Breaking: Change lambda task root directory

[aplr]

This PR changes the task root of the lambda function to use LAMBDA_TASK_ROOT, documented in the AWS documentation instead of a custom one.

The previous custom task root /app caused issues with creating and running the base image for newer lambda runtime versions.

This introduces a breaking change, as evaluation functions using the base layer need to copy their code to LAMBDA_TASK_ROOT. However, this only takes effect when deliberately choosing the new image hosted on github packages.

[KarlLundengaard]

For clarification "However, this only takes effect when deliberately choosing the new image hosted on github package", how is the deliberate choice made here? Does it mean manually updating the Dockerfile for the evaluation function, or is is this change part of the deployment procedure (i.e. updating and redeploying the evaluation function would mean choosing the new image)?

[aplr]

"Deliberately" here means, that you have to write FROM ghcr.io/lambda-feedback/baseevaluationfunctionlayer:3.8 instead of FROM rabidsheep55/python-base-eval-layer in your app/Dockerfile.

Therefore both re-deploys, and even re-builds will not be affected at all, as long as you don't consciously change the base image.

---

This is only hypothetic as I turned off pushes to rabidsheep55/python-base-eval-layer.

However, for clarification, if the base layer is rebuilt w/ python 3.8 and pushed to rabidsheep55/python-base-eval-layer, it might well be that re-deploys will lead to failing evaluation functions. I suspect that the behaviour of AWS lambda's python base image has changed in the past 6 months since the last build.

[aplr]

Btw, python 3.8 will be deprecated from the 14th of October by AWS Lambda, and will cease to function in January 2025.

https://docs.aws.amazon.com/lambda/latest/dg/lambda-runtimes.html

[KarlLundengaard]

Ok, thanks for clarifying.