Dread deploying your API? Use Cookiecutter DiscoPie!
Ever built a sik ML model only to find that it's stuck on your stupid laptop, beefy desktop, or ephemeral cloud machine? Do you need the world to have access to ur dope deep neural net? Do you require serverless deployment because what is a server? Do you insist on containerizing your code because I mean what else are you doing is this 2015 or something? Do you loathe the idea of navigating the treacherous waters of the AWS UI?
If you answered "yes" to exactly every single one of the above questions, then you've come to right place. Cookiecutter DiscoPie is here to handle all the annoying stuff so you can focus on serving that model, or whatever else people do with REST APIs.
Cookiecutter DiscoPie
- Templatizes an API endpoint based on FastAPI. You just need to call your code from the routes.
- Containerizes the service and provides commands for local interaction.
- Deploys the container as a serverless REST endpoint (currently only on AWS, but we have a vision!).
Probably not. There are other ways to do this. But they either don't support containers, or require silly command line steps. You are a Python data scientist, dammit, not Kevin Mitnick!
Plus, our FastAPI-based design means there's hope for supporting multiple serverless cloud backends (huge caveat: there needs to exist an ASGI adapter for each).
Just try it, you nerd!
All you gotta do is fork this repo, navigate to the right dir and use pip to install the DiscoPie command line tool, easy.
cd /cookiecutter-disco-pie
pip install .
cc-disco-pie https://www.github.com/caseyfitz/cookiecutter-disco-pie