Welcome to the third and final publication of our articles dedicated to the development of RAG applications based on LLM models. In this final article, we will see, based on our small example project, how we can find the most appropriate context for the question we want to send to our LLM model and for this we will make use of the vector search functionality included in IRIS.

In the previous article we presented the d[IA]gnosis application developed to support the coding of diagnoses in ICD-10. In this article we will see how InterSystems IRIS for Health provides us with the necessary tools for the generation of vectors from the ICD-10 code list using a pre-trained language model, its storage and the subsequent search for similarities on all these generated vectors.

Finally and with a little delay, we conclude this series of articles about our Workflow Engine by showing an example of the connection that we could make from a mobile application.

In our previous article we presented the general concepts as well as the problem that we wanted to solve by using the task engine integrated in InterSystems IRIS, in today's article we will see how we configure an interoperability production to provide a solution.

Workflow Engine Configuration

First we are going to define the roles of the tasks that we are going to manage, in our example we are going to define two types:

As you have seen in the latest community publications, InterSystems IRIS has included since version 2024.1 the possibility of including vector data types in its database and based on this type of data vector searches have been implemented. Well, these new features reminded me of the article I published a while ago that was based on facial recognition using Embedded Python.

As most of you probably already know, since approximately the end of 2022 InterSystems IRIS included the columnar storage functionality to its database, well, in today's article we are going to put it to the test in comparison to the usual row storage.

In the last article we presented the architecture of our SMART On FHIR project, so it's time to get down to business and start configuring all the elements that we are going to need.

We will first start with Auth0.

AUTH0 configuration

We will start by creating an Auth0 account with a valid email, once registered we will have to create our first application, and we will do it from the menu on the left:

In this article we are going to see how we can use the WhatsApp instant messaging service from InterSystems IRIS to send messages to different recipients. To do this we must create and configure an account in Meta and configure a Business Operation to send the messages we want.

Let's look at each of these steps in more detail.

Setting up an account on Meta

This is possibly the most complicated point of the entire configuration, since we will have to configure a series of accounts until we can have the messaging functionality.