A voice research companion. Right beside what you're reading.
Ask aloud. Share a screenshot. Check the sources.

Watch the demo ↗  ·  Quick start  ·  Project story

---

Something sounds off? Ask Milo. It searches public sources, talks through what it finds, and gives you links to check for yourself.

• Ask naturally. Voice or text, with follow-up questions.
• Show the claim. Capture a screen, or drop, paste or upload an image.
• Keep your place. Float Milo beside the page in supported desktop browsers, or use it in a tab.

No installation. No continuous screen watching.

Quick start

Requires Node.js 22.13+ and an OpenAI API key with access to gpt-live-1 and gpt-5.6-terra. API usage is billed to your project.

git clone https://github.com/Yun-0000/Milo.git
cd Milo
npm ci
cp .env.example .env

Set OPENAI_API_KEY in .env, then run:

npm run dev

Open localhost:5173 → Talk to Milo. The speech bubble switches to typing; the camera adds an image.

Self-hosting & development

npm test
npm run typecheck
npm run build
npm start

The production server runs at localhost:3000. Deployment options: Docker · Render.

Set OPENAI_API_KEY and a strong MILO_ACCESS_CODE as hosting secrets. Use HTTPS. The server uses in-memory storage and is designed for a single-instance preview.

React · TypeScript · Express · WebRTC · OpenAI

A note on privacy & accuracy

API keys stay server-side. Voice, questions and requested images are processed by OpenAI—avoid sensitive material. Screen sharing stops after capture; server-held results are temporary.

Milo can misread images or get answers wrong. Sources are there for you to verify, not a guarantee of accuracy.

---

_{License  ·  Report an issue}