Chalice Example

Requirements

python 3.7 and 3.9
poetry
terraform 1.1.9

How to use?

put all python code into example/chalicelib, initialize app in app.py
put all stateful deployments into infra
configure both terraform backends
configure config.py with environment variables to pass into lambda functions
use run script for all commands

Deployment flow

initialize and apply terraform from main infra folder
run config.py to get terraform outputs and pass them to config.json
generate requirements.txt from pyproject.toml
generate terraform deployment with chalice
run config_backend.py to setup backend for this deployment
initialize and apply terraform from example/infra
connect API Gateway to DNS record manually

Advantages

lambdas, layers, roles and API Gateway is generated with chalice
easy to use handlers, good integration with AWS environment
seems faster than similarly designed API using FastAPI (lower package size?)
local server for testing uses real AWS credentials to connect to the account

Disadvantages

complicated deployment, requires custom scripts to pass all values and modification of a lot of files to introduce changes
a lot of things are automatically generated so we need to remember to do it before PR to have all deployed changes in the repo itself
rigid project structure, might not be extensible for larger applications
after successful deployment, proper API Gateway needs to be connected to Route53 separately, chalice only configures API Gateway to use proper DNS but it doesn't create or change DNS records

Notes

(after setting up infra only, python API functionalities not yet properly tested)

This config is useful for small and personal projects only because of the huge configuration effort to make it work with many different environments (stg, prd). Also the project structure is very rigid and everything needs to be done in the chalice way.

Configuration is quite complicated for start and I don't think (yet) that the automatic IaC generation via chalice is worth it. It might be when adding the whole python framework into account but this needs to be further tested.

Personal quick projects (without terraform) - perfect. Huge company projects with very custom internal deployments - not so much.

Ideally if the provided API functions are good by itself it might be worth taking those and implementing into the python codebase without using the whole framework.

example works as a main or sub app, more similar can be added and managed separately to manage different contexts but domain configuration will be harder in this case and probably not worth the effort as we would have to maintain another API Gateway deployment and configuration separate from the main infra deployment.