Appsheet API Endpoint for Listing All Tables and Columns

Dear AppSheet team,

As you all know, the growth of agentic workflows and LLMs has improved exponentially in the past two years. I am working on integrating LLM workflows into my apps. A key use case is a universal agent that can answer any query to extract data from any table using the AppSheet API. To provide context for any app, we first need a list of tables, allowing agents to query the database with "Read a table record." Currently, I’ve addressed this with Apps Script and Google Sheets, but I encounter challenges with column types in AppSheet, such as EnumList. For instance, an LLM query like "SELECT(Customers[Customer ID], [Products] = 'iPhone')" results in an error due to the EnumList.

I am requesting two additional API endpoints: one for retrieving tables and another for retrieving column details, similar to:

- https://www.appsheet.com/api/v2/admin/tables 
- https://www.appsheet.com/api/v2/admin/tables/columns 

For example, the first endpoint could return:

 

{
"Tables": [
{"Table01": "Users"},
{"Table02": "Customers"},
{"Table03": "Products"}
]
}

 

And the second could return:

 

{
"User Table": [
{"User Id": "Text"},
{"User Name": "Text"},
{"User Image": "Image"}
]
}

 

These endpoints would significantly  avoid creating hundreds of workflows for various requirements, enabling a more intelligent agentic setup. Thank you!

Status Open
1 9 578
9 Comments
Rifad
Silver 5
Silver 5

Why ? because, I have solved a significant gap in data querying in Appsheet and integration with other applications by creating a universal bridge within AppSheet, utilizing LLM agents and embeddings. This agentic setup generates precise outputs for any query, any table, any data and finds relationships also and the specified API endpoints will greatly enhance workflow efficiency.


Screenshot 2025-02-18 at 11.51.52 AM.pngScreenshot 2025-02-18 at 11.52.04 AM.png

 

Check out how it queries the AppSheet API and sorts it using natural language and vector embedding.

Screenshot 2025-02-18 at 11.52.21 AM.pngScreenshot 2025-02-18 at 11.52.28 AM.png

Rifad
Silver 5
Silver 5

Here’s another example of how it SOLVES any query using AppSheet app and agents. You can see the 'Columns Required' in the payload, which automatically selects the appropriate data slice to avoid large datasets.

An API endpoint could address additional issues related to column type and curate selector query much better and get even better results.


Screenshot 2025-02-18 at 12.56.02 PM.pngScreenshot 2025-02-18 at 12.56.18 PM.pngScreenshot 2025-02-18 at 12.56.36 PM.pngScreenshot 2025-02-18 at 12.56.43 PM.pngScreenshot 2025-02-18 at 12.56.51 PM.pngScreenshot 2025-02-18 at 12.57.02 PM.png

This is a basic example, but I have other inputs that effectively solve complex problems that typically take product developers hours to solve. Agents may soon take over a lot of tasks. 🙃

MultiTech
Gold 4
Gold 4

It seems like a lot of the stuff that you're doing on this other platform  would be much easier to do if you just did it in AppSheet; you wouldn't have all these sort of problems that you're having, as everything you need is right there at your fingertips.

Rifad
Silver 5
Silver 5

@MultiTech , it seems like you misunderstood my request. I don’t comprehend how you can create LLM agentic tools using Appsheet for orchestration purposes.

My request is to create an additional endpoint that can list all the tables and columns from Appsheet.

I’m eager to understand how to accomplish these tasks I mentioned in Appsheet. 

 

 

 

 

Rifad
Silver 5
Silver 5

@MultiTech wrote:

you wouldn't have all these sort of problems that you're having


I clearly don’t understand what you mean. I was simply requesting something, and it didn’t cause any problems. It was just an additional request. 

 

MultiTech
Gold 4
Gold 4

@Rifad wrote:

I don’t comprehend how you can create LLM agentic tools using Appsheet for orchestration purposes.


The same way you're building them on this other platform. 😀 Automation is an incredibly powerful tool, capable of accomplishing anything you can with Make, Zapier, or other workflow-builder platforms.

For an Agentic type of automation, you might:

  • LLM call
  • branch on condition
  • more LLM calls
  • more branching on conditions
  • calling actions
  • API calls
  • calling processes on records
  • etc.

Many of these tasks are much easier to accomplish in AS because you've got all the data there, and ref data easily accessible, etc.  Also if you know the things you're asking the Ai to do ahead of time, you can prepare things in your app to accommodate this and make things easier.


@Rifad wrote:

My request is to create an additional endpoint that can list all the tables and columns from Appsheet.


Would it surprise you to learn that there actually is something like this already? 

Rifad
Silver 5
Silver 5

@MultiTech wrote:

For an Agentic type of automation, you might:

  • LLM call
  • branch on condition
  • more LLM calls
  • more branching on conditions
  • calling actions
  • API calls
  • calling processes on records
  • etc.

 

Ah, I now understand what you mean. Yes, it’s beneficial for Appsheet’s specific use case. In my case, Appsheet is just one part of a broader workflow, possibly a single node in a 20-node workflow. There are many out-of-Appsheet integrations, such as Llamaparse for document parsing and Cohere for reranking, supabase, converters, extractors, among others.

These nodes require handling differently, which Appsheet is not capable of. I used to rely on Appsheet as the central component and use AppScript, but I clearly had to stop using AppScript, especially due to limitations. But I still have my old workflows running like how you mentioned above.

For instance, one of my workflows would take about 20 minutes to process, especially large PDF processing with agents working to solve a problem. 

 

 

 

Rifad
Silver 5
Silver 5

@MultiTech wrote:

Would it surprise you to learn that there actually is something like this already?


Surely would love to try out!

It’s quite intriguing, even though I’m not certain if I’ve ‘solved’ a small problem within the grand scheme of all these agents and LLMs.
However, there’s something these agents do and pinpoint accurate data by traversing like a maze and extracting the right data from any database.

I find it truly remarkable. For instance, consider the p vs np problem. p represents the problem and solution we need, while np is the challenging aspect where it solves polynomial-time execution. Incorporating category theory into it allows for the construction and mapping of relationships and the understanding of scenarios. I’m not sure if all this makes sense.

 

Rifad
Silver 5
Silver 5

I hope the AppSheet team considers adding an endpoint for tables, columns, and their types. Just one more endpoint like this..

Screenshot 2025-06-17 at 11.46.35 PM.png