Installation

Add llama-crab to a Rust project when you want to load GGUF models directly from application code.

[dependencies]
llama-crab = "0.1.8"

The crate defaults to ["openmp", "metal"]. Pin the version you want explicitly; the workspace tracks 0.1.8 at the time of writing.

Native requirements

llama-crab-sys builds the native llama.cpp stack. Install the platform C++ toolchain and CMake before the first cargo build. MSRV is 1.88 — the toolchain is pinned in rust-toolchain.toml at the repository root.

Feature selection

Be explicit about compute backends in applications:

[dependencies]
llama-crab = { version = "0.1.8", default-features = false, features = ["openmp"] }

Common combinations are listed in Backends; the full current feature matrix lives in Cargo features.

Server and Tauri plugin

Both llama-crab-server and tauri-plugin-llama-crab are published separately. Install the server binary when you want a local OpenAI-compatible HTTP API:

cargo install llama-crab-server --features mtmd --force

The Tauri plugin is added to a Tauri v2 application the same way as any other tauri-plugin-* crate (see Tauri plugin). Its mtmd feature enables multimodal (vision) inference and is enabled by default in 0.1.8 alongside llama-crab's hf-hub feature so Tauri apps can load Hugging Face model IDs through load_model without extra feature wiring.