Ways to connect backend documentation- For ChatGPT ?

[Rifad]

Hey there,

I've been dabbling with something really neat, combining CHATGPT, Node.js, and Appsheet. It's all about linking up parts of my app with a cool backend API. Here’s the scoop: I use the Appsheet API and Chat GPT API to make the Action happen.

Picture this: I feed a main row and its bunch of child tables - think 5 to 10 of them - into the GPT API. What it does is like a quick tour through the whole process in different departments. It picks out the key stuff and wraps it up in a neat summary, perfect for a manager or CEO to glance over. So far, it's pretty slick. With a single click, it tells me what’s up or if there’s a hiccup somewhere.

Here’s what I put in:

{   
    "max_tokens": 1000,
    "model": "gpt-4-1106-preview",
    "temperature": 0.3,

    "ParentKeyID": "TENDOIS387",
    "ParentTableName": "Purchase Order",
    "ParentKeyColumnName": "Purchase Order ID",
    "ParentColumnToSend": 47,
    "chatGptPrompt": "{ As a CEO, I require a concise but need detail and strategic summary of the current Purchase Order (PO) in HTML plain text format without ```html. \"Length\": \"Under 150 words\", \"Format\": \"Simple HTML tags only (<b>, <blockquote>, <br>, <li>)\", \"Avoid\": \"Specific Key IDs like Purchase Order ID, Sales Order Acceptance ID ...\", \"Content Focus\": {\"Discrepancies\": \"In Price, dates, quantities\", \"Action Items\": \"List key actions with specific detail\", \"Item Issues\": \"Insights into current status\", \"Critical Aspects\": \"Decisions, delays in days, financial implications\", \"PO Status\": \"Update on status, key dates, financials\"},  \"Priority\": \"Clarity, conciseness, strategic relevance\", \"Emphasize\": \"Minimal HTML spacing and do not add unwanted linebreaks between each lines.\"} : {}",
    "childTables": [
        { ......

}

Here is the output:

And here's a glimpse of the kind of data I send over to the GPT API.

My query is:

Exploring Advanced Integration Techniques

Right now, the setup works great using just the data I feed into the GPT engine. But I'm wondering, is there a way to also send table conditions from Appsheet to GPT? Currently, we rely on a documentation PDF, but I'm thinking about giving GPT more context with the backend conditions for enhanced clarity. For instance, some columns have values like this.

CONCATENATE(

"<img src=""",ANY(My Business Settings[Whats Next Show Icon]), """ alt=""icon"" width=""50"" height=""50"">",
    
    
CONCATENATE(    
    IFS(
      [Purchase Approval Status] <> "Approved",
      CONCATENATE (
      						"<blockquote><b><i>Approve Your Purchase<i/></b>","<br>","- You need to approve this Purchase order.</blockquote>"
                  )
    ),

    IFS(
    	AND(
          [Purchase Approval Status] = "Approved",
          [Receiving Status] <> "Received"
          ),
      CONCATENATE (
      						"<blockquote><br><b><i>Awaiting Delivery<i/></b>","<br>","- Awaiting this delivery.", IFS([Due Date] <= TODAY(), " Already met the <b>Due Date</b>. You must communicate with the supplier for updates. Use the comment action to mark your communication updates.</blockquote>" )
                  )
    )
    
)

If GPT could interpret this, it would significantly improve the decision-making outputs. It's already close to perfect, and with this capability, I could craft prompts that are 100% spot-on.

It would be super helpful if anyone could offer some insights on this. Is there a way to somehow extract information from the documentation PDF and incorporate it into the prompt, or maybe some kind of workaround?