Closed BlairCurrey closed 2 years ago
settled on this architecture
this was also considered:
takes the listening/unpickling off the api but ultimately I dont think this would be necessary for this project and the above would be simpler.
another idea, maybe better. just use api to store data. publish the model to a cdn. utilize tensorflow.js to do prediction in js
cdn could actually be something else (file server?) and interact with api instead
setup python? flask app? or maybe streamlit?
Rough local scaffolding: