Deck

A modern, native, privacy-first clipboard OS for macOS

Release Notes v1.0.9

新增 / New

• 语义搜索引擎升级：引入 Apple Sentence Embedding 并与 sqlite-vec 进行静态集成，排序逻辑下沉至数据库内部执行，响应更快、内存占用更低。
  Semantic search engine upgrade: Integrated Apple Sentence Embedding with sqlite-vec (static), moving sorting into the database layer for faster responses and lower memory usage.

• 混合搜索模式优化：文本搜索与语义搜索结果融合排序，提高相关性与召回稳定性。
  Smarter hybrid mode: Merged text and semantic results for higher relevance and more stable recall.

• 中文检索体验增强：引入 FTS5 trigram 分词策略，显著提升 CJK 文本搜索的准确度与性能。
  Enhanced Chinese search: Enabled FTS5 trigram tokenization for more accurate and faster CJK retrieval.

优化 / Improvements

• OCR 性能优化：对超大图片进行下采样处理，降低卡顿概率并减少内存峰值。
  OCR performance optimization: Downsamples large images to reduce UI lag and memory spikes.

• 同步与数据库稳定性增强：优化 CloudKit 批量提交流程，并加入自动备份、完整性校验与恢复机制。
  Sync and database stability: Improved CloudKit batch submission and added automatic backup, integrity checks, and recovery mechanisms.

• 内存占用大幅下降：整体内存使用从约 300MB 降至约 50MB。
  Memory optimization: Reduced memory usage from ~300MB to ~50MB.

技术变更 / Technical Changelog

• 引入 sqlite-vec 静态链接版本，避免动态扩展加载的不确定性并提升启动稳定性。
  Integrated sqlite-vec as a static component to avoid dynamic extension issues and improve startup reliability.

• 搜索排序逻辑由应用层迁移至数据库层执行，减少数据拷贝与中间态内存占用。
  Moved search result sorting into the database layer to reduce data copying and intermediate memory usage.

• FTS5 配置升级为 trigram tokenizer，更适配中文与日韩文本。
  Updated FTS5 configuration to use trigram tokenizer for better CJK support.

兼容性与行为说明 / Compatibility & Behavior Notes

• 本次更新不包含数据库结构变更，可直接覆盖升级。
  This release introduces no database schema changes and supports in-place upgrades.

• 所有搜索、识别与处理逻辑均在本地完成，不上传、不存储任何用户内容。
  All processing is performed locally with no data uploaded or stored remotely.

升级建议 / Upgrade Notes

• 推荐所有用户升级以获得显著更快的搜索体验、更低的内存占用以及更稳定的数据同步行为。
  All users are recommended to upgrade for significantly faster search, lower memory usage, and more stable syncing.