NeuronBridge Python API

A Python API for the NeuronBridge neuron similarity search service.

See this notebook for usage examples.

Development Notes

Create a conda env with all the dependencies including Jupyter:

conda env create -f environment.yml
conda activate neuronbridge-python

Then install it as a Jupyter kernel:

python -m ipykernel install --user --name=neuronbridge-python

Install for development

You can install the module for development like this:

conda create --name neuronbridge-python python=3.8
conda activate neuronbridge-python
pip install -e .

Useful shell commands

To update conda_requirements.txt:

conda env export --from-history --file conda_requirements.txt

To update requirements.txt:

pipreqs --savepath=requirements.txt && pip-compile

Regenerate the JSON schemas:

python neuronbridge/generate_schemas.py

Run the unit tests:

pytest tests

Publishing a new release

1. Update the version in setup.py

2. Push all changes and tag a release in GitHub

3. Build PyPI distribution:

   python setup.py sdist bdist_wheel

4. Upload to PyPI:

   twine upload dist/*

Running validation using Ray

You can run validation multithreaded on a single machine like this:

./neuronbridge/validate_ray.py

To run the validation script in a distributed manner on the Janelia cluster, you must first install ray-janelia in a sister directory to where this code base is cloned. Then run a script to bsub the Ray cluster:

./scripts/launch_validation.sh