Runway Example python App

This is an example app demonstrating how to deploy a python app to runway.

• clone this repo, and navigate into that directory
• runway app create
• runway app deploy
• runway open

You can then deploy changes by git commiting them, and running runway app deploy again.

Python

For this example, we built a small API with fastapi, uvicorn and poetry.

poetry is a modern dependency manager for Python projects with deterministic installs (through a lock), run commands, automated virtualenv and much more. The idea of this example is to get you started. A similar setup with django is of course possible, but a bit more involved.

Setup

The four steps to build a small application.

Step 1

Create a new project using poetry:

$ poetry new python-runway
Created package python_runway in python-runway

Step 2

Add our framework (fastapi) and a web server (uvicorn).

To add them we go into the project directory and add fastapi:

$ poetry add fastapi
Creating virtualenv python-runway-9x66e4jD-py3.10 in /Users/till/Library/Caches/pypoetry/virtualenvs
Using version ^0.79.0 for fastapi

Updating dependencies
Resolving dependencies... (18.4s)

Writing lock file

Package operations: 15 installs, 0 updates, 0 removals

  • Installing idna (3.3)
  • Installing sniffio (1.2.0)
  • Installing anyio (3.6.1)
  • Installing pyparsing (3.0.9)
  • Installing typing-extensions (4.3.0)
  • Installing attrs (22.1.0)
  • Installing more-itertools (8.14.0)
  • Installing packaging (21.3)
  • Installing pluggy (0.13.1)
  • Installing py (1.11.0)
  • Installing pydantic (1.9.2)
  • Installing starlette (0.19.1)
  • Installing wcwidth (0.2.5)
  • Installing fastapi (0.79.0)
  • Installing pytest (5.4.3)

... and uvicorn:

$ poetry add 'uvicorn[standard]'
Using version ^0.18.2 for uvicorn

Updating dependencies
Resolving dependencies... (11.0s)

Writing lock file

Package operations: 9 installs, 0 updates, 0 removals

  • Installing click (8.1.3)
  • Installing h11 (0.13.0)
  • Installing httptools (0.4.0)
  • Installing python-dotenv (0.20.0)
  • Installing pyyaml (6.0)
  • Installing uvloop (0.16.0)
  • Installing watchfiles (0.16.1)
  • Installing websockets (10.3)
  • Installing uvicorn (0.18.2)

Step 3

For demo purposes, the application will only respond to requests at the root (/).

Create a application: the project's main module is called python_runway (as well), so let's create a main.py in ./python_runway (it did only contain __init__.py so far):

import uvicorn
from fastapi import FastAPI

app = FastAPI()

@app.get("/")
async def root():
    return {"message": "Hello World"}

def start():
    """Launched with `poetry run start` at root level"""
    uvicorn.run("python_runway.main:app", host="0.0.0.0", port=5000, reload=True)

Note: we used our default port of 5000. But you can customize this with runway app config set PORT=x and use an environment variable in your code.

Step 4

Last but not least, add a start script to poetry's pyproject.toml which is a meta file which configures all project settings. This gives you an easier way to start your application:

[tool.poetry.scripts]
start = "python_runway.main:start"

To test, run poetry run start after you added it. Adding this will also enable our builder to start your application after it's done assembling all the requirements.

Deployment

The deployment is the usual steps:

• runway app create
• runway app deploy
• runway app open