Onde CLI

Fine-tune local models from your terminal.

Manage your Onde account, fine-tune local models, and export GGUF with onde.

Download v0.3.1 Read the README

Current release: v0.3.1

Terminal UI

Open `onde`, sign in, fine-tune a model, export GGUF.

Fine-tune local models

Train supported Qwen models locally with the built-in LoRA flow.

Export to GGUF

Export a merged model to GGUF when you are ready to test or ship it.

Install it your way

Install `onde` with npm, PyPI, NuGet, pub.dev, crates.io, GitHub Releases, or Homebrew.

Install it your way.

Pick a package manager, install `onde`, done. If you want to see the full local workflow first, read Fine-Tune a Language Model on Your Mac.

npm

@ondeinference/cli

npm install -g @ondeinference/cli

Install from npm.

PyPI

onde-cli

pip install onde-cli
# or
uv tool install onde-cli
# or
uvx --from onde-cli onde

Install through pip, uv, or uvx.

NuGet

Onde.Cli

dotnet tool install --global Onde.Cli

Install as a .NET global tool.

pub.dev

onde_cli

pub.dev

dart pub global activate onde_cli

Install from Dart pub.

crates.io

onde-cli

crates.io

cargo install onde-cli

Install direct from crates.io.

Homebrew

onde

brew tap ondeinference/homebrew-tap
brew install onde

Install with Homebrew on macOS.

GitHub Releases

onde-cli

# macOS Apple Silicon
curl -Lo onde https://github.com/ondeinference/onde-cli/releases/latest/download/onde-macos-arm64
chmod +x onde
mv onde /usr/local/bin/onde

Download a prebuilt binary.

From sign-in to GGUF.

Open the app, run the tune, merge, export. If the model is headed into app code next, the Swift SDK, Dart SDK, and React Native SDKpages cover the shipping side.

1. Open the terminal UI

Run `onde` to open the terminal UI.

onde

2. Start a fine-tune

Pick a safetensors model and press `f` to start a local LoRA run.

Models tab → select a safetensors model → press f

3. Merge the adapter

Merge the adapter back into the base model.

Fine-tune complete → press m

4. Export GGUF

Export GGUF for local inference or packaging.

Merged model → press g

What you get.

Terminal UI, local fine-tuning, GGUF export, broad install support. If you are evaluating Onde as a product direction, start with What Onde Is and Privacy as a Feature, Not a Compliance Check.

Account

Terminal-first

Open the app and work from the terminal. No browser detour.

Local fine-tuning

Keep training data, model files, and logs in one place.

GGUF export

Merge the adapter, export GGUF, and test the result.

Install anywhere

Use the package manager you already have.

Native

Fast startup. Direct install. One `onde` command.