Deployment

Deploying an Lilya application is relatively easy.

What is a deployment

Deploying an application means to perform the necessary steps to make your application available to others to use outside of your local machine and/or development environment.

Normally, deploying web APIs involves putting your code in remote machines with all the necessary requirements from memeory, CPU, storage to things like networking and all of that. It will depend on your needs.

Strategies

There are many ways of deploying an application. Every case is unique and it will depends on a lot of factors that sometimes is not even related with the application itself. For example, funds.

You could want to save money not going to cloud but that also means more personal maintenance of the infrastructure.

You could also decide to go cloud and use an external provider such as AWS, Azure, GCP or even one that is very good and also affordable like render.com or Heroku. It is your choice really since it will depend on your needs.

The goal is not to tall you what to do but to give you a simple example in the case you would like to use, for example, docker and the reason why it is very simple. Every case is unique.

Lilya

We decided that we did not want to interfere with the way the people do deployments neither suggest that there is only one way of doing it but we thought that would be very useful to have at least one example just to help out a bit and to unblock some potential ideas.

We opted for using a standard, docker.

Deploying using Pydantic

Pydantic is fantastic handling with majority of the heavy lifting when it comes to read environment variables and assigning but there are some tricks to have in mind.

Loading List, dicts and complex types

When loading those into your environment variables it is imperative that you understand that Pydantic reads them as a JSON like object.

Example:

export ALLOWED_HOSTS="https://www.example.com,https://www.foobar.com"

There are many ways of doing this but in the documentation of Pydantic (even a fix), they recommend to use the parse_env and handle the parsing there.

from typing import ClassVar

from dataclasses import dataclass, field

from lilya.conf.global_settings import Settings


@dataclass
class AppSettings(Settings):
    allowed_hosts: list[str] | str = field(default=os.environ.get("ALLOWED_HOSTS", []))

    def __post_init__(self) -> None:
        self.allowed_hosts = self.allowed_hosts.split(",")