A while back when OpenAI and ChatGPT were starting to gain popularity, I took it upon myself to learn how to call the Open AI API programmatically from a website and build a layer of bespoke functionality on top of it - Something that many, many people went on to do and create so called AI startups with varying degrees of success.
At the time there was a heck of a lot of noise from alleged AI thought leaders and talking heads that had no idea how to deliver any type of value or build things with the AI technologies they were hyping.
There was just SO many people talking about Gen AI, but not many people sharing how to implement it.
So my motivation became..
Cut through the noise, ignore the hype.. learn how it works, how to build things with it and share that knowledge with others.
GenerativeAI.tours
I decided to create a website that helps the user plan a trip itinerary by allowing them to select a destination, their interests and duration. The website then composes a prompt based on their selection sends it to the OpenAI API which returns a list of popular attractions for the destination taking into consideration the trip duration and users interests.
There was little more behind it than this..
- Smart response caching was used to reduce OpenAI credits
- It was a plain html web page that called an Azure Serverless Function that talked to the Open AI API
- It called another API (Unsplash API) to retrieve images for attractions on the destination itnerary
A more detailed explanation of its workings is in the original walkthrough video.
Worldwide AI Hackathon Finalist
I was very proud that the project was a finalist in the 2023 Worldwide AI Hackathon.
Solution Description & Walkthrough Video
You can actually read about my approach in another article I created, which also contains a video of me explaining my approach and the architecture of the solution I created.
GenerativeAI all the Things!
I actually planned on creating a number of tools using this approach and similar and went on a bit of a domain buying spree, so today have various projects in a variety of states for:
generativeai.email, generativeai.photo, generativeai.page, and generativeai.report to name a few.
Though, i'm not one of these people who things Generative AI or LLM's is a silver bullet, I know which use cases they are particularly good for and others they are not.
To see a list of use cases for AI including GenAI, check out UseCasesFor.ai.
What Do I Need to Do to Get It Live?
The project submitted to the hackathon, while fully functional, ran locally.
All I need to do to get it live is change where it is persisting some of the data. Notably the cached prompts and their previous results (.json) and previously returned attraction image from the Unsplash API. I'll use Azure Blob Storage to store and retrieve these. Easy peasy.
I just hope the Unsplash API keeps working since its acquisition by Getty Images.
Why Finally Put It Live?
A few reasons..
Firstly, I don't want to just be known solely as an XR specialist, which you could easily be mistaken of looking at the XR content, experiences and videos I create these days.
I have been a Tech Lead or Software Development Manager in Microsoft web technologies for over a decade and is probably still my strength.
Putting the site live, demonstrates my expertise in AI consulting for customers.
Secondly, i've been approached by a couple of people asking for exactly the same functionality i've created in Generative AI Tours. So having completed my MSc Entrepreneurship & Innovation degree i'm dusting this off and considering monetisation and partnerships. (I'd almost certainly choose a new brand name).
Thirdly, to keep my eye in.. API's change, solution design approaches change. Being able to leverage Open AI API for other purposes is a good skill to have.
Fourthly, Whilst i've given talks on how I implemented it to share knowledge with others, having it live for people to try out would be even better.
I've still got websites live from previous side projects i've created and given talks about including GoodOrBaddy.com (trained an image classification ML model offline and hosting in a Azure Function) and IsChristmasTree.com (an image classification model trained and hosted online in CustomVision.ai, with webpage calling an API for it).
Summary
I kind of regret not putting the site live originally. But as I want to do more again in the web/AI space, this is something that can be easily remedied.
This is what I love about being a C#/.NET developer. I can build any cloud, web, desktop, mobile, AI or XR application and enjoy sharing these skills with others. Also check out VisualStudioTips.co.uk!
I've got quite a bit on my plate at the moment, so this isn't a top priority for me, but I'll let you know when it's live.
-- Lee
Update 24th January 2025: A basic version of the site is now live at GenerativeAI.tours
