Maitu Dingxin Enterprise AI Source System

Helping corporate websites become reliable sources for AI, with an all-in-one SEO + GEO digital solution. MaiTu DingXin upgrades corporate websites from "being searched" to "being cited" — making every enterprise an answer in the AI era.

Current version: V1.1.0 (Release, 2026-08-15) For: Enterprises / institutions that run their own brand website; agencies that build sites for clients.

1. Product Overview

MaiTu DingXin is a multi-template CMS for corporate websites. It cleanly separates the site framework from the page content:

• The system handles: data read/write, routing, authentication, admin, SEO / GEO, and deployment.

• The template handles: the site's visual identity and page structure (Home / Navbar / Footer can be fully taken over by a template).

Beyond that, MaiTu DingXin upgrades a corporate website from "waiting to be found by search engines" to "being cited by generative AI" — putting your content into the source pool that models such as ChatGPT, Claude, ERNIE (Wenxin) and Qwen draw on when composing answers, so your site becomes one of the cited sources in AI-generated responses. Switching styles is a single config field (settings.template) — no code change, no rebuild required. All operational content lives in the admin panel, so upgrading the system means "swap the code, keep the data."

2. Key Features

1. AI Source & GEO (Generative Engine Optimization)

• Auto-generated /llms.txt (following the llmstxt.org spec): exposes your brand, intro, columns, products, solutions, news, and compliance pages as LLM-friendly plain-text Markdown, refreshed in real time with admin content. Generative AI can understand and cite your content accurately when crawling.

• JSON-LD structured data: Organization / BreadcrumbList / content-type schemas are injected to boost credibility and citability in both traditional search engines and AI summaries.

• SEO + GEO in one: the same admin content serves traditional search engines (SEO) and generative AI engines (GEO, Generative Engine Optimization) — no need to maintain two separate content sets.

• From "being searched" to "being cited": your website is no longer just waiting to be searched; it becomes a cited source when AI composes answers, making every enterprise an answer in the AI era.

2. Multi-Template Architecture · One-Click Reskin

• Site-wide colors are driven by CSS variables; change one place and the whole site reskins.

• Three built-in templates:

• tech-blue (Tech Blue, default) — bright blue + violet, light Hero.

• natural-green (Natural Green) — for eco / health / consumer brands.

• adipo-dark (Dark Takeover Demo) — a full-page dark template inspired by high-end biopharma sites.

• Full template takeover: a template may supply Home / Navbar / Footer components to take over entire pages; if absent, the system falls back to its block-based layout.

3. Bilingual (next-intl)

• Defaults to Chinese; all columns / detail pages / SEO metadata support Chinese–English switching, with language-prefixed URLs.

4. 10 Content Types with Admin CRUD

Visual editing, no coding required: Homepage, Products, Solutions, Case Studies, News, About Us, Careers (Jobs), Navigation, Columns, and System Settings.

5. Rich Text Editing

• Tiptap-based visual editor; output is sanitized with DOMPurify before render to prevent XSS.

6. Frontend Messages

• Contact form submissions are stored in the messages table and can be reviewed and followed up from the admin.

7. Site Maintenance Mode

• One-click siteClosed shuts down the frontend (shows a maintenance page) while /admin stays fully operational for configuration.

8. Admin Experience

• Page Management: merges the former "Columns / Modules / Instances" into a single entry, each row offering Content / View / Settings.

• Workbench: todo reminders, one-click backup, last-backup time, and backup/recovery access.

• Global Search: cross-module search in the admin top bar (title / description / body) with 300ms debounce.

• Frontend Direct-Edit Bar: admins see an "Edit this page" floating bar at the bottom-right of the frontend, jumping straight to the matching admin editor.

• Card-style List Rows: admin lists render as cover + bold title + de-emphasized metadata + actions for a friendlier information density.

9. SEO & Social Sharing

• Dynamic SEO for column / detail pages: title / keywords / description generated per language; per-entry keyword override supported.

• OpenGraph / Twitter Card explicitly populated for richer share cards.

• Schema Markup (JSON-LD): Organization / BreadcrumbList / content-type schemas injected to aid indexing.

• Works together with AI Source & GEO so one content set covers both traditional search and generative-AI discovery paths.

10. Security (Enterprise-grade)

• Auth: bcrypt + JWT cookie; login validates against the database only; ADMIN_JWT_SECRET is read from environment variables.

• Rate limiting: in-process sliding-window limiter on login and contact forms against brute force / abuse.

• Write locking: serialized transactional writes prevent concurrent overwrites.

• Upload auth: image uploads require login.

• Envelope encryption for secrets: API keys etc. are stored AES-256-GCM encrypted; keep the master key separate from database backups.

11. Performance

• Automatic image compression on upload: longest edge capped at 1920px, converted to WebP (quality 80); SVG/GIF pass through unchanged.

12. Backup & Observability

• Database four-layer backup (L1–L4): local snapshots / volume backup (incl. uploads) / versioned archive / optional offsite.

• Health check: /api/health endpoint, wired into the Docker HEALTHCHECK.

13. Quality Assurance

• Vitest unit / integration tests: npm test passes in full (185 cases), covering the data layer, envelope encryption, SEO, templates, and admin logic.

3. Technical Architecture

Data model: the data table stores 12 content types as key-value (about / cases / categories / company / homepage / navigation / news / products / settings / solutions / users / jobs); the messages table holds frontend submissions.

4. Installation & Deployment

For detailed installation, migration, packaging and data-recovery steps, see the Deployment, Packaging & Recovery Guide.

Deployment constraints:

• PostgreSQL required: without it, initialization fails (routes return 500 — not a code bug).

• Always-on form: needs a persistent Node process + writable local disk + a live PostgreSQL connection; not for Vercel / Serverless / pure static export.

• Single-replica recommended: cross-replica cache invalidation is not yet implemented; run a single replica (pm2 fork 1 / docker instances 1).

5. Upgrading

Versions follow Semantic Versioning (SemVer): PATCH for copy/default-content/bugfix (no migration); MINOR for backward-compatible features; MAJOR for breaking changes (migration script required).

Core principle: swap the code, keep the data. User content lives in the PostgreSQL database and the data/ volume — preserve that database / volume on upgrade and only replace the code.

6. License & Copyright

Copyright holder: Maitu Cloud (Xi'an) Technology Co., Ltd. Contact: contact@mai-2.com ｜ Website: www.mai-2.com

This system is protected under the Copyright Law of the People's Republic of China and international copyright treaties. Copying, modifying, distributing, reverse-engineering, or commercial use without written authorization is prohibited. For licensing inquiries, contact the email above.

This overview corresponds to V1.1.0 (2026-08-15).