Show HN: Gerbil – an open source desktop app for running LLMs locally (github.com)

Gerbil is a new open-source, cross-platform desktop app that makes it easy to run Large Language Models locally by managing KoboldCpp (a heavily modified fork of llama.cpp) for you. It provides native installers for Windows, macOS and Linux (including Wayland), automatic KoboldCpp binary updates, smart process management to avoid runaway tasks, and built-in presets for image generation (Flux and Chroma). Integrations include one-click launches for SillyTavern (requires Node.js) and OpenWebUI (requires uv), and everything runs locally for privacy-minded users. The project is AGPL v3–licensed. Technically, Gerbil can act as a GUI wrapper or a CLI proxy to the downloaded KoboldCpp binary (gerbil --cli ...), passing all KoboldCpp args (model .gguf paths, --port, --gpulayers, --contextsize, --flashattention, --usemmap, --usevulkan, etc.), which enables custom frontends and lightweight operation. The Chromium-based UI is modest in resource use (~200MB RAM, ~100MB VRAM; CLI mode reduces that to about one-third). It supports Vulkan backend builds and has notes about Windows ROCm quirks (manual PATH changes often needed). Not all KoboldCpp features are yet exposed in the UI but can be supplied via “Additional arguments.” Linux users can install from the AUR; binaries are provided per platform on GitHub Releases. Overall, Gerbil lowers the barrier to local LLM experimentation while preserving flexibility for advanced users.