ocr

The OCR skill provides a robust interface for performing Optical Character Recognition (OCR) on various image formats including PNG, JPG, JPEG, GIF, BMP, TIFF, and WEBP. Built upon the powerful Tesseract OCR engine, this skill is designed for developers and autonomous agents that need to convert scanned documents, screenshots, or photographic text into machine-readable formats. It serves as an essential tool for data extraction workflows, document digitization, and automated information retrieval from visual sources.

• Full support for over 100 languages, including English, Chinese (Simplified/Traditional), Japanese, Korean, French, German, Spanish, Russian, and Arabic.

• Integrated image preprocessing capabilities such as grayscale conversion and thresholding, which significantly improve text recognition accuracy in challenging visual conditions.

• Flexible output options allowing data to be retrieved as plain text or as a structured JSON object, providing both extracted content and confidence scores.

• Capability to handle both local file paths and remote image URLs, making it highly versatile for web-scraping or agentic research tasks.

• Integration ready for trpc-agent-go, enabling agents to leverage visual data processing as part of their decision-making or data-analysis pipelines.

• Requires Tesseract OCR installed on the host system, along with Python 3.8+, pytesseract, and Pillow.

• For optimal accuracy, consider enabling the --preprocess flag on images with low contrast or noisy backgrounds.

• Supports combining language codes (e.g., eng+chi_sim) to perform multi-language OCR in a single pass, ideal for documents containing mixed character sets.

• Use the JSON output format when building downstream applications that require programmatic confidence verification or segment-based data analysis.

• The tool is designed to be invoked by agent runtimes, making it easy to incorporate into larger orchestration workflows where visual input must be interpreted and acted upon.