For those are are emerging into the AI space later than others - like myself - I found the prospects and Retrieval-Augmented Generation (RAG) to be a great capability. RAG is not new but it connects from of the underlying AI magic to decision making that makes sense.
The missing piece? Context/Background.
Retrieval-augmented generation in Language Models (LLMs) is a framework that combines the capabilities of both retrieval models and generative models to improve the quality and relevance of generated text. This approach aims to enhance the generation process by incorporating information retrieved from external sources during text generation.
Tabby is a self-hosted AI coding assistant, offering an open-source and on-premises alternative to GitHub Copilot.
HomeLab
For those who haven’t read my previous review of FauxPilot, I have been experimenting with Open Source CodeGen AI for the last few months. Using an old desktop I convert to Proxmox, I dedicated my RTX 2080 to an Ubuntu VM that is my current test-bed for AI code generation tooling.
For quite some time now I’ve helped build platforms and the focus has always been to deliver value to the end user first and foremost. What capabilities would we enable and how great the future would be amiright?
Early in this learning process was the role that dependencies played. The more you layered dependencies on top of one another - the greater the mountain of said dependencies grew from something maintainable to an overgrown monstrosity that threatened to break at any run of CI you introduced.
This is part one of multiple blogs posts about my standing desk and treadmill combo that I have planned for the forseeable future. Reason being the impact on my physical and mental wellbeing since i’ve began incorporating it into my daily working life.
In order to describe where I am at - I need to tell you about where I’ve been. Bear with me for the start of this article as it gets more interesting. If you want the TLDR - scroll to the results/conclusion below.
Open Source software is a space I feel strongly about as a core fundamental requirement to the advancement of technological capabilities that make the lives of others better. I’ve never written code for the sake of proving to anyone why I am better - rather the sole focus has always been:
Identify the problem
Iterate to solve the problem
Get feedback
Repeat
I (and many others) refer to this as the “Mission”. The North Start guiding the why behind what we build - and sharing it openly in order to collaborate with those on similar missions.
I’m far from an expert on LLMs and GenAI but - if there is one thing that makes my discovery phase for investing my time into new technology - it’s getting my hands dirty.
I could pay for Copilot and there would be some significant advantages - but I wouldn’t be learning about common constraints to the operations of such capabilities in ways that translate to future product.
That and when it comes to subscriptions - I am very stingy to spend money and after seeing what FauxPilot lays on the table, I’m starting to see how getting closer to the hype cycle of new technologies is beneficial to every day productivity.
The more you learn about me - the more you will see the experimentation is how I approach all facets of my life. From day to day learning and work, to my mental and physical wellbeing, the journey towards my daily %1 better never stops.
You’ll notice the Experiment tag across different series or blog post categories - This tag illustrates the many facets of life where I put experimentation into place.
A friend and former co-worker of mine - Alex - was really the inspiration behind a lot of the homelab infrastructure that I learn/develop on every single day. He got me started on what to look for and what services I may want to explore.
As many homelabbers know - this only snowballed from there. I learned much of what I know about Kubernetes and all of the related technologies by trial and error of setting up a variety of configurations and machines. I used this as a platform to enable my learning and testing.
Given the wide-array of available options for where to write and host blog articles - what lead me to Hugo and the markdown -> html translation workflow you might ask? It really comes down to knowing how I operate and what would make this venture successful….
Knowing myself and having tried a number of content delivery blog services - I just didn’t feel at home with writing content in a workflow that suited me. Inspiration striking and translating that into words is still somewhat of a new a skill that I am working to harness. As such, writing articles may be a labor of love that requires multiple writes/re-writes.
You might be asking yourself how you ended up here. I’m just as surprised as you are :wink:.
Nonetheless, these are the hosts for the website you’re viewing - or rather they run this workload and others with some room to spare as an experiment.
(More specifically - Two Raspberry Pi 4’s)
Why?
The answer is because I’m very extra when it comes to looking to accomplish any high-level objective in the most difficult ways possible (kidding of course). So maybe this is less of “Blogging the hard way” and more of “Blogging the extra way”.