DOCUMENTATION / Developers

Build from source

Run Noullama in development or build a standalone executable with the provided toolchain.

Install the toolchain

Source builds require Node.js 24+, Rust and Blaze for the provided build commands. On Linux, install the Tauri prerequisites, including GTK 3 and WebKitGTK 4.1 development packages.

You still need the Hugging Face CLI for downloads and a compatible llama-server for inference. These are runtime prerequisites as well as source-build requirements.

Run in development

terminal

git clone https://github.com/parasite6/noullama.git
cd noullama
npm install
npm run desktop

Once the app opens, follow the first model setup. Development builds use a separate target directory.

Build a standalone executable

terminal

npm run desktop:build             # release
npm run desktop:build -- --debug  # debug

Linux outputs are src-tauri/target/release/noullama and src-tauri/target/debug/noullama. Standalone builds include the frontend and do not require a Vite server.

Run the checks

terminal

npm run build
npm test
blaze test --manifest-path src-tauri/Cargo.toml

Optional real-model integration test

Set NOULLAMA_TEST_MODEL and NOULLAMA_TEST_RUNTIME to absolute paths for your GGUF and llama-server. Set NOULLAMA_TEST_GPU_LAYERS=99 if you want to test GPU offloading.

terminal

blaze test --manifest-path src-tauri/Cargo.toml \
  real_runtime_native_and_openai_requests -- --ignored --nocapture