We've just released a new version of Aista Magic Cloud. The primary feature in this release, is that it allows you to copy and paste a JavaScript tag into your existing website, for then to "have an intelligent conversation with your website based upon ChatGPT and AI".
Notice, you need to fine-tune your own model, but this is automatically done for you by simply pointing Magic to your existing website, at which point it will automagically crawl your website, and scrape it for data, generating a custom machine learning AI model in the process, capable of answering intelligent questions about your site.
The above website has roughly 70 pages, and after scraping our site, we're able to get an accuracy of roughly 50% in answers. However, importantly, the module logs all questions/answers, and allows you to reinforce it as you've gathered more training data, to strengthen its result. My guess is that over a month or two, the above bot will be able to provide an accuracy of 95+ percent, simply due to people asking it questions, and me "correcting" its answers, and re-training it on my corrections a handful of times.
You can see me demonstrating the thing in the video below
Top comments (13)
Can you guys turn it into a browser extension that crawls and does this for ALL API documentation websites and AWS and Azure docs?
Because that would literally save like a million man-hours a year for companies around the world.
You can upload XML, JSON, CSV or YAML files. We're not planning a browser extension at the moment - But it's a good idea, so you never know ... ;)
Yeah please!!!
Aww man, i'm bummed, went to the site to try it out and my first question crashed it lol
gyazo.com/e11f4baf1e5e3f1d3dcb9ff4...
Darn it! It's not us, it's OpenAI ... :/
oh hm, maybe add a thing to your code to catch error responses and go "chatgpt is down again" lol
VERY good idea! I just released this like an hour ago though, so everything is a bit "BETA'ish" ... :/
tried it again, dunno if it's helpful to you:
gyazo.com/2f51279dc739a7fc4cbebaea...
Thank you, at least it worked. The model I'm using currently only has about 50% accuracy. It needs to be trained multiple times to give higher accuracy. Did you watch the video? It kind of explains things.
Also, it will never answer questions that aren't relevant to the site you're browsing. So try to ask something that's relevant, such as "What us Hyper IDE?", etc ...
The last is kind of "the point" too, and how it's suppose to work ...
well no, look at the response i got, it was an API issue with the JWT token.
by the way, this time it worked!
gyazo.com/39a04b0c2e5cb2672c51dc2f...
Hehehe :D
That answer was roughly 90% wrong! But we'll get there with a couple of reinforced training sessions. First of all, we're in Cyprus, not Germany. But it's "getting there" ... ^_^
Actually no, it is a "semi relevant" answer. System endpoints is a part of Magic. It just doesn't make sense in this context ... :)