A connection test can tell you that a server responded. A schema can tell you that a column is named Code. Neither tells you, by itself, what that code represents or whether it belongs in a particular target field.

That distinction is where reliable integration design begins. Before deciding which source asset maps to which target, or which components should form a pipeline, we need to understand the interfaces—not just their shapes, but what those shapes mean in context.

This article explains how DataPipelineDemo approaches that problem on InterSystems IRIS for Health.

Hugging Face has made thousands of datasets accessible to developers working in machine learning, natural language processing, analytics, and artificial intelligence. InterSystems IRIS provides a powerful platform for storing, querying, and processing data. Bringing these two ecosystems together creates interesting possibilities for developers who want to experiment with real-world datasets directly inside IRIS.

However, importing a dataset is more than downloading a file and inserting rows into a table.

An LLM can suggest that a source column belongs in a particular target field. That is useful—but it is not yet an integration.

Someone still has to establish the connection, understand the target's constraints, choose the right components, connect them, start the pipeline, and check what actually arrived.

The Quickstart introduced DataPipelineDemo, a working demo on InterSystems IRIS for Health. This article explains the design behind it: how integration requirements and available platform capabilities become a running, verifiable pipeline.

The central idea is not “ask AI to write everything.” It is:

Let AI make integration decisions within explicit boundaries, then make those decisions executable, observable, and testable.

The facial recognition has become the most popular method for validating people's identities, thus enabling access to systems, confirmation of personal and documentary data, and approval of actions and documents.
The challenges are related to performance when the database is very large, accuracy, and especially the privacy of biometric facial data. For these challenges, nothing is better than using InterSystems Vector Search, as it allows:

1. Performing vector searches in millions of records with much faster responses than traditional methods.

Hi, Community!

Are you a Python developer starting to work with InterSystems IRIS? See how to prepare your local environment:

Setting Up Python and InterSystems IRIS on Windows

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")
    ]

## Why another management portal?

The built-in IRIS Management Portal is powerful and complete, but its key operational signals — cache efficiency, global activity, web applications, users and roles, scheduled jobs, security events — are scattered across many menu levels. When you are on call and need to answer *"is this instance healthy right now?"* in ten seconds, that matters.

**LuminIRIS** is my contest answer: a single-pane, dark-themed **glassmorphism** console where the five management scenarios an operator actually uses every day live behind one login:

1.

An Evidence-First Operations Environment for InterSystems IRIS

I wanted a management interface that made it easier to understand what InterSystems IRIS was telling me. Finding the information was part of the problem. Keeping track of what I had learned, what I was about to change, and whether the change actually happened was another part, and it became the more interesting one.

The Management Portal has a great deal of functionality.

We've published the first open benchmark measuring LLM competence on InterSystems ObjectScript and the IRIS platform: ObjectScript LLM Benchmark.

Summary

The ObjectScript LLM Benchmark evaluates 14 LLMs on 140 hand-written items covering ObjectScript and the InterSystems IRIS platform. The benchmark tests code generation, comprehension, and bug fixing. 93 items are execution-graded on a live IRIS instance, while 47 comprehension items are evaluated by a rubric-driven LLM judge. Overall accuracy ranges from 0.515 to 0.851, with the top 9 models forming a statistically overlapping cluster.

Hi Community,

We’ve launched a new free, interactive tutorial that provides a practical introduction to building interoperability productions with InterSystems Python Productions (intersystems_pyprod Python library).

🐍 InterSystems PyProd Tutorial 🧑‍💻

Technical companion to the LuminIRIS management console — a 2026 InterSystems Technology Innovation Contest submission.

Suggested DC tags: InterSystems IRIS, Embedded Python, REST API, ObjectScript, Bug report, Contest

The goal

LuminIRIS is a browser console for IRIS. The official SysAdmin API (/api/admin, shipped with IRIS 2026.2) covers monitoring, security, tasks and auditing — but it has no Python execution endpoint

Hello Community,

Want to explore InterSystems technologies without installing or configuring anything?

We now have nine free, hands-on interactive tutorials available on the Instruqt platform. Each tutorial launches a ready-to-use environment directly in your browser, so you can start experimenting with InterSystems products and technologies right away.

The tutorials cover everything from IRIS data models and application development to Python interoperability, healthcare integration, machine learning, and generative AI. Most take only 5–30 minutes to complete.

Choose your tutorial

There are a lot of useful projects on Open Exchange that are easy to miss. Here are 10 worth bookmarking if you develop with InterSystems IRIS.

Hi Community!

Another month, another round of Open Exchange gossip! This time, the numbers reveal three interesting trends: 

1. Python keeps growing
2. Tried-and-tested tools lead the install charts
3. AI agents and MCP enter the spotlight

Let’s take a closer look!

Environment

• InterSystems IRIS for Health, Community Edition, image intersystemsdc/irishealth-ml-community:2026.1
• Platform: Docker, Ubuntu 24.04 LTS host, aarch64 (Apple Silicon / ARM64)
• Embedded Python: 3.12.3 (confirmed via ##class(%SYS.Python).GetPythonInfo()), CPF_PythonPath correctly set to the packages directory
• sklearn, , all import cleanly through the embedded Python bridge

Problem

Every TRAIN MODEL ... USING {...} call fails with NoEstimatorChosen, regardless of:

• (tried , 0.50, and omitted entirely — all fail)

An election happening now

It is August 2026, and Brazil's general election campaign is underway. On October 4, 158,745,463 eligible voters may choose a president, state governors, senators, federal representatives, and state or Federal District representatives. A possible presidential and gubernatorial runoff is scheduled for October 25. The dates and electorate come from Brazil's Superior Electoral Court.

This timing turns public information into an immediate civic need as millions of people try to understand candidacies, proposals, and public careers.

I currently use pyenv on my local machine to manage my Python installations. I just installed IRIS on my Mac, and am currently trying to configure Embedded Python. I read from the docs that it's recommended to install from Homebrew. Since this was only recommended, I figured I could try to configure IRIS with my installation of 3.11.16, which I built with env PYTHON\_CONFIGURE\_OPTS="--enable-shared" to ensure that Python wasn't built statically (as is the default behavior). Building it this way then gave me a .dylib

Hi Community!

Thank you to everyone who participated in the "What functionality would you like to see in PyProd?" sweepstakes! Your ideas and suggestions are invaluable and will definitely influence the evolution of PyProd. 

Now, it's time to announce the winner – watch the video to see the sweepstake drawing:

Hi everyone, 

I created a short demo and reference repository showing how to configure Auth0 as an OAuth provider for an InterSystems IRIS FHIR server, and how to call protected FHIR resources from a Python Flask application. 

The links below include a video walkthrough and a github repo with the sample application used in the video 

Links 

Video walkthrough: 

GitHub repository: https://github.com/MarshyKid/iris-fhir-oauth-demo

Hi Community!

It's time for a new sweepstakes, and we're excited to have PyProd at the center of attention. It has been on the radar recently - we've noticed so many fresh ideas and creative projects coming up. And now the big question for you is: 

 💡 What functionality would you like to see in PyProd? 💡

We'd love to hear your suggestions on how we can make working with PyProd smoother, clearer, and more enjoyable. Whether it's a new feature, a better workflow, or something that would make PyProd easier to use, your idea could make a real difference for developers across the Community.

It has been a year since I joined InterSystems, and I have seen plenty of great articles go by on the Developer Community. The problem is finding them again when my code actually needs them.

Claude Code is already wired to our servers, the docs and Perforce. The community layer was missing, so I wrote a skill for it: one Python script, no dependencies, plus a SKILL.md.

The question I wanted to answer: I am writing this class right now, what has the community already said about the patterns in it?

Why not just wrap the site's own search? Because it is rendered client side: over HTTP, /search?search_api_fulltext=

After an extensive math lesson, we are going to put our new knowledge into practice and create one of the best context-related databases.

Get your magnifying glasses and hats ready, because Vector Holmes is back.

Hello Community,

We are excited to announce that registration is now open for the third cohort of the course:

 🧑‍💻 Developing FHIR Applications Using Python 🧑‍💻

This hands-on program is designed for developers who want to build real-world FHIR applications using Python and InterSystems IRIS for Health. 

📅 Third cohort starts on August 27, 2026