Deck

Release Notes v1.4.5

TL;DR

• 剪贴板捕获进入 pboard/XPC 事件驱动时代
  接入 Apple pboard daemon 的私有 invalidation push 链路，复制发生时由 invalidate-cache / invalidate-entries 直接唤醒 Deck 捕获流水线；常驻高频轮询退场，只保留稀疏健康检查与自动回退。
  Clipboard capture now rides the private pboard/XPC invalidation path: invalidate-cache / invalidate-entries wake Deck immediately, while polling becomes a sparse health check and safety fallback.

• SQLite 搜索性能炸裂治理
  新增 Deck SQLite Benchmark 基准体系后，锁定 FTS5 搜索里 bm25 + timestamp 全量临时排序这一颗大雷；搜索链路改为 rowid DESC 有界候选 + 主表过滤排序，medium 25,000 条数据下最坏多条件搜索从 43.1s 降到 154.6ms，约 278.6x，timeout 从 5/5 清零，整轮 benchmark 从 51.7s 降到 8.7s。
  A new Deck SQLite Benchmark suite exposed the FTS5 bm25 + timestamp full-sort trap; search now uses bounded rowid DESC candidates plus table-side filtering/sorting. On the 25k-row medium benchmark, the worst multi-condition search dropped from 43.1s to 154.6ms (278.6x) with 5/5 timeouts eliminated, and the full run fell from 51.7s to 8.7s.

• DeckClip 连接安全与信道性能
  强化防重放与校验顺序，并提升大数据帧下的读取效率。
  Stronger replay protection and verification ordering, plus larger socket reads for big frames.

• Deck MCP 进程诊断与清理
  deckclip mcp serve 增加初始化超时保护，mcp doctor 可列出当前 MCP bridge 进程、父进程与运行时长，并新增 mcp cleanup 清理疑似空闲实例。
  deckclip mcp serve now exits if no MCP initialize request arrives, while mcp doctor reports running bridge processes and mcp cleanup can clear likely idle instances.

• 复制操作反馈更快
  在按下复制/剪切后的短窗口内更快察觉剪贴板变化，音效与状态栏反馈不再拖延；Deck 面板内 Cmd+C 成功复制时也会立即播放同一套反馈音效。
  Faster copy/cut detection with near-immediate sound and menu-bar feedback; successful Cmd+C copies inside the Deck panel now play the same feedback sound immediately.

• 思考模式对话更稳定
  正确累积、保存并回传 reasoning_content，避免多轮与工具调用时的接口错误。
  Reliable reasoning_content persistence and playback across turns and tool calls.

• IDE Source Anchor 支持 Sublime Text / Zed
  从 Sublime Text session 与 Zed 剪贴板 zed-metadata 捕获文件和行号，复制源码后可从 Deck 直接跳回对应位置。
  IDE Source Anchor now covers Sublime Text and Zed by reading Sublime session state and Zed zed-metadata, so copied code can jump back to the original file and line.

• 代码预览高亮更接近编辑器
  代码预览改用轻量单遍 token 扫描与缓存，高亮字符串、注释、关键字、函数、类型、属性、数字等常见语义，同时减少多轮正则扫描开销。
  Code preview now uses lightweight single-pass token scanning with caching, highlighting strings, comments, keywords, functions, types, properties, numbers, and other common semantics, while reducing the overhead of multiple rounds of regex scanning.

• 面板与长文本预览更流畅
  基于 CPU.trace / UI.trace 的热点优化，减少主线程菜单构建、Markdown 检测与长文本系统探测开销。
  Trace-driven hot-path work reduces main-thread menu construction, Markdown detection,

Release Notes v1.4.4

优化 / Improvements

• 新增 LaTeX 公式渲染：横板模式的 Markdown 卡片预览与 Markdown/纯文本预览面板会按需渲染行内、块级、常见 equation/align/matrix/cases 环境以及 TeX 源中的公式块；Markdown 的「拷贝纯文本」会把公式导出为可读 Unicode 文本，同时保留行内 code 与代码块原文不被渲染；公式渲染结果增加内存缓存，减少列表卡片与预览切换时的重复计算，并保留预览面板原有滚动逻辑。

• Added LaTeX math rendering: horizontal Markdown card previews plus Markdown/plain-text preview panels now render inline, display, common equation/align/matrix/cases environments, and formula blocks inside TeX source; Markdown “Copy Plain Text” now exports formulas as readable Unicode text while preserving inline code and code blocks verbatim; rendered formulas are cached in memory to reduce repeated work during card and preview switches, preserving the existing preview scrolling behavior.

• 优化自动检查更新的请求调度：保留原有固定检查时段不变，在客户端为每台设备生成稳定随机延迟，将集中整点请求打散到 30 分钟窗口内，平滑后端并发峰值，降低更新服务在高峰时段的错误率。

• Improved automatic update-check scheduling by keeping the existing check windows while assigning each device a stable randomized delay, spreading requests across a 30-minute window to smooth backend concurrency spikes and reduce peak-hour update service errors.

• 新增卡片列表内的问卷反馈入口：当历史记录超过 99 条且处于默认全部列表时，自动在最前方展示不可聚焦、不可拖拽、不会进入搜索或标签过滤结果的反馈卡片；支持打开问卷、本次隐藏与永久隐藏，并针对横版/竖版、深色/浅色模式完成视觉适配。

• Added an in-list survey feedback entry: once history exceeds 99 items in the default All view, Deck shows a non-focusable, non-draggable feedback card at the front of the list that never appears in search or tag-filtered results; it supports opening the survey, hiding for the session, or hiding permanently, with tuned horizontal/vertical and dark/light appearances.

• 优化设置页「统计」页面的打开与刷新性能：统计数据上屏时避免不必要的隐式动画抖动，并在渲染前预热常用应用图标缓存，降低主线程卡顿；页面布局、图表样式与交互保持不变。

• Improved the Settings > Statistics page responsiveness by applying statistics updates without unnecessary implicit animation and warming the top-app icon cache before rendering, reducing main-thread stalls while preserving the existing UI, charts, and interactions.

• 优化后台数据库维护扫描：文件搜索文本回填、向量索引回填与缺失文件扫描从 OFFSET 分页改为基于 id 的 keyset 分批推进，大数据量下维护耗时更稳定，并避免越往后扫描越慢；不改变前台 UI、搜索语义或清理规则。

• Optimized background database maintenance scans by switching file-search backfill, vector-index backfill, and missing-file checks from OFFSET pagination to id-based keyset batching, keeping large-history maintenance time stable without changing the UI, search semantics, or cleanup behavior.

• 优化 SQLite 读写调度：新增独立只读连接与 reader 队列，让历史列表、搜索、统计和轻量读取在 WAL 模式下不再排队等待写入、迁移、checkpoint 或后台维护任务；写入与维护仍保留在 writer 串行队列中，reader 不可用时自动回退，保证行为稳定。

• Improved SQLite read/write scheduling with a dedicated read-only connection and reader queue, allowing history lists, search, statistics, and lightweight reads to avoid waiting behind writes, migrations, checkpoints, or background maintenance in WAL mode; writes and maintenance remain serialized on the writer queue, with automatic fallback if the reader is unavailable.

• 优化剪贴板解析与数据库写入热路径：剪贴板监听阶段只在主线程读取 NSPasteboard 快照，富文本、缩略图与类型解析移到后台执行；主表插入复用缓存的 SQLite UPSERT statement 并直接绑定 BLOB，减少重复 prepare 和大 payload 拷贝；后台语义 embedding 入库不再绕回 MainActor。

• Improved clipboard parsing and database write hot paths: clipboard monitoring now reads only an NSPasteboard snapshot on the main thread while rich-text, thumbnail, and type parsing run in the background; main-table inserts reuse a cached SQLite UPSERT statement with direct BLOB binding to reduce repeated prepare work and large-payload copies; background semantic embedding writes no longer hop back through the MainActor.

• 优化语义搜索与 sqlite-vec 索引路径：sqlite-vec 向量写入、恢复与查询改用原生 Float32 BLOB，避免 JSON 字符串格式化与解析；向量命中后的回表改为按候选 id 批量读取并由语义距离恢复顺序，不再引入历史时间排序；安全模式下继续禁用向量索引以保持加密语义不变。

• Optimized semantic search and sqlite-vec index

Release Notes v1.4.3

优化 / Improvements

• SQLite 数据库 P0 写放大与安全模式维护优化 Deck 对剪贴板历史数据库做了一轮最小侵入的 P0 优化：FTS 全文索引现在只会在 search_text、来源 App 名称、自定义标题或加密状态真正变化时更新，给历史记录打标签、切换临时状态等轻量操作不再触发全文索引删除与重建，减少 SQLite WAL 写放大、分词开销和后台 checkpoint 压力。安全模式下，Deck 不再维护无法直接搜索的密文 FTS 索引，全文索引重建也会跳过加密记录；批量插入路径去掉了逐条 SELECT count 的额外查询，降低导入、同步和恢复时的数据库负载。与此同时，安全模式开启/关闭时的语义向量缓存迁移改为静默探测加密状态，不再把预期中的明文探测失败误报为 authenticationFailure 错误刷屏。
  Deck received a minimal P0 SQLite optimization pass for clipboard history. FTS now updates only when search_text, source app name, custom title, or encryption state actually changes, so lightweight operations such as tagging items or toggling temporary state no longer delete and rebuild full-text index entries. This reduces SQLite WAL write amplification, tokenizer work, and checkpoint pressure. In secure mode, Deck also stops maintaining unusable ciphertext FTS entries, and FTS rebuilds skip encrypted rows. Batch insert paths no longer run a per-item SELECT count, reducing database load during imports, sync, and restore flows. The secure-mode semantic-vector cache migration now probes encryption state silently as well, preventing expected plaintext-probe misses from being logged repeatedly as authenticationFailure errors.

• Trace 驱动的规则搜索与 SwiftUI 氛围栏性能优化 基于 8 份 Instruments Trace，Deck 优先处理了最重的规则搜索、代码语言规则搜索、加密数据库规则搜索与 SwiftUI 更新风暴热点：语言规则过滤现在会先用类型、App、日期、大小等低成本条件缩小候选范围，再只对真正可能命中的文本条目预热代码语言检测缓存，避免对会被其它规则淘汰的历史记录反复执行 detectCodeLanguage；底部氛围栏也减少了每帧字符串分配，并把装饰性动画刷新率降到更克制的 10fps，在不改变交互和功能的前提下降低 SwiftUI 文本排版与 AttributeGraph 更新压力。后台复制/纯后台 Trace 中主要耗时来自 MultipeerConnectivity 系统发送/等待线程，这次保持不改，避免为了不可控系统栈引入回归风险。
  Based on eight Instruments traces, Deck prioritizes the heaviest rule-search, code-language rule-search, encrypted-database rule-search, and SwiftUI update-storm hotspots. Language-rule filtering now narrows candidates with cheaper type, app, date, and size checks before warming the code-language detection cache only for text-like items that can still match, avoiding repeated detectCodeLanguage work for records that other rules will discard anyway. The bottom ambient bar also reduces per-frame string allocation and lowers its decorative animation cadence to a more conservative 10fps, cutting SwiftUI text-layout and AttributeGraph update pressure without changing behavior or functionality. The background-copy / pure-background traces were dominated by MultipeerConnectivity system send/wait threads, so those paths were intentionally left untouched to avoid regression risk in an OS-owned stack.

• 冷启动面板查询、CPU 与内存占用优化 根据冷启动面板查询与内存 Trace，Deck 现在会优先提交更小的首屏历史快照，并把总数统计延后到首屏显示之后，避免 COUNT(*) 与首屏列表查询在打开面板瞬间争抢 SQLite 队列；历史列表的类型/标签过滤也补充了与 timestamp DESC, id DESC 排序匹配的复合索引，减少过滤分页时的额外扫描和排序。与此同时，横板卡片的缩略图缓存容量进一步收紧，图片文件大小、Base64 图片探测、智能文本分析与链接元数据加载都改为更靠后的低优先级路径，避免面板刚打开就同时触发解码、字符串扫描、网络元数据解析和 SwiftUI 布局压力，让冷启动面板更快出现、CPU 峰值更平滑，内存峰值也更可控。
  Based on cold-start panel query and memory traces, Deck now commits a smaller first-screen history snapshot first and defers total-count calculation until after the panel has rendered, preventing COUNT(*) from competing with the initial SQLite list query during panel open. Type/tag-filtered history pages also gain composite indexes matching the timestamp DESC, id DESC ordering to reduce extra scans and sorting during filtered pagination. At the same time, horizontal-card thumbnail cache limits have been tightened, while image file-size lookup, Base64 image detection, smart-text analysis, and link metadata loading now run later at lower priority. This avoids triggering image decoding, string scanning, network metadata parsing, and SwiftUI layout pressure all at once, mak

Release Notes v1.4.2

新增 / Added

• Smart Rule AI 管理工具
  Deck AI 现在不仅能创建 Smart Rule，也可以列出、读取、修改和删除已有规则；App 内审批面板与 deckclip chat/deckclip 终端审批也会显示对应的规则预览、变更摘要和删除警告。
  Deck AI can now do more than create Smart Rules. It can also list, read, modify, and delete existing rules, and both the in-app approval sheet and the deckclip chat approval overlay now show rule previews, change summaries, and delete warnings.

• Deck CLI AI 聊天审批
  在 deckclip chat 中，遇到脚本插件创建/修改/删除、剪贴板写入/删除、Smart Rule 生成等需要授权的操作时，现在会直接弹出审批面板；你可以在终端里查看摘要与预览，并用 Y / Enter 批准、N / Esc 拒绝。
  deckclip chat now shows an approval overlay for actions that require authorization, including creating, modifying, or deleting script plugins, writing or deleting clipboard items, and generating Smart Rules. You can review the summary and preview in the terminal, then approve with Y / Enter or reject with N / Esc.

• 审批补丁结构化预览
  修改脚本插件时，补丁现在会按文件、hunk 与增删行数展示；支持普通 diff 与 *** Begin Patch 风格补丁，也能识别文件移动（move）并显示旧路径到新路径的变更关系。
  Patch previews for script plugin changes are now structured by file, hunk, and added or removed lines. Both regular diffs and *** Begin Patch style patches are supported, and file moves are recognized with old-to-new path mapping.

• 审批代码高亮
  审批面板会对 manifest.json、JSON 和 JavaScript 代码块做基础语法高亮，长预览也支持独立滚动。
  The approval overlay now applies basic syntax highlighting to manifest.json, JSON, and JavaScript blocks, and long previews can be scrolled independently.

优化 / Improvements

• Smart Rule 自动删除粒度
  “多久之后自动删除” 现在最低支持 1 分钟；在 5 分钟以内可直接选择 1 / 2 / 3 / 4 / 5 分钟，超过 5 分钟后继续沿用原来的 5 分钟步进逻辑。AI 生成或修改规则时也同步支持 1 分钟的最小自动删除时间。
  “Auto delete after” now supports a minimum of 1 minute. Within the first 5 minutes, you can choose 1 / 2 / 3 / 4 / 5 minutes directly, while values above 5 minutes continue to use the original 5-minute step cadence. AI-generated or AI-modified rules now support the same 1-minute minimum.

• 聊天过程工具状态反馈
  AI 在搜索剪贴板、保存记忆、读写会话上下文、运行脚本插件等阶段时，顶部状态与对话尾部会显示更明确的进行中文案；连续搜索还会带上递增计数，减少“像卡住了一样”的观感。
  During clipboard search, memory saving, session context reads or writes, script plugin execution, and similar steps, the chat UI now shows clearer in-progress status text in both the header and transcript tail. Repeated searches also display an incrementing count to reduce the feeling that the session is stuck.

• macOS 审批键盘体验
  审批面板打开时，会临时切到可输入 ASCII 的键盘布局，关闭后再恢复原输入法；在中文输入法等场景下，Y / N 快捷批准或拒绝更可靠。
  When the approval overlay opens on macOS, Deck temporarily switches to an ASCII-capable keyboard layout and restores the previous input source afterward, making Y / N approval shortcuts more reliable under Chinese IMEs and similar input methods.

• 聊天模块结构与可维护性
  deckclip chat/deckclip 的实现拆分为 app_impl、approval、render 和 tests 模块，后续继续迭代聊天与审批功能时更容易维护。
  The deckclip chat implementation has been split into app_impl, approval, render, and tests modules, making future iteration on chat and approval features easier to maintain.

• 局域网共享后台策略平衡优化
  重新调整了局域网共享在后台、熄屏与连接状态切换时的发现策略：相比频繁暂停、恢复或重建网络对象，Deck 现在更倾向于保持发现链路稳定运行，以优先降低异常退出与崩溃风险，同时减少因重复重建带来的额外开销。
  Rebalanced LAN sharing discovery behavior across background, display sleep, and connection state transitions. Instead of frequently pausing, resuming, or rebuilding networking objects, Deck now prefers to keep the discovery path stable, reducing crash risk first while also avoiding the overhead caused by repeated teardown and rebuild churn.

• 剪贴板后备轮询更省电
  剪贴板监听的后备轮询现在会根据用户空闲状态、低功耗模式和设备热状态动态放宽间隔，并减少无变化时的重复调度，降低后台闲置时的不必要唤醒。
  Clipboard fallback polling now ad