在线PDF红线批注对比与变更摘要工作台
目标：把两份PDF的批注、涂鸦、文本框、印章等注释元素做结构化对比，并输出“新增/删除/修改”清单与可下载报告。
英文 Prompt:
You are to generate a complete, runnable web application called "PDF Annotation Diff Workbench". Output: (1) a full file tree, (2) all source code, (3) exact install/run commands, (4) Dockerfile + docker-compose, (5) deployment notes for Linux VPS, (6) at least 8 test cases or a QA checklist.
Requirements:
- Stack: Next.js (App Router) + Type + Node.js API routes.
- Upload two PDFs, parse annotations (including highlights, underline, strikeout, free text, ink, shapes) using a reliable PDF library.
- Normalize annotations into a JSON schema with page number, bounding boxes, color, author, timestamp, contents, subtype.
- Compute a diff: added/removed/changed annotations and generate a human-readable change summary.
- Provide an export: JSON + Markdown + CSV.
- No image generation features.
- Include security: file size limits, input validation, antivirus hook placeholder, and memory-safe streaming.
- Include UI: side-by-side page navigator, filter by author/type, search annotation contents.
中文释义:让AI生成一个可部署的在线工具：上传两版PDF后自动提取注释并对比，输出变更清单与摘要，适合法务审阅/合同修订留痕。
批注元素结构化抽取与统一JSON规范
目标：把不同PDF库返回的注释字段统一成可比对的规范数据，便于后续审计、检索与导出。
英文 Prompt:
Generate a reusable Type module for extracting PDF annotations and converting them into a normalized schema. Provide full project code, file tree, and unit tests.
Constraints:
- Provide types: Annotation, Rect, Color, DiffRecord.
- Implement parsers/adapters for two libraries (e.g., pdf-lib and pdfjs-dist) with a common interface.
- Include deterministic sorting and stable IDs for annotations.
- Include tests with sample fixtures and assertions.
- Provide a CLI command example: node ./ s/extract.mjs input.pdf --out annotations.json
- Include at least 6 edge cases (rotated pages, missing timestamps, duplicated IDs, multi-quadpoints, zero-size rects, non-UTF8 strings).
中文释义:让AI产出可复用的抽取模块与测试，统一注释数据结构，解决“不同库字段不一致导致无法对比”的问题。
在线批注检索与合规审计（不做图片生成）
目标：把批注内容做全文检索、敏感词扫描、作者与时间范围筛选，并生成审计报告。
英文 Prompt:
Build a web tool "PDF Annotation Audit" that accepts a PDF and produces an audit report.
Must include:
- Full Next.js + TS project with file tree, run/deploy commands, Docker.
- Extract annotations + data; index annotation contents in an in-memory search (or SQLite if running locally).
- Rules engine: configurable regex patterns, whitelist, severity levels.
- Output: audit report page + downloadable Markdown/CSV.
- QA checklist: include at least 10 items covering performance, correctness, and security.
- Do NOT generate images; only process existing PDF data.
中文释义:生成在线审计工具：针对PDF批注做检索与合规检查，输出报告，适合项目交付前的批注清理与留痕审计。
批注合并与冲突检测工作台
目标：多方评审产生多份带批注的PDF，在线合并批注并检测冲突（同位置不同意见/重复批注）。
英文 Prompt:
Create a web app that merges annotations from multiple PDFs that share the same content.
Functional requirements:
- Upload 1 PDF + N annotated PDFs.
- Extract annotations from each annotated PDF, align by page + geometry, detect duplicates and conflicts.
- Provide a merge plan UI and export a merged annotation JSON.
- Provide a command to apply merged annotations back to a PDF output file (server-side).
Deliverables: full code + file tree + run commands + deployment + tests/QA.
中文释义:让AI生成一个“批注合并+冲突检测”的在线工具，适合多人评审收敛意见，避免遗漏与重复。
批注变更时间线与审批流导出
目标：把批注按时间轴排序，生成“谁在什么时候提了什么意见”的可追溯记录，并导出审批流材料。
英文 Prompt:
Generate a full-stack web tool that creates an annotation timeline from a PDF: authors, timestamps, contents, and page references.
Specs:
- Next.js UI + Node API.
- Timeline view with filters.
- Export: text- d PDF summary, Markdown, JSON.
- Include authentication (simple password- d session) and rate limiting.
- Provide at least 5 automated tests and a QA checklist.
中文释义:生成可部署的时间线工具，把批注变成审批证据链，便于归档与复盘。
批注坐标校准与页旋转/裁切适配
目标：解决批注坐标在不同渲染器/旋转裁切条件下偏移的问题，输出可复现的坐标转换报告。
英文 Prompt:
Create a developer-focused online tool that visualizes PDF page coordinate systems and annotation rectangles.
Requirements:
- Upload PDF, list pages with rotation/cropbox/mediabox.
- Overlay annotation rects on a canvas preview ( d on existing PDF rendering, not AI images).
- Provide coordinate transform utilities and export a calibration report.
- Deliver full code, file tree, run/deploy commands, and tests.
中文释义:生成坐标校准工具：可视化页框与批注矩形，帮助定位“批注偏移”问题，并输出转换报告。
批注内容脱敏与一键清理（保留审计记录）
目标：对批注中的手机号、邮箱、地址等敏感信息做脱敏处理，同时保留审计日志与变更记录。
英文 Prompt:
Build an online PDF Annotation Redaction tool.
Features:
- Extract annotation contents.
- Detect sensitive patterns (phone/email/address) with configurable rules.
- Offer redaction options: mask text in exported reports; optionally remove annotations from an output PDF.
- Always generate an audit log and a before/after diff.
Deliverables: full runnable project + file tree + commands + Docker + tests/QA.
中文释义:生成脱敏清理工具：处理现有批注文本，不生成图片；适合对外发版前的合规处理。
批注与正文引用定位（段落锚点）
目标：把批注关联到正文附近的文本片段（锚点），导出“批注-引用-页码”的对照表，便于沟通与修改。
英文 Prompt:
Generate a web tool that s annotations to nearby extracted text snippets.
Must include:
- Extract text per page; for each annotation rect, find the nearest text blocks.
- Export a mapping table: annotation id, page, author, comment, text snippet, confidence score.
- Provide UI to review and correct mappings.
- Provide full code + file tree + run/deploy commands + tests.
中文释义:生成“批注与正文引用定位”工作台，把意见精确落到正文片段，导出对照表用于改稿沟通。
