Finhealth / BD Plaid Pilot

This repo contains our pilot Django application intended to collect and annotate two years of retrospective financial data.

Local deployment

First, clone our repo using git clone https://github.com/brozena/finhealth-pilot.git. The sqlite branch is configured to use Plaid's Sandbox and sqlite3 (rather than postgresql) for easier local deployment. After cloning, navigate to the project's folder using cd finhealth-pilot and then use git checkout -b sqlite origin/sqlite before proceeding.

We'll tend to stick to feature branch workflows.

pipx & poetry

We use pipx and poetry for package/environment management. On a Mac, the easiest way to install pipx is using Homebrew,

brew install pipx
pipx ensurepath

If you do not use Homebrew, go ahead and follow the installation instructions for your platform.

Poetry can then be installed using pipx install poetry. Then run poetry install from within the project's root directory to get dependencies installed as specified in pyproject.toml.

A note on virtual environments: By default, poetry install will create a Python virtual environment according to Poetry's configuration. Internally, we set virtualenvs.path to a specific directory and then set virtualenvs.in-project to false. We then configure mod_wsgi to make use of (and isolate to) Poetry's virtual environment, which is necessary for the asynchronous Plaid calls.

Environmental variables

Environmental variables required by this project are listed in finhealth/.env.example. Copy that file to finhealth/.env and enter a Django secret of your choosing and your Plaid API keys. These can be found in the Plaid dashboard.

.env is listed in .gitignore; do not hardcode API keys or other authentication info. Note that the sqlite branch does not require postgres authentication details to run.

Handling Plaid calls

This repo makes use of celery (installed via Poetry) and rabbitmq-server (installed manually, as a service) to handle the Plaid calls. If you want to test calls to Plaid (using either Sandbox or Production modes), you can get rabbitmq-server installed and then run poetry run celery -A finhealth worker -l debug.

In our case, Apache is configured to use Poetry's venv through mod_wsgi by setting WSGIPythonPath and WSGIDaemonProcess python-home within the httpd.conf virtualhost definition. Also, since we don't host this prototype on the public internet, we use a ngrok agent to establish a tunnel with a selector pointing at /webhook_transactions. This lets us receive Plaid's HISTORICAL_UPDATE webhook when transactions are ready.

Starting Django

So long as pipx/poetry are installed correctly (i.e., no issues with pipx ensurepath or the Poetry venv), you should be able to get a Django instance up and running in a shell using the following commands:

poetry run python3 manage.py makemigrations
poetry run python3 manage.py migrate
poetry run python3 manage.py migrate --run-syncdb
poetry run python3 manage.py runserver

You'll also want to edit finhealth/settings.py to adjust ALLOWED_HOSTS to fit your needs.

Email me if you have any questions or comments!