Lessons learned

:fire:
------

:man_shrugging:
------

- lambda has an uncompressed code size limit of 250mb. This causes it to reject deployment packages that are 700+mb (if you include a spark dependency in this project, e.g.), but not if they're 329mb. It's not obvious where the line is.
- there are some rough edges still, which makes sense given that serverless includes tools for like every language. for example, it doesn't seem like you can set the integration response to "convert to binary (if needed)" from the serverless template. That's annoying.

:man_facepalming:
------

- `serverless deploy` doesn't run `sbt assembly`. No matter how many times you make changes and follow them with `serverless deploy`, your code in AWS will never change. That step that takes a minute where it tells you it's "Packaging service..."? That's building the deploy package from whatever jar it finds where you tell it to look for one.

The issue records lessons learned about AWS Lambda package limits, Serverless deployment, and the sbt assembly step, but names no files or tests and does not request a specific change. Start by reviewing the deployment setup and build workflow described here; a documented improvement would need a defined location and clear acceptance criteria.

Written by the indexing model from the issue text.