Deploy the Base Application

Run the Application

5 minutes

With the database loaded and secrets in place, start the app and exercise the two main agent paths.

Exercise Run and try the assistant
1

Deploy the healthcare assistant app

Run the following command to deploy the healthcare assistant app:

bash 
cd ~/workshop/healthcare-assistant/1-base-app
kubectl apply -f k8s.yaml

Ensure that the application pod is running:

bash 
kubectl get pods -l app=healthcare-assistant

Using the IP address of your EC2 instance and port 81, open the healthcare assistant app using your browser:

text 
  External URL: http://98.86.181.9:81

Healthcare assistant home screen

2

Try the RAG path

Click the example button (or type) for the medicine question:

What is the dosage and common side effects of Lisinopril?

The agent calls the search_medicine_qa tool, which retrieves matching chunks from pgvector, and returns a grounded answer.

Healthcare assistant home screen

3

Try the text-to-SQL path

Now ask the patient-lookup question:

Can you look up information for patient P001?

The agent calls get_patient_info, which generates SQL against the healthcare_patient table and returns the patient’s details.

Healthcare assistant home screen

Notice what’s missing

The app works, but right now you have no record of what the agent did. You can’t see which tool was called, what was retrieved, how many tokens were used, or whether the answer was even correct. If the assistant told a patient to take double their dose right now, you’d find out from social media, not from your tooling. That’s exactly the gap you’ll close in the next chapter by instrumenting the app with Splunk Agent Observability.

When you ask the patient-lookup question, which tool does the agent invoke, and what backend does it hit?

Click here to see the answer
The agent invokes get_patient_info, which uses text-to-SQL to generate a SELECT against the healthcare_patient table in PostgreSQL and returns the matching row.