📺 Lower That

Mute/unmute your Samsung TV based on what your webcam sees. The web app captures an image every 5 seconds, the server analyzes it with an OpenAI vision model, and the server decides whether to toggle mute.

Use Docker Compose to run both services: the Express API and the web app.

Quick start:

# from repo root
# Use the provided example to create your Compose env file
cp env.example .env

# Edit .env and set your values (at minimum OPENAI_API_KEY).
# If enabling TV control, set TV_CONTROL_ENABLED=true and provide
# SAMSUNG_TV_IP_ADDRESS and SAMSUNG_TV_MAC_ADDRESS.

docker compose up -d --build

Once containers are running, access the app at http://localhost:5173

• Web client (Vite + React + Tailwind + MUI) shows the webcam, lets you set a preferred content description, and streams snapshots to the server.
• Server (Express) exposes POST /api/analyze; it calls the OpenAI API and controls your TV.
• Testing is set up end-to-end: Vitest + RTL + MSW on the web, and Vitest + Supertest on the server.

• server/: Express API
  • src/server.js: entry point
  • src/http/app.js: Express app wiring
  • src/http/routes/analyze.js: POST /api/analyze (multipart or JSON base64)
  • src/services/visionAnalysisService.js: OpenAI integration
  • src/services/tvRemoteService.js: Samsung TV control
  • src/services/tvRemoteNoopService.js: no-op TV control (for dry runs)
  • src/config/environment.js: env parsing/validation
  • src/utils/logger.js: structured logging
• web/: Vite React app
  • src/App.tsx: single-page UI (webcam, input, last image, results log)
  • src/api/client.ts: client for /api/analyze
  • vite.config.ts: dev proxy /api → target from VITE_API_PROXY_TARGET (defaults to http://localhost:3000)

cd server
cp env.example .env

Edit .env with your values:

OPENAI_API_KEY=your_openai_api_key
SAMSUNG_TV_IP_ADDRESS=your_tv_ip
SAMSUNG_TV_MAC_ADDRESS=your_tv_mac

# Optional: disable real TV control (dry run)
TV_CONTROL_ENABLED=true

# Optional: logging level (info, debug, warn, error)
LOG_LEVEL=info

Notes:

• Samsung IP/MAC are required when TV_CONTROL_ENABLED=true.
• With TV_CONTROL_ENABLED=false, the server analyzes but does not send TV commands.

Install dependencies:

cd server
npm install

Run the API:

npm start
# http://localhost:3000

Install deps and run dev server:

cd web
npm install
npm run dev
# http://localhost:5173 (proxies /api to :3000)

The page:

• Shows live webcam
• Lets you set your preferred content description
• Captures a JPEG every 5s and posts to /api/analyze
• Displays the last capture and a rolling results log

• Server: cd server && npm test (Vitest + Supertest; coverage enabled)
• Web (unit): cd web && npm test (Vitest + RTL + MSW; coverage enabled)
• Web (e2e): cd web && npx playwright install && npm run e2e

• OpenAI calls incur cost; monitor usage.
• Webcam captures include your environment; handle artifacts and logs securely.
• The server applies mute logic on every analysis request — the client does not need to manage TV state.

PRs welcome! Please include:

• Rationale and scope of changes
• Testing steps and logs (use LOG_LEVEL=debug where helpful)
• Any env var changes (update env.example)