In my first article, I described the baseline version of the FHIR Patient Snapshot Agent: a Streamlit and Python application that retrieves FHIR resources from InterSystems IRIS for Health and uses an LLM to generate a concise patient summary.

This follow-up article explains how I extended the project with two additional InterSystems-focused features:

• Source context vector search
• Embedded Python artifacts for IRIS-compatible review

The goal was to make the project more useful as a clinical summarisation prototype while keeping the design small enough to understand and reproduce.

Hi Community,

In this article, I will introduce my application iris-fhir-agents A multi-agent clinical AI platform powered by InterSystems IRIS for Health. Features agents for triage, specialist consultation, pharmacy safety, and FHIR server exploration — all grounded by IRIS Vector Search RAG. Includes a no-code Agent Builder that lets you design and deploy custom clinical agents without writing a single line of code.

The successful construction and implementation of AI agents to address diverse use cases in the healthcare sector depend on high-quality data and APIs, effective governance, and management. The InterSystems IRIS FHIR server delivers all of this and is also fluent in Python, Vectors, and Interoperability. Combined with a strong LLM, patients, physicians, caregivers, and managers gain access to state-of-the-art technology for personal and public health.

Hi!

I'm trying to compiling a class with an Embedded Python and got this error:

Compilation started on 06/12/2026 11:44:17 with qualifiers 'cuk'
Compiling class User.VectorSearch
ERROR #7802: Worker job/s '749:33' unexpectedly shut down in group '#Default:(446070926892):0'.
ERROR #7812: Work queue unexpectedly removed, shutting down.
ERROR #5002: ObjectScript error: <THROW>WaitForComplete+215^%SYS.WorkQueueMgr *%Exception.StatusException ERROR #7802: Worker job/s '749:33' unexpectedly shut down in group '#Default:(446070926892):0'.
ERROR #7812: Work queue unexpectedly removed, shutting down.
Detected 3 errors during compilation in 1.020s.

1. Introduction

Epidemiological surveillance is one of the foundational pillars of public health. Régis Júnior et al. (2026) define it as a continuous system of data collection, analysis, interpretation and dissemination of health events — a function whose effectiveness depends critically on the quality of information systems, data analysis capacity, and coordination between different levels of care.

Hi Community,

Have you ever wished your EHR could think? Not just display data. Not just fire alerts. But actually read a patient record, reason over clinical guidelines, and write a structured referral order back to the system — in response to a single message from a clinician

In this article, I am going to show you how to create your own custom clinical AI agent.

🏥 About iris-fhir-agents App

iris-fhir-agents is a multi-agent clinical AI platform built entirely on InterSystems IRIS for Health.

In the previous article, we used pyprod to create production components while relying on the UI for production configuration. That same production can now be defined entirely in Python:

from intersystems_pyprod import Production, ServiceItem, ProcessItem, OperationItem

iris_package_name = "HelloWorld"

class MyProduction(Production):
    services = [
        ServiceItem(
            "MyServiceName",
            "HelloWorld.MyService",
            host_settings={"target": "MyProcessName"},
        )
    ]
    processes = [
        ProcessItem(
            "MyProcessName",
            "HelloWorld.MyProcess",
            host_settings={"target": "MyOperationName"},
        )
    ]
    operations = [
        OperationItem("MyOperationName", "HelloWorld.MyOperation")
    ]

In the previous IoP article, I showed how IoP can expose Python messages to DTL by generating JSON schemas. That is useful when the message is primarily a Python object and we want the IRIS tooling to understand its structure.

This time, the direction is a little different.

Starting with IoP 3.7.1, a PersistentMessage can now be a native IRIS message body class. The Python class is still the source code you write, but the generated IRIS class extends Ens.MessageBody

With Embedded Python and the Native API, it is becoming increasingly natural to write part of IRIS application logic in Python. But one question quickly comes up: how can you manipulate IRIS persistent objects from Python without losing the connection to the native object model, class dictionary, indexes, storage, and SQL projections?

iris-persistence explores that question. The project provides a Python object persistence layer for InterSystems IRIS, inspired by %Persistent

InterSystems IRIS globals are one of the platform's core strengths: they store hierarchical data in a direct, ordered, and efficient structure. But when working from Python, manipulating globals can sometimes feel closer to a low-level API than to the natural habits of the language.

The iris-global-reference project provides a Python layer on top of IRIS globals. Its goal is simple: make access to globals more readable, more idiomatic, and easier to integrate into modern Python code, without hiding the underlying hierarchical model.

When developing Python applications with InterSystems IRIS, you can quickly end up with several execution contexts:

• Python launched directly by IRIS with Embedded Python;
• a regular python3 process that loads the Embedded Python libraries from a local IRIS installation;
• an external Python application that connects to IRIS through the official native driver.

These three cases are useful, but they do not behave exactly the same way for imports, system configuration, object APIs, and SQL access.

Data privacy regulations such as GDPR, LGPD, and HIPAA demand that organizations know exactly where Personally Identifiable Information (PII) lives inside their databases. Yet in practice, most teams rely on manual inventories, tribal knowledge, or external scanning tools that require data to leave the database engine — a process that itself creates privacy and security risks.

This article presents an MVP that takes a different approach: it runs PII detection inside InterSystems IRIS using Embedded Python, analyzing data where it lives and never exporting it to an external process.

Abstract

Solar irradiance forecasting is critical for grid stability in photovoltaic (PV) power plants. This article replicates and extends the methodology of Lara-Benítez et al. (2023) "Short-term solar irradiance forecasting in streaming with deep learning" replacing the original offline simulation with a fully operational streaming pipeline built on InterSystems IRIS. We leverage IRIS Interoperability Productions as the streaming backbone, Embedded Python to run MLP, LSTM, and CNN deep learning models, and IntegratedML as an AutoML baseline.

Hey Community, 

Yeah, I know we have a lot of tips & tricks articles - we even have a special tag for it, right? But I can not share my own collection. In this collection, you can find snippets for beginners, as well as some everyday ObjectScript constructions. In fact, it's also a way to learn something new about my favorite programming language. So, if you guys have anything to add, you are welcome!

Today I have published a new Open Exchange package for generation of Synthetic Data directly into IRIS.

 It can be a frustrating process to find decent datasets when you are looking to make a demo app. Maybe the dataset doesn't matter that much, but you still want it to appear somewhat genuine and with several linked tables that are usable directly within IRIS with the neat implicit joins with ->. Maybe you just want linked tables that are easily installable with IPM to benchmark queries, this dataset generation would be perfect.

I need a %SYS.Python object that will be accessible from different processes. The goal is to avoid reinitializing the Python object whenever it is called. Ways that I checked and decided that it is NOT what I want:

• %-variables (process variables). Reusing only in the same process
• Save-and-restore via globals. Restoring means heavy initialization, which I want to avoid

Next, my ideas (none of them look like a silver bullet):

• Using Interoperability Production.

What is a Microservice?

A microservice is an architectural style that structures an application as a collection of small, autonomous services. Each component is developed around a specific business capability, can be deployed independently, and is typically managed by a miniature, specialized, self-governing team. (Source: https://microservices.io/)

Since I started using Claude Code, my motivation to create things has skyrocketed.

Previously, even if I wanted to build something, actually doing the coding felt like a hassle, so unless there was a very strong need, I rarely went as far as programming. But now, if I just jot down the specifications, Claude Code handles the rest automatically, resulting in a dramatic improvement in productivity.

I come from a generation native to ObjectScript, so I used to feel some hesitation when it came to switching to Python.

Working with files often starts off simple. open the file, read, and process. That approach works perfectly well, until the file happens to be an Excel file.

A Common Assumption

At first, an Excel file (.xlsx) looks like just another data file, rows, columns and values. nothing unusual. So it's natural to assume it can be read the same way as a .txt ot .csv file. But that's where things start to break.

Why Excel files behave differently

The key difference is how the data is stored:

-> .txt / .csv - plain text, line-by-line.

-> .

Global Guard AI

1 Introduction

In environments that use InterSystems IRIS, globals are the physical foundation of data storage. Although system queries and administrative tools exist for metric inspection, global growth analysis is usually reactive: the problem is generally only noticed when there is disk pressure or performance impact.

In the modern healthcare landscape, finding clinically similar patients often feels like looking for a needle in a haystack. Traditional keyword searches often fail because medical language is highly nuanced; a search for "Heart Failure" might miss a record containing "Congestive Cardiac Failure."

I am excited to share iris-medmatch, an AI-powered patient matching engine built on InterSystems IRIS for Health. By leveraging Vector Search, this tool understands clinical intent rather than just matching literal strings.

Hello everyone,
I’m looking to implement Continuous Training (CT) as part of an MLOps strategy for some data science projects in IRIS. I want to automate the full cycle:

- Monitoring model performance & accuracy degradation.
- Retraining models automatically.
- Validating and updating production models.

I’ve looked into IntegratedML, but it seems more focused on the SQL interface for training (AutoML). Even with the new Custom Models (beta), which allows for more flexibility with Python, it doesn't seem to provide the "Continuous" orchestration out of the box.

I’d like to know: