This example is based on https://rh-aiservices-bu.github.io/insurance-claim-processing/ but implemented in Java using Quarkus framework.
It uses MongoDB to store content and Minio to store images.
Interesting things:
-
DJL processing images
-
Using Jib with special base Java image with graphics libraries enabled
-
Mutiny for sending parallel requests to two different models
-
Langchain4J with ChatGPT or Llama
The following image shows the schema of the app:
To deploy this example, you can compile by your own, or use Kubernetes.
In deployment folder you’ll find the deploy_openshift.sh script to deploy this solution to OpenShift. If you don’t have an OpenShift cluster, we provide a free cloud OpenShift solution at https://developers.redhat.com/developer-sandbox
Or you can, adapt the YAMLs files easily by changing the OpenShift Route objects to Ingress objects.