GitHub - BasedHardware/omi: AI that sees your screen, listens to your conversations and tells you what to do
4 hours ago
- #wearable-technology
- #open-source-ai
- #real-time-transcription
- Omi is an open-source AI tool that captures screens and conversations, transcribes in real-time, generates summaries and action items, and provides an AI chat with memory across devices including desktop, phone, and wearables.
- The installation supports macOS 14+ with requirements like Xcode and Node.js; for local development, prerequisites include installing Xcode, Rust, cloning the repository, and configuring environment variables before building.
- The system architecture involves devices (wearable, macOS app, mobile app) connecting via BLE or HTTPS/WS to a backend with components for listening, WebSocket pushing, voice activity detection, diarization, and services like Deepgram for speech-to-text, Firestore, Redis, and LLMs.
- Key components and their tech stacks: macOS app (Swift, SwiftUI, Rust), mobile app (Flutter), backend API (Python, FastAPI, Firebase), firmware (nRF, Zephyr, C), Omi Glass (ESP32-S3), SDKs for various languages, and AI personas (Next.js).
- It includes development guides, example apps, API references, SDKs for Python, Swift, and React Native, MCP server integration, and open-source hardware options like Omi and Omi Glass Dev Kit with build and integration guides, all under the MIT license.