Test-Driven Development with FastAPI and Docker

Table of Contents

• Introduction
• Prerequisites
• Installation
• Usage
• Testing
• Deployment
• License

Introduction

This project is an implementation of the course on Test-Driven Development to build the API for summarizing articles, with additional features and optimizations. Key enhancements include:

• Dependencies updated to the latest version at the moment
• CORSMiddleware used to manually control allowed origins
• Venv replaced with PDM
• Added versions of the API
• Refactored the code
• Gunicorn added to manage the uvicorn
• BackgroundTasks replaced with Celery, Redis and RabbitMQ
• Optimized CI/CD pipeline in GitHub Actions
• Migrated to the Container registry from the Docker registry
• Implemented authentication and authorization using OAuth2
• Tortoise-ORM has been replaced by SQLAlchemy
• Transformers models are used instead of the NLP from Newspaper3k

Prerequisites

• Docker and Docker Compose installed on your machine.
• Basic knowledge of Python, FastAPI, and Docker.

Installation

1. Clone the repository:

git clone https://github.com/spyker77/fastapi-tdd-docker.git
cd fastapi-tdd-docker

2. Build and start the containers:

docker compose up -d --build

3. Generate the database schema:

docker compose exec web alembic upgrade head

Usage

1. Access the API documentation at: http://localhost:8000/docs

2. Create a test user at: http://localhost:8000/docs#/users/create_user_api_v2_users__post

   Example of the payload:

{
  "full_name": "Cute Koala",
  "username": "cute",
  "email": "[email protected]",
  "password": "supersecret"
}

3. Use the Authorize button (simplest way) at the top and enter the username and password you've just created, then click Authorize.

4. At this point you're authorized. Now use the endpoint http://localhost:8000/docs#/summaries/create_summary_api_v2_summaries__post to send the article you want to summarize, for example like so:

{
  "url": "https://dev.to/spyker77/how-to-connect-godaddy-domain-with-heroku-and-cloudflare-mdh"
}

5. This triggers the ML models to download, which may take a few minutes for the first run (in the current implementation). After that, reach the endpoint http://localhost:8000/docs#/summaries/read_all_summaries_api_v2_summaries__get and in the response you should see something like:

[
  {
    "id": 1,
    "url": "https://dev.to/spyker77/how-to-connect-godaddy-domain-with-heroku-and-cloudflare-mdh",
    "summary": "If you struggle to connect newly registered domain from GoDaddy with your app at Heroku, and in addition would like to use advantages of Cloudflare – this article is for you. Hope it will help you and without many words, let's jump in!Sections: Heroku settings, Cloudflare settings and GoDaddy settings.",
    "user_id": 1
  }
]

Testing

Run the tests using the following command:

docker compose exec web pytest -n auto --cov

Deployment

For production deployment, don't forget to change the ENVIRONMENT variables. The default CI/CD pipeline is set up to build images with GitHub Actions, store them in GitHub Packages, and deploy the application to Heroku. But the deployment part is currently disabled/commented out.

Note

The current implementation of text summarization using the transformer models is not ideal for production due to the following reasons:

• The requirement to install a heavy transformers library along with its dependencies.
• The necessity to download several gigabytes of the model.
• The need for powerful hardware to run the model.

Typically, in a production environment, the models would be provided via an API using services like AWS SageMaker or Paperspace Gradient.

License

This project is licensed under the MIT License. See the MIT License file for details.