YOLO

1.5.1.5 ⚡ Quick Ask Overhaul & Refined Chat Feedback

✨ Quick Ask & Inline Editing

• 🛠️ Unified Editing Engine: Migrated Quick Ask to a more stable, unified editing engine. This improves the reliability of selection rewrites and structured content editing.
• 📉 Enhanced Review Diffs: Selection rewrite reviews now support Markdown block-level diffs. We've also addressed issues with title grouping, scrolling jitters, and content mismatches during complex changes.
• 🎨 Selection Highlighting: Added a new "Selection Highlight Background Color" for both Sidebar Chat and Quick Ask (configurable in settings #165), making it easier to identify the active context.
• 🧠 Logic Optimization: Refined selection prompts and the injection logic for skills and tools within the Quick Ask workflow.
• 🐞 Session Fix: Fixed a bug where Quick Ask would incorrectly inherit session states from the sidebar.

💬 Chat Experience & Rendering

• 🌊 Smoother Streaming: Optimized the streaming rendering and scrolling strategies, resulting in improved fluidity and reduced performance overhead during AI response generation.
• 🎭 Animation & Style: Refined the visual design, status feedback, and transition animations (#166) across the sidebar to provide a more polished feel.
• 🔄 History Accuracy: When reopening historical conversations, the expanded message editor now correctly displays the model that was actually used for that specific turn.
• 📋 LaTeX Support: You can now copy LaTeX math formulas directly from the chat interface (#171).

🔧 Core Engine & Bug Fixes

• 🤖 DeepSeek Fix (#87): Resolved a 400 error occurring in DeepSeek "Reasoning Mode" when consecutive tool calls failed to pass back the reasoning_content.
• ⚡ Optimized Apply Engine: Merged the "Fast" and "Precise" apply modes into a single, more efficient middleware for a more consistent code application experience.
• 👀 Better Previews: Improved the streaming feedback during code replacements. This reduces the exposure of raw JSON and provides a cleaner preview of changes, including better handling of "Aborted" states.
• 🧹 Maintenance: Normalized the default assistant logic to prevent unnecessary updateAt timestamp refreshes (#174).

1.5.1.5 ⚡ Quick Ask 体验升级与交互细节优化

✨ Quick Ask 与内联编辑

• 🛠️ 统一编辑引擎：Quick Ask 现已改用更稳定的统一编辑引擎，显著提升了选区改写与结构化内容编辑的可靠性。
• 📉 审阅体验改进：选段改写的 Diff 对比现支持按 Markdown 段块展示。同时修复了标题分组显示、滚动抖动以及复杂变更时的内容错配问题。
• 🎨 选区背景高亮：新增侧边栏 Chat 与 Quick Ask 的「选区高亮背景色」能力，支持在设置中自主开启或关闭 (#165)。
• 🧠 逻辑优化：优化了 Quick Ask 的选区提示词算法以及 Skills 与工具的注入逻辑。
• 🐞 会话隔离修复：修复了 Quick Ask 助手选择异常错误继承侧边栏会话状态的问题。

💬 Chat 交互与性能优化

• 🌊 流式渲染增强：优化了聊天流式渲染与滚动策略，在回复生成时拥有更好的流畅度，并降低了浏览器的性能消耗。
• 🎭 动画与样式：进一步打磨了侧边栏对话的样式设计、状态反馈及过渡动画 (#166)，提升视觉一致性。
• 🔄 历史记录模型显示：重新打开历史对话时，展开的消息编辑栏现在能够正确回显该轮对话当时所使用的模型。
• 📋 公式复制支持：支持在对话栏直接复制 Latex 数学公式 (#171)。

🔧 核心引擎与 Bug 修复

• 🤖 DeepSeek 适配修复 (#87)：修复了 DeepSeek 思考模式下，因连续工具调用未回传 reasoning_content 导致的 API 400 报错。
• ⚡ Apply 引擎重构：优化了对话 Apply 的实现方式，将原有的快速与精准应用模式合并为统一的过渡处理件。
• 👀 预览体验优化：改进了聊天应用修改时的流式反馈，减少了原始 JSON 的暴露，并优化了内容替换预览及中止生成时的前端过渡效果。
• 🧹 助手逻辑归一化：修改了默认助手的归一化逻辑，避免了无实际变动时频繁刷新 updateAt 时间戳的问题 (#174)。