crop-health-api

API for crop health prediction. For more information check out the developer portal.

Running locally

There are two docker images, one for TorchServe (which runs the models), and another image for FastAPI.

Start by creating a local network with docker, this will allow the containers to communicate with each other:

docker network create my-network

Now, in the torch_serve/docker folder, run one of the following to build the TorchServe image.

# For most CPUs
DOCKER_BUILDKIT=1 docker build --file Dockerfile -t crop_torch_serve --target production-image  ../

# For arm CPU
docker buildx build --platform=linux/amd64 --file Dockerfile -t crop_torch_serve --target production-image  ../

Then run the container:

docker run --rm -it --network my-network --name local_torchserve -p 8080:8080 -p 8081:8081 -p 8082:8082 crop_torch_serve

In the root of crop-health-api, run the following to build the FastAPI image:

docker build --file Dockerfile -t crop_fast_api ./

Then run the container:

docker run --rm -it --network my-network -p 127.0.0.1:5000:5000 crop_fast_api

Now requests can be sent to:

• http://localhost:8080/predictions/binary
• http://localhost:8080/predictions/single-HLT
• http://localhost:8080/predictions/multi-HLT

And the documentation can be found at:

• http://localhost:5000/redoc

Example request: curl -X POST http://localhost:8080/predictions/binary -T cocoa.jpg