Hi guys,
I am a no coder who is getting better and learning about coding, I run a start-up that streamlines sports clubs adminsiative burden while helping their online community thrive.
I am looking to integrate GPT into my application so that users have a all seeing eye into all their data, my question is what would the flow diagram look like for setting this up for my users and what coding would be required and can I use GPT itself to achieve this while I am learning!
You are a software start-up with no developers? From what I read; it looks like you need to use a mix mode strategy. You are dealing with structured data and SSRS or PowerBi would be best. If you want to use an LLM, use gpt-3.5-turbo-0613 or 4 and use the function capabilities as well as LlamaIndex to do sql queries. Lots of people are trying to cram LLM in everything. vector databases are great for unstructured data Code search results · GitHub
Pinecone will also give you sticker shock.
Hi Larry,
Thanks for your response and suggestions. I appreciate your insight into the mix mode strategy and the potential use of tools like SSRS or PowerBi for structured data. However, I’m particularly interested in leveraging the capabilities of GPT-3.5-turbo-0613 or 4 for my application.
As you rightly pointed out, I’m indeed a non-coder running a software start-up. I’m currently learning and improving my coding skills, but in the meantime, I’m looking for ways to integrate GPT into my Bubble application to provide users with a comprehensive view of their data.
Bubble has a unique data structure and requirements, and I’m keen on understanding how I can best utilize GPT within this framework. Could you possibly provide a more detailed roadmap or flow diagram on how I could go about integrating GPT into my Bubble application? Specifically, I’m interested in understanding the steps involved, the coding requirements, and how I could potentially use GPT itself to achieve this while I’m still learning.
Honestly, I’m not a 100% sure what you’re trying to accomplish. I do understand that having a natural language connection to your data is great and users can ask questions without knowing SQL. I’ve done what you’re trying to do (I think) to a SQL server using pyodbc and having ChatGPT create the SQL and return the results of what you’re asking for. I’m using langchain and a lot of prompt engineering tweaks. It actually works very well. I am going to change the app to use llamaindex Natural language SQL A Guide to LlamaIndex + Structured Data - LlamaIndex 
0.7.2 (gpt-index.readthedocs.io). As for using Pinecone for your structured data? Well, I personally would not use a similarity search DB for it, IMO at least not now. Anyone else have any ideas?
Do you think we could have a chat on google meet about this?