Document OCR Service
Internal HTTP service that returns the text recognised in uploaded images and PDFs, callable by other applications, with synchronous, streaming or asynchronous processing.
- Area
- AI · Operations & tools
- Main stack
- Node.js · Express · OAuth2 · JWT
Key features
- • Text recognition from images and PDFs over HTTP
- • Synchronous, streaming and asynchronous processing modes
- • OAuth2 client credentials with JWT and distinct scopes
- • Per-client rate limits from a periodically reloaded file
- • Configurable timeouts and retries with increasing backoff
About the project
Service that extracts the text from an image or a PDF and returns it over HTTP, so other applications can use text recognition without embedding it. Callers choose one of three modes: a synchronous call, a streaming response, or an asynchronous job whose state is polled. Access is protected by OAuth2 client credentials with JWT, with read and write scopes and per-client rate limits declared in a file reloaded at intervals. Timeouts, retries and job retention are configurable, and the Express service ships as a Linux binary.
Technologies used
Node.js 
Express 
OAuth2 
JWT 
Mistral AI 
pkg