repocost

Estimate the human effort and cost to build any GitHub repository from scratch. Powered by COCOMO II with per-language market rates.

How it works

1. Fetches language byte data from GitHub's API
2. Converts bytes → SLOC using per-language bytes-per-line ratios
3. Applies complexity weights (Rust 1.4×, Java 0.8×, etc.)
4. Runs COCOMO II: Effort = 2.94 × (KSLOC)^1.0997 person-months
5. Costs each language's share of effort at its market hourly rate

Stack

• Next.js 15 (App Router, Edge Runtime)
• Recharts for visualization
• Canvas API for downloadable receipt cards
• Vercel OG for dynamic social preview images
• In-memory cache with 1-hour TTL (API proxy)

Deploy to Vercel

# 1. Clone and install
git clone https://github.com/sfw/repocost.git
cd repocost
npm install

# 2. Set up environment
cp .env.example .env.local
# Edit .env.local — add a GitHub token for higher rate limits

# 3. Run locally
npm run dev

# 4. Deploy
npx vercel

Or just click:

Environment Variables

Variable	Required	Description
GITHUB_TOKEN	Recommended	GitHub PAT for 5K req/hr (vs 60 without)
NEXT_PUBLIC_URL	For OG images	Your deployed URL (e.g. https://repocost.dev)

Shareable URLs

Every estimate gets a shareable URL:

repocost.dev/facebook/react
repocost.dev/vercel/next.js

These generate dynamic OG images, so sharing on X/LinkedIn shows a preview card with the cost estimate.

Caching

Three layers:

1. In-memory server cache — 1hr TTL per repo (survives within a single serverless instance)
2. Vercel Edge Cache — s-maxage=3600 on API responses (shared across all users)
3. Browser cache — max-age=300 (5 min client-side)

For high traffic, swap the in-memory cache for Vercel KV (Redis) — the interface is the same, just change getCached/setCached in lib/github.ts.

License

MIT