What it does Omi is a life-logging system that captures your screen activity and ambient conversations, transcribes speech in real-time, and feeds everything into an AI chat that claims to remember what you’ve seen and heard. It spans a macOS app (Swift/Rust), a Flutter mobile app, a Python/FastAPI backend, and two pieces of open-source hardware: a small wearable and smart glasses built on ESP32-S3.

The interesting bit Most “AI memory” projects stop at the browser extension. Omi goes the other direction: it sells actual hardware with open schematics, runs BLE + WebSocket pipelines for continuous audio streaming, and exposes the whole stack—from nRF/Zephyr firmware to a Next.js persona marketplace—for modification. The “–yolo” quick-start flag that builds and launches the macOS app without credential setup is a small but telling UX choice.

Key highlights

• Full capture stack: screen + audio + conversation transcription with speaker diarization
• Cross-platform: macOS app, Flutter iOS/Android, browser version, and dedicated wearables
• Backend uses Deepgram for STT, GPU-based VAD/diarization, Firestore, Redis, and pluggable LLMs
• Hardware is open-source and purchasable: Omi wearable (24h+ battery) and Omi Glass dev kit with camera
• SDKs for Python, Swift, React Native, plus an MCP server for Model Context Protocol integration
• MIT licensed throughout

Caveats

• The “300,000+ professionals” trust claim is stated without sourcing or methodology
• macOS quick-start requires Xcode and code signing; full local backend needs Rust, Python, and Firebase credentials
• The hardware ecosystem is ambitious but adds manufacturing and supply-chain complexity most software projects avoid

Verdict Worth a look if you’re building ambient intelligence, testing wearable form factors, or want a reference architecture for continuous audio+screen pipelines. Skip it if you just need a meeting summarizer—there are lighter tools for that.