How to start the backend server without using docker?

[BladeSun]

How can I start the backend server without using docker?

[achaosss]

I have the same question.

[auh83]

Everything works for me with Docker. Without Docker, I tryed with virtualenv but it didn't find the solution for the moment...

[machov]

you guys can run it using uvicorn main:app --reload @auh83 @achaosss @BladeSun

[DanielGuarnizo]

in which directory we should run this command ?

[machov]

cd backend the backend folder root, where the fastapi backend lives

[machov]

I'm struggling to understand which command is intitializing the backend in production? because the Dockerfile doesn't have this uvicorn main:app --reload

[carlos-dz]

The docker-compose.override runs command: /start-reload.sh, it most be already on the image from where the backend is build tiangolo/uvicorn-gunicorn-fastapi:python3.10

[machov]

the Dockerfile runs uvicorn main:app --reload when importing the Dockerfile from FROM tiangolo/uvicorn-gunicorn-fastapi:python3.10
with command CMD ["/start.sh"] is there!

however `COPY ./app /app/app takes place till the end here .

How is unicorn being executed without the app? does this mean CMD at the end of tiangolo/uvicorn-gunicorn-fastapi:python3.10 is somehow delayed?

I see that docker-compose.override.yml has
command: /start-reload.sh
which executes uvicorn

however I can't understand how does docker-compose.yml work in production

[carlos-dz]

The Dockerfile just have instructions tu build an image, copy the files, install necessary packages on the image...doesn't run or execute the container. To run a single container from an image you would use the command docker run [image]... but docker compose let you build multiple images from different Dockerfiles and then run the containers for each image.

If you run "docker compose up -d" and then "docker compose exec backend batch" you can cd and explore the files inside the container and your will find start-reload.sh which executes uvicorn once the backend container is running.

[gospodima]

I have made a similar setup for backend without using docker. async sqlalchemy + pydantic schemas (without sqlmodels) are used there though.
Take a look: Async FastAPI SQLAlchemy Template.

[auh83]

Thank you @gospodima !