- Category
- AI Tools
- Rank
- No. 339Tools index
- Listed in
- #3 Capture meetings
- Pricing
- Open Source
- Type
- APP
- Builder
- basedhardware
- GitHub
- 13.5k stars
- Latest release
- v0.12.352+12352-macos
- Date
About
Omi is an open-source personal memory system built around a wearable pendant plus desktop, mobile, and web apps. It captures audio from conversations (and screen activity on desktop), transcribes it in real time, and generates summaries, action items, and a searchable history you can chat with. The hardware side — an Omi pendant and Omi Glass — is what separates it from software-only meeting recorders.
What it does
Omi is an open source personal knowledge system built around a timeline of conversations and computer activity. It turns incoming material into transcripts, summaries, action items, memories, and a searchable chat context. Users access that record through desktop, mobile, browser, and wearable clients. Developers can connect devices or add behavior through documented APIs, SDKs, apps, chat tools, and Model Context Protocol support.
Why it's ranked here
Its strongest case is unusually broad ownership of the stack. The repository covers client applications, cloud services, device firmware, protocol SDKs, extension points, and hardware documentation. It also exposes the resulting conversations, memories, and tasks through developer interfaces. That scope makes Omi compelling for builders who want an extensible memory platform, but it also creates substantial setup and operational complexity.
What's good
The project exposes useful outputs instead of stopping at raw transcription: summaries, memories, action items, and contextual chat. Its REST API covers conversations, memories, and action items. Device support spans shared Bluetooth framing plus Python, Swift, React Native, and several protocol languages. The repository also includes firmware, open hardware guidance, an app system, MCP integration, emulator tests, replay harnesses, and explicit backend recovery behavior.
Tradeoffs
The complete system depends on a large service graph, including Firebase, Redis, speech recognition, language models, WebSockets, and GPU-backed voice processing. The simplest macOS launch connects to Omi's cloud rather than running everything locally. Platform support is uneven: Ray-Ban Meta camera capture requires a special iOS build, physical hardware, and Meta developer access, while Android integration is unfinished. That build also disables pendant firmware updates because of a dependency collision.
How to use it well
Omi fits people who want meetings and daily conversations converted into durable notes, follow-ups, and searchable context. It also suits teams building memory-aware apps or integrating supported audio devices. Start with the hosted desktop or mobile path, then adopt the API, SDKs, or self-hosted backend only when customization justifies the extra machinery. It does not serve as a turnkey, self-contained offline recorder.
Technical notes+
backend/main.py assembles a large FastAPI application with routers for transcription, conversations, memories, action items, chat, MCP, integrations, desktop services, and account workflows. backend/pusher/main.py provides the WebSocket-side service with readiness and drain handling, while backend/modal/main.py exposes speaker identification and voice activity detection endpoints. backend/models/app.py defines extension metadata for webhooks, chat tools, MCP servers, OAuth tokens, capabilities, and app catalog records. plugins/main.py hosts bundled integration routes. package.json primarily exposes Firebase emulator and replay-oriented test commands. The Makefile provides baseline environment setup, preflight checks, local stack control, fixture management, and runtime image checks.
Observed
- License
- MIT
- Languages and stacks
- Python and FastAPI backend, Swift and SwiftUI macOS client, Flutter mobile client, and C firmware
- Install surface
- Source launch instructions cover macOS, Windows, iOS, Android, and a full local backend setup
- Interfaces
- REST API, WebSockets, device SDKs, app integrations, chat tools, and Model Context Protocol server support
- Platforms
- macOS, Windows, iOS, Android, browser, Omi wearables, Omi Glass, and iOS Ray-Ban Meta support
- Device protocols
- Bluetooth Low Energy framing is shared across TypeScript, Go, Rust, C++, and Dart protocol SDKs
- Testing structure
- The root package exposes Firebase emulator, replay harness, lifecycle, transaction, and parity test commands
Read from README.md, Makefile, package.json, docs/README.md, docs/rayban-meta-device.md, docs/rayban-meta-dat-setup.md, docs/rayban-meta-troubleshooting.md, docs/rayban-meta-founder-acceptance.md, backend/main.py, plugins/main.py, omiGlass/App.tsx, backend/modal/main.py, backend/models/app.py, backend/pusher/main.py, plugins/hume-ai/app.py.
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.