1.5.1.4
1.5.1.4 🚀 Performance Boost & Fluid Editing
⚡ Performance Optimizations
- 🚫 Non-blocking Sending: Removed the blocking token-counting and RAG-splitting logic before sending messages. This reduces lag when
@mentioning multiple files (#149). - 📸 Lightweight Session Storage: Refactored the conversation storage to a "Message + Snapshot Reference" mechanism. This ensures lower memory usage and better response speeds during long-running conversations or when working with large documents.
- 🧹 Snapshot Management: Added a "Clear Snapshots" option in settings to help maintain a clean workspace.
- 🛡️ Interaction Stability: Optimized the performance of selected mention text, addressing potential page deadlocks when interacting with mentions (#125).
✍️ Refined Editing & Review Experience
- 🪄 Inline Review: Refactored the "Selection Rewrite" module into an inline review experience within the editor. AI suggestions now flow more naturally with your writing process.
- 📉 Paragraph-level Diffs: The Apply view now utilizes paragraph-level diffing, making it easier to track and review granular changes in natural language sections.
- 🔢 LaTeX & Format Robustness: Improved parsing for Obsidian-specific formatting. Fixed a bug where Agent mode would mis-parse LaTeX backslash commands (
\) during precise replacements (#158). - 📦 Better Rendering: Enhanced
<smtcmp_block>rendering to prevent occasional blockquote nesting failures.
🧠 Skills & Agent Intelligence
- 🤖 Anthropic Skills Compatibility: The Skills module now supports the Anthropic standard (e.g.,
YOLO/skills/**/SKILL.md). It features automatic ID fallback and stricter file identification to prevent accidental document loading (#162). - 📂 YOLO Root Directory: You can now configure a custom YOLO plugin root directory, centralizing the discovery and loading path for all your skills (#136).
- 🔍 Smart
@Filtering: The@reference menu now supports instant filtering for files, folders, and assistants as you type, with a cleaner layout and improved placeholder hints.
1.5.1.4 🚀 性能优化与创作体验改进
⚡ 性能改进
- 🚫 发送流程优化：移除了发送前统计 Token 与文件超限 RAG 分流的阻塞逻辑。降低了在
@引用多个文件时点击发送后的界面卡顿现象 (#149)。 - 📸 轻量化会话存储：将会话存储重构为“消息内容 + 快照引用”机制。在处理长程对话或大文档时，能提供更低的内存占用与响应速度提升。
- 🧹 快照管理：在设置中新增“清空快照”功能，支持手动清理历史缓存以释放空间。
- 🛡️ 交互稳定性：优化了选中 Mention 文本时的性能表现，解决了在高频交互下可能导致的页面死锁（卡死）问题 (#125)。
✍️ 创作与审阅体验重构
- 🪄 内联审阅模式：重构了“选中改写”模块，将选段改写升级为编辑器内联审阅。AI 的改写建议现在直接在文内呈现，创作流更加连贯。
- 📉 段级 Diff 对比：Apply 视图现已统一采用段级差异对比（Diff），在处理自然语言段落时视觉上更清晰、更易读。
- 🔢 LaTeX 与格式增强：优化了针对 Obsidian 格式解析的鲁棒性。修复了 Agent 在执行精确替换时对 Latex 反斜杠命令（
\）的解析错误 (#158)。 - 📦 渲染稳定性：优化了 Obsidian 渲染渲染逻辑，修复了在复杂嵌套输出场景下引用块偶尔无法正常渲染的问题。
🧠 Skills 与 Agent 智能