Installation

Echodataflow is available and tested for Python 3.13-3.14. We recommend using a Conda environment because of its scientific and geospatial dependencies.

Create a Conda environment¶

Create and activate an environment:

conda create --name echodataflow -c conda-forge python=3.13 uv
conda activate echodataflow

If your system does not already have rclone, include it when creating the environment by using the command below, then activate the environment:

conda create --name echodataflow -c conda-forge python=3.13 uv rclone
conda activate echodataflow

Once done, verify that rclone is installed by running:

rclone version

Install Echodataflow¶

Echodataflow v0.2.0 is currently under development and has not been released, and therefore cannot be installed directly from PyPI or conda yet.

Instal as an installed package¶

To run Echodataflow as an installed package, install it directly from the repository:

uv pip install "echodataflow[dev,mission,inference] @ git+https://github.com/echostack-org/echodataflow.git"

Install for development¶

To install Echodataflow for development, clone the repository and install from the local repository where you will be making code changes:

# clone the repo
git clone https://github.com/echostack-org/echodataflow.git

# switch into the repo directory
cd echodataflow

# install in editable mode from the local repo
uv pip install -e ".[dev,mission,inference]"

Adding [dev,mission,inference] in the uv pip install command above install the full suite of dependencies specified in pyproject.toml:

• dev: testing, linting, code quality, documentation tools, and Jupyter kernel support.

• mission: mission-specific dependencies for the example pipelines.

• inference: the hake segmentation model, needed for running the classify_backscatter flow.

Testing the installation¶

After installation, verify that the package is correctly installed by running:

python -c "import echodataflow; print(echodataflow.__version__)"

The installation enables the echodataflow-deploy command, which can be run from any directory. Testing that this works by running

echodataflow-deploy --help

Get the deployment recipes¶

Example recipes are maintained separately on a companion repository echodataflow-recipes. You can download the recipes directly from the repo or get a copy by cloning it:

git clone https://github.com/echostack-org/echodataflow-recipes.git