OCR Skill

This skill provides stand‑alone OCR functionality without any Telegram or Discord bot integrations. It can be used by other agents via the core ocr_core.py module.

Features

• Supports JPG, PNG, WEBP.
• Languages: German (de) and English (en) (easy to extend).
• Uses EasyOCR as primary engine, falls back to Tesseract when needed.
• Optional pre‑processing (grayscale, contrast‑enhancement) for noisy scans.
• Fully async – multiple simultaneous requests are handled without blocking.
• Clean error handling & logging.

Structure

ocr/
├─ SKILL.md               # this file
├─ HEARTBEAT.md           # optional periodic health check
├─ package.json           # metadata for ClawHub
├─ main.py                # optional entry‑point (can be omitted)
├─ ocr_core.py            # core OCR logic, importable by other agents
├─ ocr_engine.py          # wrapper for EasyOCR/Tesseract
└─ utils.py               # download / cleanup helpers

Setup & Deployment

1. Install system dependencies – Tesseract is required for the fallback engine:

   sudo apt-get update && sudo apt-get install -y tesseract-ocr libtesseract-dev   # Debian/Ubuntu
   # or on macOS: brew install tesseract
   

2. Create a virtual environment (recommended):

   cd /root/.openclaw/workspace/ocr
   python3 -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt
   

3. Start the skill – you can run the optional entry point if you need a simple CLI:

   python -m ocr.main
   

   Or import the core directly from other agents:

   from ocr.ocr_core import OCR
   text = OCR().process_image("/path/to/image.jpg")
   

4. Test – call the core function with any supported image file and verify the extracted text.