This is the typical sentiment AI example, but Kafka and Debezium are used instead of synchronous calls.
The sentiment analysis example analyzes a comment made on a product, analyzes the sentiment, and rates it from 1 to 5 stars.
Regarding architecture, there are two services implemented in Quarkus: one manages the product reviews, and another contains the HuggingFace model for analyzing the content.
Kafka and Debezium indirectly manage the communication between both services, which is the correct way to avoid the double-writes problem.
Debezium reacts to this event (Change Data Capture) when a new comment is stored in the MongoDB database and places the data into a Kafka topic.
Then, the sentiment service gets the event, processes it, and sends it to another topic.
Finally, the first service consumes the event and updates the stars counter.
The following schema summarizes the architecture:
This example works in any Kubernetes cluster; in the deployment directory, you can find the manifests ordered by execution.
The only specific manifest is the 10_openshift_routes.yaml that creates the OpenShift routes to access the service.
Moreover, if you don’t want to execute all the manifests individually, the deploy_openshift.sh script is provided to manage all at once.
This example works on Red Hat Developer Sandbox, with free access to a shared OpenShift and Kubernetes cluster.
After you are registered, run the oc login in a terminal window, and then the script will deploy the solution.
This solution doesn’t come with a frontend; it offers two important endpoints to interact with the solution:
// Returns the information of the products with the comments and the ratings
curl "$reviewUrl/productsWithComments"
// Generates a random comment to a random product
curl "$reviewUrl/review/generate"
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent