• cargo test --manifest-path openless-all/app/src-tauri/Cargo.toml --lib openai_chat_body -- --nocapture
• cd openless-all/app && cargo test --manifest-path src-tauri/Cargo.toml --lib codex_oauth -- --nocapture
• cd openless-all/app && cargo test --manifest-path src-tauri/Cargo.toml --lib disabled_thinking_config -- --nocapture
• cd openless-all/app && cargo test --manifest-path src-tauri/Cargo.toml --lib build_generate_body -- --nocapture
• cd openless-all/app && cargo check --manifest-path src-tauri/Cargo.toml
• cd openless-all/app && npm run build
• git diff --check

备注：Rust 与 Vite 输出均只剩仓库既有 warning。

---

PR Type

Enhancement, Tests

---

Description

• Add channel-level LLM thinking toggle in settings

• Pass thinking flag to polish, translate, QA, and validation

• Inject provider‑specific thinking fields (OpenAI, DeepSeek, Alibaba, Gemini, OpenRouter, Codex)

• Simplify Gemini thinking to channel‑level budget config, dropping per‑model table

---

Diagram Walkthrough

flowchart LR
  Settings["Settings: Toggle thinking mode"] -- "saves" --> Prefs["UserPreferences.llm_thinking_enabled"]
  Prefs -- "passed to" --> Dispatch["Polish / Translate / QA / Validate"]
  Dispatch -- "openai/codingPlanX" --> OpenAI["OpenAI‑compatible providers"]
  Dispatch -- "gemini" --> Gemini["Gemini native"]
  Dispatch -- "codex" --> Codex["Codex OAuth"]
  OpenAI -- "applies" --> Fields["reasoning_effort / enable_thinking / reasoning.effort / thinking.type"]
  Gemini -- "applies" --> GemConfig["thinkingConfig: budget=0 (off) or omitted (on)"]
  Codex -- "applies" --> CodexField["reasoning.effort: medium/low"]

Loading

File Walkthrough

	Relevant files
Enhancement	8 files commands.rs Read thinking preference and pass to provider configs        +21/-9    coordinator.rs Thread thinking flag through polish/translate/QA paths      +39/-18  dictation.rs Pass thinking preference to dictation session handling      +3/-0      llm_gemini.rs Simplify Gemini thinking control to channel‑level budget  +50/-115 polish.rs Add channel‑specific thinking control fields for OpenAI‑compatible +208/-20 types.rs Add llm_thinking_enabled field to UserPreferences                +10/-0    types.ts Add llmThinkingEnabled to frontend UserPreferences type    +2/-0      Settings.tsx Add thinking toggle UI component in settings                          +45/-5
Documentation	5 files en.ts Add English translations for thinking mode toggle                +4/-0      ja.ts Add Japanese translations for thinking mode                            +4/-0      ko.ts Add Korean translations for thinking mode                                +4/-0      zh-CN.ts Add Chinese translations for thinking mode                              +4/-0      zh-TW.ts Add Traditional Chinese translations                                          +4/-0
Tests	2 files ipc.ts Add llmThinkingEnabled to mock settings                                    +1/-0      stylePrefs.test.ts Include thinking field in test preferences                              +1/-0

---

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis ✅ 402 - PR Code Verified Compliant requirements: 已在 LLM 设置页新增思考模式开关 已将开关状态持久化到用户偏好 已贯通到润色、翻译、QA 流式回答和连接验证等调用路径 Requires further human verification: 需要在实际 UI 中确认开关布局、文案和交互是否符合预期

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ No major issues detected