How haokunwen/awesome-composed-image-retrieval works
Key Architectural Components
a. Content Storage
• Git repository (Markdown files—currently only README.md, but can grow)
• Directory structure for categorization (future “docs/”, “_data/”, etc.)
b. Static Site Generator (SSG) Layer
• Tool of choice (e.g., Jekyll, Hugo, MkDocs, Docusaurus) that transforms Markdown into HTML
• Plug-ins for navigation, search indexing, theming
c. Continuous Integration / Deployment (CI/CD)
• GitHub Actions or other CI that:
– Automatically builds the site on push or pull-request merge
– Runs link-checker, Markdown linter
– Deploys artifacts to hosting
d. Hosting and Delivery
• GitHub Pages, Netlify, or similar (serves static assets)
• Optional CDN (e.g., Cloudflare) for caching and faster global delivery
e. Consumers / Users
• Web browsers (researchers/developers)
• Optional programmatic consumers (if an API or JSON index is provided)
Relationships and Data Flows
1. Developer clones or forks the Git repo →
2. Adds or updates Markdown content →
3. Pushes to remote → triggers CI/CD →
4. SSG builds the site →
5. CI/CD deploys to hosting →
6. CDN caches and serves static files →
7. End-users browse documentation in their browsers.
Architectural Patterns and Principles
• Static Site Generation (no server-side code at runtime)
• Infrastructure as Code (CI scripts, Git workflows)
• Separation of Concerns:
• Content (Markdown)
• Presentation (theme/templates)
• Build & Deployment (CI/CD)
• Idempotent builds and content versioning via Git
Technologies, Frameworks, and Libraries
• Markdown (content format)
• Hugo / Jekyll / MkDocs / Docusaurus (choose one)
• Git + GitHub (version control & hosting)
• GitHub Actions (CI/CD)
• Optional: Algolia DocSearch or Lunr.js (client-side search)