This article introduces SHAP explainability methods as an approach to understand the reasons behind predictions in machine learning black-box models. It also includes a simple Jupyter notebook that you can use and modify to gain hands-on experience with these concepts:

https://www.kaggle.com/code/jorgeivnjh/explainability-in-ml-models

https://github.com/JorgeIvanJH/Explainability-in-ML-models

We will leverage these concepts for a future implementation in our Continuous Training Pipeline: https://community.intersystems.com/post/complementing-iris-mlflow-continuous-training-ct-pipeline

Okay… first things first - sorry it took me so long to write this up. There was just too much happening, too many conversations, too many photos, too much travel and honestly… too much excitement to sit down and write right away.

READY 2026 started the way it always does, with that unique mix of anticipation and energy in the room. Look at these excited faces! 

@Gabriel Ing, @Mariam Makhmutova, @Olga Zavrazhnova, @Dean Andrews, @Anastasia Dyubaylo, @Iryna Mykhailova 

(Let's be honest, this was the end, and everyone was happy that we survived 😁)

When managing critical healthcare data through an integration engine you want to know the moment a queue starts backing up or a service drops. 

Unfortunately, this usually leads to an email inbox stuffed full of notifications that can sometimes seem impossible to maintain.

Well, I’ve not solved that problem. 

But I have created something that could help...

Hey Community,

Round 1 of the Community Bounty Program "Idea to Application" is officially open. Three ideas from the Ideas Portal are ready to be implemented — pick one or more, build it, publish it on Open Exchange, and earn Global Masters points. 

🏅Every qualifying submission earns a Credly badge, Global Masters badge, and 10k+ points. Implement all ideas this round and unlock Tier 2 rewards.

Hi Guys,

I've the below code to build JSON and send it via httpRequest called from business Operation.

But I'm getting this error, which I'm guessing that the formulated JSON doesn't have the correct format, so I was looking to convert the JSONStream to JSON string and try to validate it so see where the JSON is broken, problem is that I'm expecting that there will thousands of records in this JSON so it will hard to go through it, so is there a way to track for error or why this JSON is failing to send    

Thanks

Hi Guys,

I've the following code:

and I would like to convert the resulting json from JSONStream to JSON string?
thanks  

#North American Demo Showcase entry. 

>> Answer the question below to be entered in the raffle!

⏯️ Message Operational Data Store

Create an operational data store using the data flowing through your production. Create user-defined analytics tables based on fields and paths to their data from incoming documents with varying standards (FHIR, CDA, HL7v2, etc.).

Presenters: 
🗣 @Kwabena Ayim-Aboagye, Sales Engineer at InterSystems
🗣 @Christopher Carmichael, Sales Engineer at InterSystems
🗣 @Jeff Semmens, Senior Sales Engineer at InterSystems

#North American Demo Showcase entry. 

>> Answer the question below to be entered in the raffle!

⏯️ AI-Assisted Rare and Complex Disease Detection

This demo shows how InterSystems Health Gateway can be used to pull in outside patient records from networks like Carequality, CommonWell, and eHealth Exchange, creating a more complete longitudinal view in a clinical viewer. That full record is then analyzed by AI to surface potential rare disease considerations with clear reasoning, helping clinicians see patterns they might otherwise miss.

Presenters: 
🗣 @Jesse Reffsin, Senior Sales Engineer at InterSystems
🗣 @Georgia Gans, Sales Engineer at InterSystems
🗣 @Annie Tong, Sales Engineer at InterSystems

#North American Demo Showcase entry. 

>> Answer the question below to be entered in the raffle!

⏯️ AI Assistants for the Unified Care Record Powered by Gemini

In this demo, you will see how Gemini works directly with FHIR data, and how it leverages the harmonized dataset provided by InterSystems Unified Care Record. It also showcases multiple AI assistants helping multiple groups of users, e.g. clinicians, patients.

🗣 Presenter: @Simon Sha, Sales Architect at InterSystems

#North American Demo Showcase entry. 

>> Answer the question below to be entered in the raffle!

⏯️ HL7 Validation Error Profiler: Rapid Insights into Batch HL7 Data Quality Issues

This demo highlights how HL7 validation logs can be transformed into scalable, actionable data quality insights using a lightweight application built on top of Health Connect.

• Batch ingestion and parsing of HL7 validation error logs into structured data
• Aggregation of errors by segment, field, and type to reveal systemic issues
• Interactive dashboard (Streamlit) for filtering, exploration, and root cause analysis
• Rapid identification of top error patterns across entire data feeds
• Natural language chatbot for intent-driven data investigation and querying
• PDF report generation for sharing clear, evidence-based feedback during data source onboarding 

Special thanks to @Henry Wojnicki for his contributions to designing and refining the application workflow.

🗣 Presenter: @Lynn Wu, Sales Engineer at InterSystems

As an experiment in agentic coding in ObjectScript I'm using VS Code to try and create an implementation of an SFTP server.

SFTP is built on top of SSH, so the first phase involves implementing an SSH server. While working on the KEX part of that the agent (using GPT-5.3-Codex) reported:

1. IRIS has what we already need for signing and hashing: RSASHASign, RSASHAVerify, RSAGetLastError, SHAHash, SHA1Hash, SHA3Hash.
2. It does not list a Diffie-Hellman key exchange or modular exponentiation API in %SYSTEM.Encryption.

It then offered these options:

1. Keep using IRIS APIs for cert/key/sign/hash.

Despite the fact that LOCK (docs) is a foundational part of InterSystems IRIS, responsible for concurrency, there is not a lot of discussion on the Developer Community about it. Which is understandable, considering it's stable and fairly low-level command. In this article, I will show a simple example of how to use locks with interoperability.

Below is an update to the IRIS release calendar for 2026 and an early look at the planned changes for 2027. The key takeaway for 2026 is that maintenance release numbering will differ slightly from prior years.

2026: IRIS 2026.1 maintenance release numbering

• July 2026: first maintenance release for IRIS 2026.1 (planned version: 2026.1.4)
• October 2026: next maintenance release (planned version: 2026.1.5)
• February 2027: following maintenance release (planned version: 2026.1.6)

This numbering change prepares us for the expanded maintenance cadence planned for 2027.

Every great community is enriched by those who work tirelessly behind the scenes, crafting the knowledge and resources that help others grow. In the InterSystems Developer Community, one such person is @Derek Robinson — a dedicated educator, whose passion for teaching has helped shape how developers everywhere learn to work with InterSystems technologies.

👏 Let's take a closer look at Derek's journey in the InterSystems ecosystem.

Hey Community!

We know not everyone could join our live stream from the READY 2026, so we’re happy to share that you can now watch our session online.

👉 Streamlining Development Workflow with InterSystems Developer Community and Ecosystem

I am working on setting up OAUTH with FHIR and Microsoft Entra.  I have configured the Client and configured it in the FHIR Server successfully. However, I get 401 when authenticating with a token.   Looking at ^ISCLOG I see:
^ISCLOG("Data",9,0)="accessToken=<jwt token>, scope=, aud="
^ISCLOG("Data",10)=$lb(3,"OAuth2","[OAuth2.ServerDefinition:ValidateJWT]","171430","%SYS","2026-04-28 17:41:25.305549397","OAuth2.ServerDefinition.1","","zwKKakZZZx2")
^ISCLOG("Data",10,0)="JWT valid? sc=1"
^ISCLOG("Data",11)=$lb(3,"HSFHIRServer","[HS.FHIRServer.

In v2026.2 (currently available as a Developer Preview), we are adding a feature that can help in a FHIR Endpoint SMART/OAuth authorization - more out-of-the-box flexibility in audience value validation.

We have a large, existing production taking a feed from a single up-stream system and sending to a wide-variety of downstream systems. One of the downstream systems has been decommissioned, and we have deleted the classes that implemented the routing/transformations from the namespace, and deleted the components from the production (Wardview and WV Out in the diagram below). We now want to delete the messages associated with those components without deleting the messages associated with any other components.

READY 2026 has come to an end and we’re sure it brought a lot of great moments along the way 🚀 🚀

What stood out to you the most?

It could be a session, a conversation, an idea or just a moment that made you think “this is why I came”.

Share your experience in a few lines 👇

We’d love to hear what you’ve discovered and feel free to explore other replies and like the ones you relate to 💙