Mouse

AI-powered smart autocomplete for any browser text field

Getting started

1. Clone the repo
2. Go to chrome://extensions
3. Enable Developer mode
4. Click Load unpacked and select the repo root folder

Running the Project

Extension

1. npm install
2. npm run build / npm run dev for development

Backend

The backend is built with FastAPI and uses Amazon Bedrock (Nova for generation + Titan for embeddings).

1. Install dependencies:

   pip install -r backend/requirements.txt

2. Create a .env file with your configuration:

   AWS_ACCESS_KEY_ID=your_access_key_id
   AWS_SECRET_ACCESS_KEY=your_secret_access_key
   AWS_REGION=us-east-1
   
   # Optional overrides
   BEDROCK_GENERATION_MODEL=amazon.nova-lite-v1:0
   BEDROCK_EMBEDDING_MODEL=amazon.titan-embed-text-v2:0
   BEDROCK_EMBEDDING_DIM=1024
   
   AUTH_TOKEN=your_secure_auth_token_here

3. Start the development server:

   fastapi run backend/server.py

Debugging

Console output from content script is visible respective tab. Service worker can be inspected from chrome://extensions

Extension Architecture

src/background: Service worker, a process that runs tab-independent, interacts with the db and ai models
src/content: Runs in every tab, listens for DOM events & co, sends them to the service worker with chrome.runtime.sendMessage and then receives completion suggestions
src/popup: UI for configuring settings that opens when users click our extension icon

Extension is built with Vite: background and popup are compiled as ES modules, while content has a separate build config (vite.content.config.ts) and uses an IIFE bundle, because content scripts cannot use ES module imports. npm run dev runs two vite build --watch commands concurrently, so the output is a bit messy.