GitDigest

Discover what's trending on GitHub with AI-powered summaries, a trend wordcloud, and a reading activity heatmap — all in one split-pane dashboard.

🚀 Live: https://gitdigest.td-rootx.workers.dev/

Highlights

• Split-pane dashboard — browse a feed of trending repos on the left, AI summary / wordcloud / README on the right. Mobile-first responsive layout.
• Multi-provider AI — bring your own key for OpenAI, Groq, OpenRouter, or Gemini. No vendor lock-in.
• Zero cold start — deployed to the edge on Cloudflare Workers, cached per region.

Features

• 📊 Trending repositories — daily / weekly / monthly, with language and topic filters
• 🤖 AI summaries — structured technical breakdowns (overview, stack, features, traction)
• ☁️ Trend wordcloud — visual technology trend analysis + category / insight panels
• 💬 Chat with AI — interactive Q&A per repository (grounded on the summary)
• ⭐ Favorites & filters — star repos, filter by favorites / unread, one-click clear all
• 📅 Reading activity — GitHub-style contribution heatmap that tracks what you've read
• 🌍 Multilingual — English & Vietnamese with automatic translation between cached entries
• 🎨 Polished motion — staggered card entrance, slide-in panels, water-flow toggles, subtle hover lifts
• 🧩 Custom UI primitives — dropdowns, compact filter toolbar, status pills, tooltip
• 🛡️ Built-in security — per-IP rate limiting, input validation, prompt-injection escaping
• 📱 Responsive — split-pane on desktop, stack-with-back on mobile

Requirements

• Node.js 18 or newer (for Wrangler)
• API key from one of the supported providers:
  • OpenAI (starts with sk-)
  • Groq (starts with gsk_)
  • OpenRouter (starts with sk-or-)
  • Gemini (starts with AIza...)

Setup

1. Install dependencies:

   npm install

2. Log in to Cloudflare (once):

   npx wrangler login

3. Create the KV namespace for the shared cache layer:

   npx wrangler kv namespace create gitdigest_cache

   Paste the returned id into wrangler.jsonc under kv_namespaces[].id (replacing the placeholder). The binding name must stay CACHE_KV — that's what the worker reads via env.CACHE_KV.

   Local dev (npm run dev) uses miniflare's in-memory KV simulation, so no preview namespace is needed. If you want to hit the real KV from dev, run npm run dev -- --remote.

4. (Optional) For local npm run dev, create .dev.vars in the project root (do not commit it):

   API_KEY=sk-...        # or gsk_... for Groq, sk-or-... for OpenRouter, or AIza... for Gemini
   GITHUB_TOKEN=ghp_...          # GitHub Personal Access Token
   
   # Optional: Custom AI models (defaults shown)
   OPENAI_MODEL=gpt-4o-mini
   GROQ_MODEL=llama-3.3-70b-versatile
   OPENROUTER_MODEL=nvidia/nemotron-3-super-120b-a12b:free
   GEMINI_MODEL=gemini-2.0-flash-lite

   Note: GITHUB_TOKEN is required for production to avoid rate limits (60 requests/hour unauthenticated vs 5,000 requests/hour authenticated).

5. Run locally (Worker + static assets):

   npm run dev

   Wrangler prints the local URL (often http://localhost:8787).

6. Set up GitHub API token (required for production):

   npx wrangler secret put GITHUB_TOKEN
   # Create a Personal Access Token at: https://github.com/settings/tokens
   # Scopes needed: public_repo (read-only access to public repositories)

7. (Optional) Set default API keys and models on the deployed Worker:

   # Default AI API key
   npx wrangler secret put API_KEY
   
   # Optional: Custom AI models
   npx wrangler secret put OPENAI_MODEL
   npx wrangler secret put GROQ_MODEL
   npx wrangler secret put OPENROUTER_MODEL
   npx wrangler secret put GEMINI_MODEL

8. Deploy:

   npm run deploy

API

Routes are implemented in TypeScript (src/handlers.ts):

Repository Endpoints

• GET /api/repos?page=1&period=daily&lang=javascript — Paginated trending repositories (cached 15 minutes)
• GET /api/repo?id=<repoId> — Repository details and README content
• GET /api/summarize?id=<repoId>&lang=<iso> — AI-powered repository summary
• POST /api/ask — Interactive Q&A about repositories
• GET /api/wordcloud?period=daily&lang=en — Technology trend word cloud

Admin Endpoints

• GET /api/admin/stats — Cache statistics and system health
• POST /api/admin/clear — Clear specific or all caches (optional { type } body to clear a single cache)

Authentication

Send Authorization: Bearer <API key> header or set API_KEY secret.

Rate Limiting

• Summarize: 5 requests/hour per IP
• Ask: 10 requests/hour per IP
• WordCloud: 20 requests/hour per IP

Architecture

• Frontend: Modular JavaScript with ES modules (public/js/)
• Backend: TypeScript with Cloudflare Workers (src/)
• Caching: Hybrid L1 (in-memory Map per isolate) + L2 (Cloudflare KV) — shared across isolates, survives redeploy
• Security: Input validation, rate limiting, and abuse prevention
• AI: Multi-provider support with configurable models

Configuration

Bindings

Binding	Kind	Purpose
CACHE_KV	KV Namespace	L2 cache shared across isolates (summaries, Q&A, wordcloud, repo, trending)
RATE_LIMIT_KV	KV Namespace	IP-based rate limit counters (optional)
ASSETS	Static assets	public/ served via Workers Assets

Environment Variables

Variable	Default	Description
API_KEY	-	Default AI API key fallback (stored via wrangler secret put)
GITHUB_TOKEN	-	GitHub Personal Access Token
OPENAI_MODEL	gpt-4o-mini	OpenAI model
GROQ_MODEL	llama-3.3-70b-versatile	Groq model
OPENROUTER_MODEL	nvidia/nemotron-3-super-120b-a12b:free	OpenRouter model
GEMINI_MODEL	gemini-2.0-flash-lite	Gemini model

Cache Configuration

Every cache runs as a HybridCache<T> — L1 (per-isolate Map, LRU-evicted) backed by L2 (KV) for cross-isolate sharing.

Cache	L1 max	TTL	KV prefix
Trending lists	50	30 min	trending:
Repository details	200	30 min	repo:
Summaries	500	24 h	summary:
Ask Q&A	1000	6 h	ask:
Word cloud	100	30 min	wc:

KV writes are non-blocking via ctx.waitUntil() so response latency isn't affected. On a cold isolate the first read that hits L2 warms up L1 in ~10-30 ms — no need to re-call upstream APIs.

If CACHE_KV isn't bound (e.g. during a quick test), caches transparently degrade to L1-only (in-memory per isolate).

Security

• ✅ Input Validation: Type checking and sanitization
• ✅ Rate Limiting: IP-based with atomic operations
• ✅ Prompt Injection: Escaped user input
• ✅ Cache Security: Hash-based keys prevent collisions
• ✅ Abuse Prevention: Pattern filtering and validation

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

License

This project is licensed under the MIT License — see LICENSE.

---

Built with ❤️ by ThuanD

Deployed on Cloudflare Workers: https://gitdigest.td-rootx.workers.dev/