[WIP] Python refactor - Githubissues

drewbo commented 5 years ago

Towards #3

drewbo commented 4 years ago

Will also close #4

Deployed and ran the stack successfully. The new workflow for a problem looks like:

Generate a TFserving/fastai-serving docker image, note it's input and output format carefully
Write a new lambda/download_and_predict/handler.py function. If you are (1) hitting a tile endpoint (2) using the standard b64 encoded API and (3) saving an array or JSON object, no changes are necessary. Otherwise there are example_lambdas which should how to subclass or otherwise modify the base example for easy modification.
Update the database migrations if needed. This is an unfortunate extra "coupling" problem (because the database schema needs to match the lambda function output and docker image return) that I'm not sure how to solve.
[WIP] Define minimal config
[WIP] Ideally the lambda will be tested and type checked to prevent errors in production
[WIP] Deploy via serverless

cc: @wronk @Geoyi @vincentsarago

drewbo commented 4 years ago

For ease of earlier merging, we will decouple the lambda refactor to python with (1) the deployment and (2) the utility scripts. I'll update README + build and merge soon cc: @vincentsarago @ingalls

developmentseed / chip-n-scale-queue-arranger

[WIP] Python refactor #5