Wednesday, June 10, 2026, 12:00 noon, CDT (UTC−5:00) Download .ics

Special Topic:
Taking AI Kit off the Grid – How to build your own private semantic search engine – Keisuke Miyako

“How does your application use AI?” Heard that question recently? I sure have! We are all living through a major change in the world of technology, and as application developers ourselves, we are expected to deploy AI features as soon as possible or risk falling behind the competition and/or be viewed as, brace yourself, old and outdated.

We have already seen a great demonstration of integrating a 4D application with AI last year with Ricardo Mello and Mathieu Ferry’s fantastic presentation, 4D AI Platform – the Future Is Now. Their demo showed how to add AI chat into an app to perform searches, generate reports, and serve other useful enquiries. However the best AI chat style functionality is still powered by cloud services like Claude.ai and ChatGPT. Cloud-based APIs are convenient and capable, but there are issues to consider with tying your application to them such as cost, vendor lock-in, and privacy.

What about taking AI off the grid instead?!?

Keisuke Miyako returns to the user group to deliver a tour de force presentation (as per usual but especially so in this case) on how to run your own local LLM, including model considerations and setup configuration, to provide semantic search functionality in 4D applications that leverages 4D AI Kit and 4D.Vector search capabilities. “A privately hosted open-source model, running entirely on your own hardware, is producing results in the same range as a state-of-the-art cloud API. No API key, no data leaving your machine, no per-token billing.”

Check out this post on discuss.4d.com from Keisuke Miyako that jumps right into the content that will be presented. A bit of familiarity with the terminology and tools will certainly help keep up with this packed demo. We will see how to run models with llama.cpp directly from 4D using a component in order to vectorize and search a localized cache of 4D documentation.

Agenda

1. All About Embeddings
2. What is Cosine Similarity
3. Ride the llama.cpp (component)
4. Startup Explained
5. Specification
6. Sample Application

This meeting is being hosted live with real-time discussion on Zoom.  Click here to join…

To follow along with the meeting slideshow, access it here when it is available…
Download a copy of the slides used in this meeting
Full presentation details
Demo application

Have comments or questions after the meeting?  Either post them here or to discuss.4d.com.

If you cannot attend the live Zoom meeting, no worries!  A recording will be posted to the 4DMethod Youtube channel after the meeting ends.