AI Platform

This repo contains the code for a AI platform with 2 environments:

1. dev: A development platform running on docker
2. prd: A production platform running on aws + k8s

Setup

1. Create + activate virtual env:

python3 -m venv .venv/dpenv
source .venv/dpenv/bin/activate

2. Install + init phidata:

pip install phidata
phi init -l

If you encounter errors, try updating pip using python -m pip install --upgrade pip

3. Setup workspace:

phi ws setup

4. Copy secrets:

cp -r workspace/example_secrets workspace/secrets

5. Run dev platform on docker:

phi ws up dev:docker

If something fails, run again with debug logs:

phi ws up -d

Optional: Create .env file:

cp example.env .env

Using the dev environment

The workspace/dev directory contains the code for the dev environment. Run it using:

phi ws up dev:docker

Run Airflow

1. Set dev_airflow_enabled=True in workspace/settings.py and run phi ws up dev:docker
2. Check out the airflow webserver running in the airflow-ws-container:

• url: http://localhost:8310/
• user: admin
• pass: admin

Run Jupyter

1. Set dev_jupyter_enabled=True in workspace/settings.py and run phi ws up dev:docker
2. Check out jupyterlab running in the jupyter-container:

• url: http://localhost:8888/
• pass: admin

Validate workspace

Validate the workspace using: ./scripts/validate.sh

This will:

1. Format using black
2. Type check using mypy
3. Test using pytest
4. Lint using ruff

./scripts/validate.sh

If you need to install packages, run:

pip install black mypy pytest ruff

Install workspace

Install the workspace & python packages in the virtual env using:

./scripts/install.sh

This will:

1. Install python packages from requirements.txt
2. Install the workspace in --editable mode

Add python packages

Following PEP-631, add dependencies to the pyproject.toml file.

To add a new package:

1. Add the module to the pyproject.toml file.
2. Run: ./scripts/upgrade.sh to update the requirements.txt file.
3. Run phi ws up dev:docker -f to recreate images + containers

Add airflow providers

Airflow requirements are stored in the workspace/dev/airflow/resources/requirements-airflow.txt file.

To add new airflow providers:

1. Add the module to the workspace/dev/airflow/resources/requirements-airflow.txt file.
2. Run phi ws up -f --name airflow to recreate images + containers

Stop workspace

phi ws down

Restart workspace

phi ws restart

Add environment/secret variables

The containers read env using the env_file and secrets using the secrets_file params which by default point to files in the workspace/env and workspace/secrets directories.

Airflow

To add env variables to your airflow containers:

1. Update the workspace/env/dev_airflow_env.yml file.
2. Restart all airflow containers using: phi ws restart dev:docker:airflow

To add secret variables to your airflow containers:

1. Update the workspace/secrets/dev_airflow_secrets.yml file.
2. Restart all airflow containers using: phi ws restart dev:docker:airflow

Test a DAG

# ssh into airflow-worker | airflow-ws
docker exec -it airflow-ws-container zsh
docker exec -it airflow-worker-container zsh

# Test run the DAGs using module name
python -m workflow.dir.file

# Test run the DAG file
python /mnt/workspaces/data-platform/workflow/dir/file.py

# List DAGs
airflow dags list

# List tasks in DAG
airflow tasks list \
  -S /mnt/workspaces/data-platform/workflow/dir/file.py \
  -t dag_name

# Test airflow task
airflow tasks test dag_name task_name 2022-07-01