AI Analysis Cache

AI Analysis Cache

关键词：整本书问答、AI 分析、embedding、缓存、检索、PDF 问答、文档问答。

AI analysis data is the local embedding/cache layer used for document-aware Q&A.

Flow

Open document
  -> Build text chunks
  -> EmbeddingClient
  -> PDFEmbeddingStore

Question
  -> Retrieval from cached analysis data
  -> Prompt
  -> AIChatPanel

Files

• ReaderWindowController+Embedding.swift: background AI analysis state, progress, cache restore, controls.
• PDFDocumentAgentIndex.swift: chunking and retrieval scoring.
• PDFEmbeddingStore.swift: local SQLite-backed embedding cache.
• EmbeddingClient.swift: embedding API client.
• AISettingsPanelController*.swift: model and AI analysis settings UI.

User-Facing Terms

• UI should prefer “AI analysis data” or “AI analysis cache” over “vector index” unless the setting is explicitly about an embedding model/provider.
• Button labels should match scope: 生成/更新本书缓存 fills missing current-book embeddings, 清除本书分析缓存 deletes only the current book's embeddings, and 清除全部缓存 deletes all local embedding cache rows.

Related Files

• mac-app/ReaderWindowController+Embedding.swift
• mac-app/ReaderWindowController+EmbeddingBackfill.swift
• mac-app/ReaderWindowController+EmbeddingStatus.swift
• mac-app/PDFDocumentAgentIndex.swift
• mac-app/PDFEmbeddingStore.swift
• mac-app/EmbeddingClient.swift
• mac-app/EmbeddingActionPolicy.swift