Lyra is built as a small constellation of its own: a backend that thinks, a client that you talk to, and packaging that gets it onto your system in the first place.
The brain
Backend daemon for Lyra. Exposes a Unix socket for the CLI and desktop UI, plus an HTTP API with streaming for external integrations.
- Runs a local Qwen2.5-1.5B model via llama-cpp-python
- Remembers past conversations with a semantic memory store
- Resolves packages across pacman, apt, PyPI, npm, crates.io
Stack Python 3.13 · FastAPI · llama-cpp-python · SQLite · ChromaDB
The face
The Electron desktop client. Talks to lyra-server over a Unix socket, renders responses as markdown, and keeps an eye on your system.
- Renders responses as markdown with syntax highlighting
- Shows live RAM, CPU, and disk usage
- Starts lyra-server automatically if it isn't already running
Stack Electron · Node.js 20 · pnpm · markdown-it · Chart.js
The delivery
The packaging and release pipeline that turns lyra-server and lyra-ui into installable packages for your distribution.
- Builds .pkg.tar.zst archives for Arch Linux
- Builds .deb archives for Debian and Ubuntu
- Handles CI/CD and release announcements
Stack Shell · Python · GitHub Actions