OCR

OCR is a full-stack document processing app that turns uploaded PDFs into structured outputs through a queued OCR pipeline.

App: https://ocr.tuturu.io

Features

• PDF upload and process tracking
• PDF split pipeline
• OCR transcription and post-processing
• Structured output download
• OpenAPI docs exposed by the backend outside production

Stack

• Frontend: React, TypeScript, Vite, TanStack Router
• Backend: Node.js, TypeScript, tRPC
• Database: PostgreSQL with Drizzle ORM
• Queue and cache: RabbitMQ, Redis
• Object storage: MinIO / S3-compatible storage
• Monorepo: pnpm workspaces

Development

Install dependencies:

pnpm install

Start infrastructure with Docker:

docker compose up -d

Start the app in dev mode:

pnpm dev

pnpm dev runs a runtime build first, then starts the frontend, backend, and workers.

Build

Build runtime packages and apps:

pnpm build

Run lint across workspaces:

pnpm -r --if-present lint

Production Docker

Production compose file:

docker compose -f docker-compose.prod.yaml up -d --build

Environment template:

• .env.docker.example

Main exposed ports in production compose:

• Front services: configure separately if you add the frontend service
• Backend: 4010
• Postgres: 5436
• Redis: 6380
• RabbitMQ: 5673
• MinIO API: 9002

Notes

• The production Dockerfiles build workspace artifacts and run compiled files from dist.
• The frontend is currently not included in docker-compose.prod.yaml.