Forum Discussion

MohamedBennouf's avatar
MohamedBennouf
Qrew Member
10 months ago

ChatGPT like search engine?

Hello all.

Long time no talk!

Anyway, I started using QB for my company in 2006 to records problems/solutions found in the field by our field service engineers related to our products. We know have 65K+ cases! Of course now the issue is to search for similar issue in this immense database!  For instance, it will be useful four junior engineers confront with similar problems in the field. At least to give them a list of similar issues cases to help them fix the tool. 

But as you can imagine a QB keyword Seach type is almost useless with the database is that large. One because if the keyword is not present (or miss-spelled) then QB will not find the similar cases. Or sometimes it returns 100's cases!  For instance say the "beam is unstable" and "the beam is What we need is a semantic search option with QB. Or even better a tool like chatGPT applied to the 1000's of cases already in the databases. 

I have been experimenting with chatGPT, chatPDF and also with CLAUDE (different from OpenAI company) and got some great results. But it means I needed to export a version of the QB tables (repair database) into a PDF and use one of those tools above. Which also means we cannot have an update system since dozen + are added by our FSE Worldwide!

Anyway, I just wanted to start the conversation about large language models (LLM's like chatGPT) applied to QB. Of course, this will not be limited to Q/A of a QB applications. They could also be used to analyze QB data and so on.


Will love your point of view and hopefully QB team to look into it! I think SaleForce has already integrated AI in their system..


Cheers.

Mo



------------------------------
Mohamed Bennouf
------------------------------

1 Reply

  • Hi Mohamed!

    In order to do semantic search within a Quickbase application, we would be reliant on their product team to support some type of vector database feature. Not really sure if they are tooled to build that type of capability into every customer environment, or what the performance implications would be, but that is their challenge to solve.

    I have been able to get this type of thing going for my own Quickbase data, but it was necessary to first replicate the data in an environment that already had support for vector storage and search. I've had some success with xata.io and their Vector Search API.

    I've also had some success creating my own ChatGPT app in Quickbase using OpenAI's Chat Completion API. I've uploaded the app to the exchange, but I'm waiting on Quickbase's approval.

    Happy to talk more about it anytime if you're interested. 



    ------------------------------
    Justin Torrence
    Quickbase Expert, Jaybird Technologies
    jtorrence@jaybirdtechnologies.com
    https://www.jaybirdtechnologies.com/#community-post
    ------------------------------