@chiragbuilds/glos-core

Core engine for glos.io - Give your i18n pipeline eyes.

Overview

This package provides the core functionality for capturing UI screenshots and extracting context using Vision LLMs. It includes:

• Playwright-based screenshot capture - Headless browser automation for capturing all routes
• Gemini Vision integration - AI-powered UI element extraction from screenshots
• Context file generation - Maps translation keys to their visual context
• Quality scoring - Analyzes translation quality based on UI context

Installation

npm install @chiragbuilds/glos-core

Usage

Capture Screenshots

import { captureApp } from '@chiragbuilds/glos-core';

const result = await captureApp({
  url: 'http://localhost:3000',
  outputDir: '.glos/screenshots',
  onProgress: (event) => {
    console.log('Progress:', event);
  }
});

Build Context File

import { buildContextFile } from '@chiragbuilds/glos-core';

const context = await buildContextFile(
  allElements,
  appUrl,
  localeMessages
);

Environment Variables

• GEMINI_API_KEY - Required: Gemini API key from Google AI Studio
• GEMINI_API_KEY_2 through GEMINI_API_KEY_5 - Optional fallback keys for rate limiting

Files Structure

dist/
├── capture.js      # Screenshot capture logic
├── vision.js       # Vision LLM analysis
├── context.js      # Context file builder
├── quality.js      # Quality scoring
├── translate.js    # Translation pipeline
├── types.js        # TypeScript type definitions
└── index.js        # Main entry point

License

MIT

Links

• npm: https://npmjs.com/package/@chiragbuilds/glos-core
• GitHub: https://github.com/chiragkhatri19/GLOS
• Dashboard: https://glos-dashboard.vercel.app