First public release.
LLMFuse watches memory across every local LLM provider you run — Ollama, oMLX, LM Studio, llama-swap
and koboldcpp — and lets you eject a model from any of them in one place. Each provider manages its
own memory and is blind to the others; LLMFuse is the only thing with a whole-system view.
Highlights
- Menu-bar popup: memory as a stacked bar with one colour-blind-safe segment per resident model, the
full list across every provider largest-first, and one-click eject. - Zero configuration. Each provider's port and API key are detected continuously from its settings
file, launch arguments, environment, or the kernel's socket table — so a provider that restarts on
a different port is followed automatically. The one exception is LM Studio's API key, which it
stores only as a hash. - Sizes are measured, not estimated. Every provider was verified against actual memory consumption:
oMLX +0.2%, Ollama −0.8%, koboldcpp −3.1%, llama-swap −5.6%, LM Studio −8.4%.
Requirements: macOS 15 or later. Signed and notarized.