Installation#
sktime-cli runs on Python 3.10 to 3.14 and installs sktime as a dependency.
Install the CLI#
To install sktime-cli as a standalone tool with its own isolated
environment, use uv:
uv tool install sktime-cli
To install it into the active environment instead, use pip:
pip install sktime-cli
Both commands provide one executable, sktime-cli. There is no short
sktime alias, which keeps the command from colliding with the library.
Check the installation#
doctor reports whether sktime imports, whether the cache directory is
writable, and which optional dependencies are available:
sktime-cli doctor
The output is similar to the following:
doctor exits with 0 even when optional dependencies are missing. It exits
with 1 only when sktime itself fails to import.
Optional dependencies#
sktime ships most of its estimators with optional dependencies, and the CLI adds one extra of its own. Nothing is installed by default.
parquetAdds pyarrow so that
data convert,data inspect, anddatasets loadcan read and write parquet files.uv tool install "sktime-cli[parquet]"
- Estimator dependencies
Estimators such as
AutoARIMAandProphetneed packages that sktime doesn’t install. When a command needs one, it exits with code3and a hint that names the install command. To see what an estimator needs before you run it, useregistry describe NAME, or filter searches down to what you can already run withregistry search --installable-only.
Install from source#
To work on the CLI, clone the repository and sync the development environment:
git clone https://github.com/siddharth7113/sktime-cli
cd sktime-cli
uv sync
This resolves sktime from PyPI, like any other dependency.
For the full development workflow, see Contributing.