Home

NInfer Studio Wiki

A from-scratch, native Linux & Windows desktop app (Tauri 2 — WebKitGTK on Linux, WebView2 on Windows) for the NInfer local LLM inference engine — turning a raw ninfer-serve binary into a product.

NInfer Studio is the desktop shell: engine configuration with per-GPU presets, model management with Hugging Face downloads, streaming chat with vision, an agentic coding harness, and live token/throughput/VRAM metrics — with the control plane running in-process. The inference engine itself (ninfer-serve) is a separate project you point Studio at.

Pages

• Getting Started — prerequisites, build & run, dev modes
• Architecture — control plane, desktop shell, engine, lifecycle behaviors
• Screens — Chat, Engine, Models, Settings, Coder mode
• Configuration — settings, profiles, data directory
• Security — hardening notes, dependency status, reporting a vulnerability
• Packaging and Releases — CI builds, tagging a release
• Roadmap — planned work

At a glance

Frontend	React 19 + Vite + Tailwind 4 (apps/web)
Control plane	Rust ninfier-control (axum + tokio), desktop/control
Desktop shell	Tauri 2, ninfier-studio, desktop/app
Engine	ninfer-serve (C++/CUDA), a separate checkout you point Studio at

See the repo README and DESIGN.md for the full source-of-truth documents this wiki is derived from.