Install¶
pip¶
pip install "conch-triton[all]"
The package is published as conch-triton (the name triton is taken on PyPI by
NVIDIA's GPU compiler), but it is still imported as triton and the command is triton.
Optional extras keep the base install small. Pick what you need:
| Extra | Adds | Needed for |
|---|---|---|
transcribe |
openai-whisper |
Transcribe tab, triton transcribe |
classify |
torch, transformers |
Classify tab |
features |
torch, transformers, openai-whisper |
Features tab (word onsets, neural layers) |
all |
all of the above | everything |
Without an extra, the rest of Triton still works; using a feature whose extra is
missing shows the exact pip install command to run.
ffmpeg is a system program and must be installed separately
(brew install ffmpeg on macOS, sudo apt install ffmpeg on Debian/Ubuntu).
Pixi (for development)¶
Pixi locks every dependency, including ffmpeg, to ensure
reproducibility across macOS, Linux and Windows.
- Install Pixi: https://pixi.sh
- Clone Triton and
cdinto the repo - Create the environment (includes all extras, tests and docs):
bash pixi install
Verify Installation¶
triton --help # or: pixi run triton --help
triton gui # or: pixi run gui
The GUI opens in your browser (usually http://localhost:8501). To try it without your own
recordings, run triton examples download or use Add example files in the GUI.
The first use of Transcribe, Classify or neural Features downloads the model weights (~140 MB to ~1.5 GB depending on the model); later runs use the local cache.
Publishing a release (maintainers)¶
Releases are published to PyPI by .github/workflows/publish.yml when a version tag is pushed,
using PyPI trusted publishing (no API token stored in the repo). One-time setup:
- On pypi.org, create the
conch-tritonproject's trusted publisher: ownerAliTafakkor, repositorytriton, workflowpublish.yml, environmentpypi. - In the GitHub repo settings, create an environment named
pypi.
Then, for each release: bump version in pyproject.toml, commit, and run
git tag v0.1.0 && git push --tags.