tfserving-api-test

main.py

This is the api with endpoints. When receiving a request, it sends the image to the correct model. The models are hosted in separate docker containers.

start_models.sh

Add tfserving models in models/ and add a new line to the startup script.

Dockerfile

This file has instructions on how to build the api image. Move files into the docker image, install the requirements etc.

docker-compose.yml

Use docker stack deploy api -c docker-compose.yml to deploy the stack. This will deploy the api image and the models in separate docker services. The services are on the same virtual network and can communicate with each other.