fastapi-plan

Dead simple but powerful template manager for FastAPI applications.

• About
• Quickstart
  • Initialization
  • Development
  • Debug (http)
  • Production (https)
• Project structure
• High level overview
• How to add new endpoint (step by step)

About

Features:

• postgresql database with Tortoise ORM as ORM
• well organised, rock solid project structure (see section Project structure)
• ready-to-use user model, authentiaction system (JWT), hashing with Bcrypt
• easy to undarstand config.py with settings (there is only one file for changes: .env)
• out-of-the-box well-tested routes for login and user (register, read, read_me, update etc.)
• aerich for migrations
• well-designed tests folder filled with tests for existing user model/user endpoints
• auto-generated strong passwords for database, secret_key and superuser password
• poetry or pip
• deployment ready docker-compose.prod.yml file with poetry, you will only need own domain

Furthermore:

• full project structure schema
• high level overview how this project is organised and why, questions like where do the settings live or what every variable in .env file is used for
• step by step explanation how to add new endpoint, from creating new model, adding schemas and routes to migrating database and writting tests (it's always better to have it and optionally adopt it, than wasting time trying to figure out the best dev path)

Quickstart

NOTE: you will need docker and optional but recommended poetry installed!

install via pip (or poetry) globally:

pip install fastapi-plan

there are 3 docker-compose files available, one for development, one for running project via http, and the last one is for production with enabled https using traefic as a proxy (and letsencrypt for ssl), steps to initialize new project are the same for every approach, you can then choose from 1-3.

0. INITIALIZATION

initialize new FastAPI project:

fastapi-plan

enter project_name and other information and after project is ready, cd project_name and continue installing dependencies:

poetry install

# optional if you selected "requirements.txt" (with venv installed)
pip install -r requirements.txt

1. DEVELOPMENT

since we wanna use uvicorn in development, create only postgres container using docker-compose.yml file like that:

docker-compose up -d

now run aerich migrations and configure tortoise (and add first superuser)

aerich upgrade
python app/initial_data.py

finally you can run this command to start uvicorn server

uvicorn app.main:app --reload

2. DEBUG (http)

To make it available from http://localhost on your local machine or http://your-host-name on VM just run

docker-compose -f docker-compose.debug.yml up -d

The diffrence between development approach is that web server automatically runs aerich and initial_data.py using shell script (app/initial.sh), so you don't have to do anything except changing some lines in .env file:

1. PROJECT_NAME - it will show up in docs view as a name of project.
2. FIRST_SUPER_USER_EMAIL - first account email
3. DEBUG - when it's false, the POSTGRES_SERVER is set to localhost for development, so change it to DEBUG=true to use db postgres server.

3. PRODUCTION (https, own domain)

To make it available from https://your_domain.com on VM run

docker-compose -f docker-compose.prod.yml up -d

The diffrence between development approach is that web server automatically runs aerich and initial_data.py using shell script (app/initial.sh), so you don't have to do anything except changing some lines in .env file:

1. PROJECT_NAME - it will show up in docs view as a name of project.
2. FIRST_SUPER_USER_EMAIL - first account email
3. DEBUG - when it's false, the POSTGRES_SERVER is set to localhost for development, so change it to DEBUG=true to use db postgres server.
4. DEFAULT_FROM_EMAIL - your private email for ssl purposes, e.g. they will inform you shortly after some problems with you certificate.
5. MAIN_DOMAIN - your own domain e.g. example.com

Plesae also note that to get no-test certificate, you should comment line "--certificatesresolvers.myresolver.acme.caserver=https://acme-staging-v02.api.letsencrypt.org/directory" in docker-compose.prod.yml file, by default you will use test certifactes (to be sure that everything works, there are some hard limits on number of certifiactes you can ask per week!). You should comment line "--log.level=DEBUG" also (but it can be useful when debugging traefik). There would probably be problems anyway, just be sure that everything works via http using 2. DEBUG apropach. If it then doesn't with the https, you should refer to traefik docs.

Project structure

|── app
|    ├── api                              # endpoints/dependecies
|    |
|    ├── core                             # settings and security algorithms
|    |
|    ├── crud                             # CRUD operations
|    |
|    ├── migrations                       # for aerich migrations
|    |
|    ├── models                           # tortoise models
|    |
|    ├── schemas                          # pandatic schemas
|    |
|    ├── tests                            # tests
|    |
|    ├── initial.sh                       # initial shell script used by docker
|    ├── initial_data.py                  # init database and add first superuser
|    ├── main.py                          # main fastapi application file
|
├── config                                # nginx server config file
|
├── .env                                  # .env file with settings
|
├── Dockerfile                            # dockerfile for web app
|
├── aerich.ini                            # aerich (migrations) configuration
|
├── docker-compose.prod.yml               # puts it all together in prod (https)
|
├── docker-compose.debug.yml              # puts it all together in debug (http)
|
├── docker-compose.yml                    # puts it all together (development)
|
├── (optional) pyproject.toml             # python dependencies (poetry)
|
├── (optional) poetry.lock                # python dependencies (poetry)
|
├── requirements.txt                      # python dependencies (pip)

High level overview

This project strucutre is mostly based on the official template (but not only) which is really great but unfortunatly does not support Tortoise ORM and is... (too?) complicated. All the security or problematic stuff (app/core/security.py with verify_password function, login and token routes, JWT token schemas) are just copied from there, so you can be pretty sure it will work as expected.

The main thougts are:

• There two sorts of settings, first one located in .env file for the ENTIRE project, and python-specific settings which lives in app/core/config.py, the file is based on pydantic solution (using dotenv lib). Why? Well, that's simple, this is due to 12factor methodology, python-specific settings inherit from .env file, so this is the only place where you actually change something. If you have any problems understanding mentioned config.py file, just refer to pydantic - settings management, it's pretty clear.

• Models, crud, schemas, api routes, tests... it might be confusing how to actually ADD SOMETHING NEW here, but after following next section (learn by doing, step by step), it should be pretty easy

• Database-related stuff is very convinient, taken mostly from Tortoise ORM docs and just working. There is register_tortoise function in main.py, TORTOISE_ORM variable in app/core/config.py. Please, be aware that if you don't run initial_data.py SOMEHOW (in development- you have to do it yourself, in debug/production it is handled by shell script initial.sh, which also runs tests and migrations), you won't be able to connect to database. initial_data.py is hearbly based on the same named file in official template mentioned earlier. It has two responsibilities, first is running init function from Tortoise to initialize connection, and the second - creating first superuser (defined in .env) if one doesn't yet exists.

• Migrations are also provided by Tortiose (the tool is aerich), docs can be found here in aerich repo. The default migration (default user model) file is already included. After changes in models (e.g. new model Cars), just run aerich migrate, aerich upgrade and you are good to go.

• All tests lives in tests folder, with some pytest-specific content included. If you feel unconfortable with pytest, feel free to read articles about using it, and if you just want to see how to test new enpoints/models, just read next section.

How to add new endpoint

Let's imagine we need to create API for a website where users brag about their dogs... or whatever, they just can crud dogs in user panel for some reason. We will add dummy model Dog to our API, with relation to the default table User and crud auth endpoints, then test it shortly.

1. Create file dog.py in app/models folder:

from tortoise import fields
from tortoise.models import Model


class Dog(Model):
    name = fields.CharField(max_length=100)
    age = fields.IntField(null=True, default=None)
    breed = fields.CharField(max_length=100, null=True, default=None)
    owner = fields.ForeignKeyField("models.User", related_name="dogs")

2. Add import in app/models.__init__.py:

from .dog import Dog # type: ignore

3. Migrate changes

aerich migrate
aerich upgrade

4. Create file dog.py in app/schemas folder (pydantic schemas with typing support):

from typing import Optional
from tortoise import Tortoise
from tortoise.contrib.pydantic.creator import (
    pydantic_model_creator,
    pydantic_queryset_creator,
)
from pydantic import BaseModel
from app.models import Dog


# Pydantic models from Tortoise models, pls refer
# https://tortoise-orm.readthedocs.io/en/latest/examples/pydantic.html#basic-pydantic

Tortoise.init_models(["app.models"], "models")
DogPydantic = pydantic_model_creator(Dog, exclude=("owner",))
DogPydanticList = pydantic_queryset_creator(Dog, exclude=("owner",))

# Unfortunately, it doesn't work the other way around


class DogCreate(BaseModel):
    name: str
    age: Optional[int]
    breed: Optional[str]


class DogUpdate(BaseModel):
    name: Optional[str]
    age: Optional[int]
    breed: Optional[str]

5. Add import in app/schemas.__init__.py:

from .dog import DogUpdate, DogCreate, DogPydantic, DogPydanticList # type: ignore

6. Create crud_dog.py in app/crud folder

from app.schemas import DogCreate, DogUpdate
from app.crud.base import CRUDBase
from app.models import Dog, User


class CRUDDog(CRUDBase[Dog, DogCreate, DogUpdate]):
    def get_dogs_by_user(self, user: User, skip: int = 0, limit: int = 100):