Poor man's linux voice automation.
- Wake phrase "Tempest rise" and sleep phrase "Tempest rest"
- Shortcut customization using config files
- Recording built into the binary
- Sending fuzzy questions to an LLM through Ollama's API
- Built-in understanding of fuzzy terms (similar sentences are treated as equal)
- Static builds
- Rust toolchain, either via your package manager or rustup
- C++ standard library
clangcmakepkg-configlibvoskinLD_LIBRARY_PATH- Audio library for your OS (linux distros have package names like
alsa-dev,alsa-liboralsa)
If you use Nix flakes, run nix develop in the project directory to get a dev shell with the dependencies installed.
git clone https://github.com/lavafroth/tempest
cd tempest
Change the bindings in the config file to suit your needs.
Run the following to build the daemon and the client:
cargo build --workspace --releaseThe daemon is optional and is only needed if you want phrases in your bindings to perform keyboard shortcuts. Since performing keystrokes is a privileged action, you must run the daemon as root.
sudo ./target/release/tempest-daemonThis will give a token to authenticate with the daemon.
If you have the daemon running in the background, in a different terminal tab, run
./target/release/tempest-client \
the_token_from_the_daemonwhere the_token_from_the_daemon is the token provided by the daemon.
You can alternatively run the client as standalone. However, config bindings with keyboard shortcuts will not work.
./target/release/tempest-clientOn first run, the client will prompt you to download models.