Source Library v2

A Next.js application for digitizing and translating historical texts. Built for the Embassy of the Free Mind.

Stack

• Framework: Next.js 14 (App Router)
• Database: MongoDB Atlas
• AI: Google Gemini for OCR and translation
• Storage: Vercel Blob for images
• Deployment: Vercel

Getting Started

npm install
npm run dev

Open http://localhost:3000

Architecture

Image System

All page images go through /api/image for consistent sizing and cropping:

Split Pages

Books with two-page spreads can be split. Each page stores:

• crop.xStart and crop.xEnd (0-1000 scale)
• cropped_photo (optional pre-generated Vercel Blob URL)

Cropping happens on-demand via Sharp. OCR automatically crops inline and saves the result for future use.

Processing Pipeline

1. Import - Upload images or import from Internet Archive
2. Split - Detect and split two-page spreads (ML or manual)
3. OCR - Extract text using Gemini Vision
4. Translate - Translate to English using Gemini
5. Summarize - Generate summaries and key themes

Key Directories

src/
├── app/
│   ├── api/          # API routes
│   ├── book/         # Book pages (detail, read, pipeline)
│   └── page.tsx      # Homepage
├── components/       # React components
└── lib/              # Utilities (mongodb, ai, types)