misty_dc

Deploy ML with Django and Docker, I think it's kinda overkill, but a good learning project This is a more fancy setup, if you just deploy one model (not gonna do anything about, refer to misty_dc_light

1. Create venv and install Django

virtualenv env
env\Scripts\activate.bat
pip3 install django

2. Start project

cd backend
django-admin startproject mysite
cd mysite folder
pip3 install djangrestframework
pip3 install markdown
pip3 install django-filter

4. Start endpoints in the project

python3 manage.py startapp endpoints

5. Create apps, put appname in it
6. Change apps.py such that it knows we have moved the endpoints directory into apps

class EndpointsConfig(AppConfig):
    default_auto_field = 'django.db.models.BigAutoField'
    name = 'apps.endpoints'

7. Edit apps/endpoints/models.py
8. Good time to migrate the fields into the db.

python3 manage.py makemigrations
python3 manage.py migrate

10. Add apps/endpoints/serializers.py
11. Edit apps/endpoints/views.py
12. Add the links to urls
13. Edit settings

  INSTALLED_APPS = [
      'django.contrib.admin',
      'django.contrib.auth',
      'django.contrib.contenttypes',
      'django.contrib.sessions',
      'django.contrib.messages',
      'django.contrib.staticfiles',
      'rest_framework',
      'apps.endpoints',
      'apps.ml'] 
  ROOT_URLCONF = 'misty_dc.urls'
  SECRET_KEY = 'django-insecure-' # delete just for security reason

12. Edit asgi.py

  os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'misty_dc.settings')

13. Edit urls.py

  from apps.endpoints.urls import urlpatterns as endpoints_urlpatterns
  urlpatterns += endpoints_urlpatterns

14. Make registry to add connect the ML to server side (the worst nightmare)
15. use wsgi.py to add MLAlgorithm