NetCare RFC 方案与 AI 审计

NetCare RFC 方案与 AI 审计助手是一款专为网络运维与工程技术人员打造的方案导出、附件归档与 AI 辅助复核效能工具。 【核心功能】 1. 方案一键导出与在线预览：支持输入一个或多个 RFC 变更单号，自动对接 NetCare 原生导出入口生成标准 DOCX 数字化方案，并可在后台联动打开在线原页进行画中画浏览。 2. 章节结构扫描与全附件归档：自动扫描方案完整目录树，智能识别各章节中的各类离线附件及正文内容；支持按规则提取指定章节并打包生成标准 RFC 归档材料压缩包。 3. 章节规则模板与名称纠错：内置章节模板，支持层级继承匹配；当实际方案章节编号缺失或变更时，可按章节完整标题自动纠错定位，保证提取完整性。 4. 多模型 AI 辅助合规审计：支持用户自主接入 OpenAI（Chat / Responses）、Anthropic（Messages）、Gemini 及兼容协议模型（含私有内网模型与自定义端口）。系统按章节针对性提示词执行三态判定（通过 / 不通过 / 需人工核查），提供中英双语摘要与风险发现。 5. 原始区域截图与证据留存：内置离线渲染组件，对提取章节与关键附件所在 DOM 区域捕获清晰 PNG 原图证据，避免纯文本还原失真。 6. 结构化报告与材料导出：一键导出带合并单元格、详细原因与原图排版的标准审计表格，支持下载包含原始方案、附件、截图与审计结论的完整材料压缩包。 7. 本地安全与配置备份：所有模型 API Key、审计记录与快照均保存在本地浏览器 IndexedDB 及加密存储中；支持通过 AES-256-GCM 密码加密导出与恢复完整配置。 【安全与隐私声明】 - 本工具为前端辅助效能扩展，所有网络请求均在用户本地发出，无第三方中转服务器，开发者不收集任何用户数据。 - 方案下载依赖用户当前浏览器中已登录的有效会话，扩展不拦截、保存或上传任何用户身份凭据（Cookie/SSO Token）。 - AI 审计功能默认关闭，且必须由用户自行配置模型接口及 API Key。调用 AI 时仅外发用户明确勾选的章节文本证据，敏感字段（密码、Token、私钥）会自动脱敏处理。 - 审计结论仅供技术人员复核参考，不代表官方审批意见，亦不替代人工技术复核。 NetCare RFC Plan & AI Audit Assistant is a professional productivity extension designed for network engineers and IT operations teams to streamline RFC document export, attachment archiving, and intelligent compliance review. Key Features: 1. One-Click Document Export: Batch retrieve and export standard DOCX digital plans by RFC change ticket numbers, with optional background picture-in-picture tabs for native online review. 2. Full Directory Scan & Attachment Packaging: Automatically parse native section structures, recognize associated offline attachments and body content, and bundle designated sections into organized archive packages. 3. Hierarchy Templates & Title Correction: Match sections with hierarchical templates. If section numbers vary, smart title-based matching accurately identifies targets without manual intervention. 4. Flexible BYOK AI Audit: Bring your own API keys for OpenAI (Chat/Responses), Anthropic (Messages), Google Gemini, or compatible self-hosted private models. Perform three-state compliance checks (Pass, Fail, Needs Review) with dual-language summaries and findings. 5. High-Fidelity Visual Evidence Capture: Utilizes offline-bundled canvas components to capture screenshot evidence of exact DOM elements and attachment tables, preserving the visual integrity of original plans. 6. Professional Audit Spreadsheet & Materials Export: Generate structured audit spreadsheets with merged cells and embedded screenshot rows. Download complete audit dossiers including source plans, attachments, screenshot evidence, and audit conclusions. 7. Local Isolation & Encrypted Backup: All credentials, audit snapshots, and logs are kept strictly within local browser storage (IndexedDB). Full configurations can be exported and restored using AES-256-GCM encryption. Security & Privacy Standards: - Zero Developer Tracking: Fully client-side architecture. The publisher operates no remote servers and collects no telemetry or user data. - Session Protection: Leverages existing user authentication in the active browser session. Never intercepts, extracts, or transmits login cookies or account passwords. - Explicit AI Data Dispatch: AI audit is disabled by default. When enabled, only user-selected section texts are transmitted directly to the user-configured model endpoint. Sensitive credentials (keys, tokens) are strictly redacted before dispatch. - Purely Advisory: AI findings are advisory aids for engineering review and do not execute automated site actions or replace human authorization.