shenlei
3e60bf1869
阅读器在多会话朗读、加密 EPUB 资源、并发打开同一本书和多 UserDefaults 容器等场景下，存在旧异步结果覆盖新状态、锁屏控制串会话、大型明文资源被误拒绝、解压半成品被复用及标注跨容器混用的风险；同时高亮、书签、搜索和设置面板的空状态与自动化可访问性入口不完整。\n\n本次为朗读会话引入代次和当前 utterance 校验，远程控制改为仅响应当前会话并按自身 token 清理；加密资源 provider 先判定是否实际返回解密数据，明文大资源继续流式读取；解压缓存串行化以避免并发复用未完成目录。书签与高亮迁移至受保护文件，按 UserDefaults 容器隔离命名空间，保留标准容器旧文件兼容并确保新副本成功落盘后才删除历史数据。\n\n同步调整缓存淘汰、FoundationModels 弱链接、搜索/标注/设置面板交互与 UI 测试，并更新 Pod 生成配置及 API、风险文档。已执行 git diff --check 和 ReadViewDemo Debug Simulator 构建，结果为 BUILD SUCCEEDED。
RDAIReaderView
RDAIReaderView is the local indexing and retrieval foundation for ReadViewSDK AI reader features. Version 0.1 provides Natural Language sentence chunking, language detection, named-entity candidates, source-backed citations and local lexical retrieval. It does not upload book content.
Install
pod 'RDAIReaderView/NaturalLanguage'
pod 'RDPDFReaderView/AI'
pod 'RDEpubReaderView/AI'
Use with PDF
let service = reader.makeAIReaderService()
try await service.prepareIndex(
options: RDAIIndexOptions(scope: .wholeDocument)
)
let matches = await service.retrieve(
query: "主角为什么离开",
options: RDAIRetrievalOptions(scope: .wholeDocument)
)
if let match = matches.first {
let citation = RDAICitation(
passageIdentifier: match.passage.id,
quote: match.passage.text,
locator: match.passage.locator
)
try await reader.makeAIContentProvider().aiShowCitationHighlight(citation)
}
PDF citations retain native-text versus OCR provenance. EPUB citations use normalized href and derive CFI ranges when navigating.
Foundation Models generation, persistent SQLite storage and the AI UI are planned follow-up modules; the Core API is intentionally independent of those capabilities.