fitse-yotor/Amharic-OCR-Telegram-BOT

# Amharic OCR Platform Multilingual OCR platform with: - **React frontend** for image upload/paste, crop, and OCR preview - **Telegram bot** that reads text from images and replies with extracted text - **Hybrid OCR architecture**: - Python **EasyOCR** service (primary) - Node **Tesseract.js** fallback (automatic) Supports Amharic-focused extraction with preprocessing and text cleanup. ## Features - Upload image or paste from clipboard - Crop target area before OCR - Multi-language OCR (default `amh+eng`) - Telegram bot `/start` welcome + OCR from photo messages - Waiting indicator while bot processes image - OCR post-processing to reduce noisy characters and broken lines ## Tech Stack - Frontend: React + Vite - Bot: Node.js + `node-telegram-bot-api` - OCR service: Python + FastAPI + EasyOCR - Fallback OCR: Tesseract.js ## Project Structure `src/` - frontend app `bot/` - Telegram bot `python-ocr/` - Python OCR microservice `BOT_SETUP.md` - detailed bot setup notes ## Prerequisites - Node.js 18+ (recommended) - Python 3.10+ - pip ## Environment Variables Create `.env` in project root: ```env TELEGRAM_BOT_TOKEN=your_new_token_from_botfather OCR_LANG=amh+eng OCR_SERVICE_URL=127.0.0.1 ``` ## Installation ```bash npm install pip install -r python-ocr/requirements.txt ``` ## Run Commands - Frontend only: ```bash npm run dev ``` - Bot only: ```bash npm run bot ``` - Python OCR service only: ```bash npm run ocr:service ``` - Frontend + Bot: ```bash npm run dev:all ``` - Frontend + Bot + Python OCR service (recommended): ```bash npm run dev:stack ``` ## Build ```bash npm run build ``` ## How OCR Flow Works 1. Telegram user sends photo. 2. Bot downloads best-quality image. 3. Bot calls Python OCR service (`/ocr`). 4. If Python service is unavailable, bot falls back to Tesseract pipeline. 5. Bot cleans extracted text and replies in one or multiple messages. ## Troubleshooting - **`Missing TELEGRAM_BOT_TOKEN in .env`** - Ensure `.env …