Webrewind

Git history, but for websites.

Webrewind pulls a site's snapshots from the Wayback Machine, deduplicates unchanged captures, and stitches them into a scroll-driven timeline with AI-written captions and a looping GIF reel.

How it works

1. Query the Wayback CDX API for captures in a given year range.
2. Dedupe consecutive snapshots by HTML digest — no ten copies of the same page.
3. Capture each unique snapshot in parallel with headless Chrome.
4. Store frames and an encoded GIF in MinIO/S3.
5. Caption each frame on demand with Gemma 3n (NVIDIA).
6. Stream progress over SSE; render a vertical timeline in React.

Stack

• Client — React 18, TypeScript, Vite, Tailwind, Radix, Framer Motion
• Server — Node, Express, Puppeteer, gif-encoder-2, axios
• Storage — MinIO (S3-compatible)
• AI — NVIDIA Gemma 3n (google/gemma-3n-e4b-it)
• Monorepo — Turborepo + Yarn workspaces

Layout

webrewind/
├── apps/
│   ├── client/   # Vite + React    (@webrewind/client)
│   └── server/   # Express + Puppeteer  (@webrewind/server)
├── turbo.json
└── package.json

Quick start

git clone https://github.com/iamspathan/webrewind.git
cd webrewind
yarn install
yarn dev

Client on http://localhost:5173, server on http://localhost:3200.

You need MinIO reachable on :9000 with a public webrewind bucket. The simplest path is docker compose up -d if a compose file ships with your checkout; otherwise point MINIO_ENDPOINT at your own instance.

Configuration

Copy .env.example to .env in each app.

Server — the essentials:

Variable	Purpose
PORT	server port (default 3200)
MINIO_ENDPOINT	S3 endpoint the server writes to
MINIO_PUBLIC_URL	S3 endpoint the browser reads from
MINIO_BUCKET	bucket name (default webrewind)
MINIO_ACCESS_KEY / MINIO_SECRET_KEY	credentials
NVIDIA_API_KEY	enables AI captions (optional)
CAPTURE_CONCURRENCY	parallel Puppeteer pages (default 4)

Client:

Variable	Purpose
VITE_API_BASE_URL	API base (empty in dev — Vite proxy)
VITE_API_PROXY_TARGET	dev proxy target (default http://localhost:3200)

Scripts

yarn dev                              # run client + server (turbo)
yarn build                            # build everything
yarn lint                             # lint everything
yarn workspace @webrewind/client dev  # client only
yarn workspace @webrewind/server dev  # server only

API

• POST /screenshots — start a capture job; returns { jobId, streamUrl }
• GET /screenshots/events/:jobId — SSE progress stream
• DELETE /screenshots/:jobId — cancel a job
• POST /summaries — caption a single frame
• GET /health · GET /metrics · GET /docs — ops

Full OpenAPI at /docs.

Contributing

PRs and issues welcome. Open one and we'll take a look.

License

MIT.