lilbee-cuda
by tobocop2
The whole local AI stack in one executable (NVIDIA CUDA build)
Support the creators
Flatsonar is just the shop window. The people below made lilbee-cuda.
About
lilbee is the whole local AI stack in one executable: it runs and manages local AI models across every GPU you have, indexes your files and code, crawls the web pages you point it at, and answers questions with citations down to the file and line. One install covers the TUI, CLI, HTTP server, and MCP server. You don't need Ollama or LM Studio, and it works with both. This build offloads inference to an NVIDIA GPU via CUDA. It needs a recent NVIDIA proprietary driver on the host; without a usable GPU it falls back to the CPU. Run it with GPU access enabled (flatpak --device=all).
Permissions
2 permissions weaken the sandbox. Flatsonar asks before installing.
- full access to your entire file system
--filesystem=host - all devices (/dev): webcams, disks, raw hardware
--device=all - network
--share=network
Who publishes this
The creator demonstrably controls this app id: Flathub verification, the hosting account owns the namespace, or a well-known file on their domain.
- downloads github.com at install time, after any scan has run
source:extra-data - installs a prebuilt binary instead of building from source
source:prebuilt - hosted by github.com/tobocop2, which owns the io.github namespace of io.github.tobocop2.lilbee.cuda
publisher:namespace
Maintenance
Recent activity on the upstream repository, or built and reviewed by Flathub.
No staleness signals: recent activity, or built and reviewed by Flathub.