Unlock a world of possibilities! Login now and discover the exclusive benefits awaiting you.
Generative AI is here.
Generative AI is there.
Generative AI is everywhere.
Welcome to life in mid-2023 my friend. Whether you are caught up in the hype that it will solve all of the worlds problems and make trillions of dollars for your organization or not, the facts are simple:
Your bosses boss wants it.
Your boss wants it.
You want it.
In my previous post Creating, Calling and Implementing the OpenAI Connector I shared with you how to utilize the brand spanking new Qlik Cloud Connector for OpenAI.
Before the ink was even dry, and I had time to celebrate, there were so many questions about whether or not the connector was going to be available for Qlik On-Premise/Client Managed solutions.
Then more questions arose about whether or not Qlik could connect to the other "9,390,029" Generative AI platforms or privately trained Large Language Models.
The Answer is a Resounding Yes
I responded that of course Qlik can connect to them because nearly all provide REST API's. Thus, you simply need to use the Qlik REST Connector. Easy for me to say, but it's not like the parameters for REST API's just build themselves. So I've created a few videos to help you get access to, what I know you NEED access to.
Calling OpenAI via REST
Nobody is sure what the question is, but they know OpenAI is the solution. Right? Well, you and I probably aren't falling for all of that hype, but we can agree that it will be a lot easier for your organization to start dreaming up use cases, once you know you can execute on them.
In this video I walk you through all of the necessary REST API settings to successfully connect Qlik Sense to OpenAI via a REST connection. While the values I plug in are for OpenAI it's easy to correlate them to other Large Language Model (LLM) solutions including privately trained LLM's as well.
Of course I also added some technical fun like making the calls via a Partial Reload so that you can let the end user input the prompt you want to to send OpenAI on the fly.
Calling Hugging Face via REST
While OpenAI has all of the buzz, Hugging Face is also a very highly utilized and valuable Large Language Model provider as well. I figured it only made sense to demonstrate at least 1 alternative system being called via the Qlik Sense REST API connector as well.
The use case is a fun one, viewing a list of translation models available and choosing one and executing it to translate input text on the fly. In the previous video I demonstrated how to collect input prompts from the end user and call OpenAI via a Partial Reload. In this video, I go even deeper and demonstrate how you can debug errors that can arise during the partial reload. I mean, why not? You have to learn some time.
Many thanks
As I shared in the Hugging Face demo I love Generative AI Art. What I also love are the art collectives where folks around the world collaborate on ideas. Prompt Engineering is an art form in and of itself. What you often see online are the completed works, but if you could see the pure prompt poetry that goes into them on the front end you would be equally amazed.
The collective I am part of has a channel dedicated to prompts that are simply 1 or 2 words. The goal is to see how concise a prompt can be, and determine how creatively the Generative Art tools can respond. The opening illustration was produced by 2 of my friends in such a collaboration: @RuttenbergKyle and @FoxTrotFrog(on Twitter) came up with the term ArterialTechnocracy. Normally, I like to use my own generated images, but when I saw that image I knew I simply had to use it for this post. They were kind enough to permit me to do so.
I've been writing code for 42 years. Two things I know are that while algorithms allow for creativity, the code you input must follow strict rules. I'm still learning from masters like @RuttenbergKyle and @foxtrotfrog when it comes to completely opening my brain to create words that don't even exist to try and convey concepts that otherwise might take 1,000 words.
While I tagged them in the image description, I thought it fitting to share even more thanks here. Primarily to share my formal thanks, but also to encourage you to consider building your prompt prompt engineering skills. Generative AI tools, regardless of their flavor, all accept prompts. The better your prompts are as input, the more deterministic the responses will be as output. Or in this case, the more creative your prompt is, the more you allow for creativity as the output.
