Use Cases for AI Support?

I’m curious why integrating AI functions directly into Lucee will be beneficial. What’s the big picture and/or roadmap.

For context, Lucee 7 introduces some AI functions. AI / LLM Integration :: Lucee Documentation

So I could find out that gemma3:1b hallucinates that I’m a contributor to the spring project :slight_smile: (small models are so much worse at this)

Lucee makes integrating with different technologies easy, just like good old CFQUERY does for JDBC (well originally CF was C++, but you get the idea!)

I first discovered CF after getting frustrated trying to send mails with Perl way back in 95

3 Likes

I do like the idea of abstracting some of that stuff, but it’s hard to imagine what kind of landscape we’re going to see in the future. Obviously cfquery is something that devs can’t live without. Not trying to be a contrarian, but CFML also did stuff with Forms, Ajax, Graphs, etc. which turned into a nightmare for both the CFML engine and developers who used it. However, the big difference I guess is that all that stuff was for front-end development.

Maybe others can relate, but I found it hard to understand AI and what it can do.

To me, it’s more like a search engine that responds in sentences (or images) rather than SERPS.

But I’m sure that is just the beginning.

So to follow your question, why would Lucee want to provide a connection to this?

For one, AI is not going away any time soon, so it might be good to stay ahead of the curve.

And for another, it would be great to use CFML.

Bill (@noursecode), given that challenge you share, you might enjoy this talk by Dan Wilson, “AI for devs”, which I hosted today on the online cf meetup. His 60-min talk, covering many uses and scenarios, is available at YouTube via the playlist whose link is easiest remembered as https://recordings.coldfusionmeetup.com

1 Like

Blockquote

What’s frustrating about -

use Net::SMTP;

?

:wink:

1 Like

phew! we have a topic to discuss at CFCAMP 2025 when we meet IRL, no awkward silences

Looking forward to it! :beers: :dromedary_camel::beers: :dromedary_camel::beers: :dromedary_camel:

I dont think of chatGPT/copilot (same thing)/ollama AI like a search engine.

I think of it as a group of matter experts with one spokesperson, who reads (and understands) documentation or a website or code (in this context) in microseconds and respond, recommend changes or ways to achieve something.

It knows design patterns, best practices and can also ‘duck type’ like ignore them to achieve a goal. It’s like having 10 keynote speakers at your desk on call.

Now as long as you give it ‘good’ context, it will give you amazing output.

I’ve been using github copilot (with cluade 3,.7 sonnet thinking) this week to analyse our teamcfml framework as well as extensions and code.

I can ask, how does the functionality of my component for say ‘twilio’ compare to other cfml components or libraries in other programming languages.

Or ask, is there any new endpoints I need to add to the xero-cfml github project to update it to the latest endpoints on the xero api?

or generate a series of components to interact with every method on this api.