README.md

Visibabel extension

Chrome extension to capture images, send them to a local Ollama (GLM-OCR) endpoint, and return translated text.

Source of truth

• TypeScript files in src/**/*.ts are the canonical source.
• JavaScript files under dist/ are runtime artifacts emitted by npm run build:extension.
• Re-run npm run build:extension after TypeScript changes before loading the unpacked extension.

API reference

• ../docs/ENDPOINT_API_REFERENCE.md

Features

• Capture visible tab screenshot
• Select region on page
• Right-click image to OCR/translate
• Configurable endpoint, prompt, and settings
• All processing local (no cloud)

Setup

1. Install Ollama and pull the model:

   ollama pull glm-ocr:latest

2. Start Ollama (see ../ollama/README.md).
3. Install dependencies and build:

   cd extension
   npm install
   npm run build:extension

4. Load this extension unpacked in Chrome from the extension/ folder.
5. Set endpoint URL in options (default: http://localhost:11434/).
6. Optionally start the GLM-OCR layout service (see ../glm-ocr/README.md).

Development commands

npm run build:extension
npm run verify:emit-sync
npm run verify:browser-esm-imports
npm run test:unit
npm run test:e2e:runtime

From repository root:

npm --prefix ./extension run test:unit

Troubleshooting

• If the extension cannot connect, ensure Ollama is running at the configured URL.
• Check the options page for diagnostics and logs.