Ruby, Faraday and LM Studio API

I’ve been playing with the LM Studio REST API v1 using Ruby and Faraday to communicate with the local server – it’s a great system and I’ve had fun creating feedback loops where the LLM reacts to its own creations.

I’ve found IBM Granite Tiny to be a great LLM to play with on a laptop – very fast and decent responses.

I still need to try all the various JSON properties you can embed in the request and examine all the details of the response object.

Two New Languages

I’ve been learning Ruby and Python lately – Python because it’s become required to dive into AI experiments, and Ruby because it’s a beautiful language.

Years ago I used Python in a cybersecurity class, and it’s nice, but I wish Ruby had become the AI language as it’s way more ergonomic and enjoyable. At least it has Rails for a practical application.

I’ve been reading the “pickaxe book” by Pragmatic Programmers for Ruby and “Learning Python” for Python – both classics apparently, and both an interesting read. Also started playing with scripts in PyCharm and RubyMine (JetBrains IDEs).

Wonderful Nespresso

I recently purchased a Nespresso Vertuo machine, and if you’re a coffee enthusiast like me it’s simply incredible. Not only is it delicious as simply black coffee, the crema on top is like a milkshake. By far the best coffee I’ve ever had.

Amazon puts the machines on sale regularly and the Nespresso site is really excellent. The coffee is pricey, but it’s well worth it.

Diving into the world of XAML

I’ve made progress reading a long book on WPF and XAML, and creating a WPF application to process Word and Excel documents, with a form interface.

XAML is quite fascinating, and I love declarative languages; the end-goal is to move to WinUI 3 instead of WPF (WinUI being much more modern), but I’m starting out with the latter as it’s more familiar to me. They both use XAML.

I hope to post some small utility programs built in WPF on my GitHub page eventually.

Experimenting with HTML5 Canvas

Started playing with HTML5 Canvas to draw graphics in the browser – I used it a long time ago to make 3D charts using Three.js, now I’m just experimenting with its native 2D context for fun.

It reminds me a lot of when I used to make games in QBASIC, manually drawing each pixel or line for each frame of an animation or scene. Fun!

LM Studio for local AI

If you want to experiment with AI models locally, check out LM Studio. It lets you choose a model to download, and then you can run the LLM locally on your machine. Expect maxed out processors and RAM, and a longer runtime to respond than online AI clients, but it’s cool to run it on your own machine.