#Embedded Python

Picture a dark room. The only light comes from a wall of displays and a blinking prompt that politely informs you:

“LINK TO M-U5K-3T: HIGH LATENCY – DIRECT CONTROL DISABLED.”

Perfect.

You are supposed to be a Rover Commander in the Musketeer Corps. Somewhere absurdly far away, you have an autonomous unit parked on a hostile exoplanet. It wants you dead, but it’s also covered in minerals that could fund a small civilization.

The ore on M-U5K-3T is everywhere. Unfortunately, everything else on the surface is actively trying to kill you.

So, we don't pilot. We script.

You aren't driving a car; y

One objective of vectorization is to render unstructured text more machine-usable. Vector embeddings accomplish this by encoding the semantics of text as high-dimensional numeric vectors, which can be employed by advanced search algorithms (normally an approximate nearest neighbor algorithm like Hierarchical Navigable Small World). This not only improves our ability to interact with unstructured text programmatically but makes it searchable by context and by meaning beyond what is captured literally by keyword.

In this article, I aim to demonstrate a couple of methods for easily adding validation to REST APIs on InterSystems IRIS Data Platform. I believe a specification-first approach is an excellent idea for API development. IRIS already has features for generating an implementation stub from a specification and publishing that specification for external developers (use it with iris-web-swagger-ui for the best results). The remaining important thing not yet implemented in the platform is the request validator. Let's fix it!

PEP 578 added Python Audit hooks. A rich variety of events (module load, os interactions and so on) triggers audit events which you can subscribe to.

Here's how to do that. First create an embedded python hook:

Class User.Python
{

/// do ##class(User.Python).Audit()ClassMethod Audit() [ Language = python ]
{
import sys
import time
def logger(event,args):
     if event=='import':
        module = args[0]
        print(f"Loading {module}")
        if module == "numpy":
            print(f"Module {module} forbidden. Terminating process in 3.")
            time.sleep(3)
        

Here are the technology bonuses for the InterSystems Full Stack Contest 2026, which will give you extra points in the voting:

• IRIS Vector Search usage -3
• InterSystems Native SDK for Python or Embedded Python usage -3
• Developer Community Idea implemented - 2
• Docker container usage -2 
• IPM Package Deployment - 2
• Online Demo -2 
• Find and report a bug - 2
• Article on Developer Community - 2
• The second article on Developer Community - 1
• Video on YouTube - 3
• YouTube Short - 1
• First Time Contribution - 3

See the details below.

Vector search is a retrieval method that converts text, images, audio, and other data into numeric vectors using an AI model, and then searches for items that are semantically close. It enables “semantic similarity search” from free text, which is difficult with keyword search alone.

However, in real use, I encountered cases where results that are “close in meaning” but logically the opposite appeared near the top of the search results.

This is a serious issue in situations where affirmation vs. negation matters. If the system returns the wrong answer, the impact can be significant, so we cannot ignore this problem.

This article does not propose a new algorithm. I wrote it to share a practical way I found useful when semantic search fails due to negation.