airops

An AI-powered platform that enables businesses to build, deploy, and scale AI-driven workflows and applications for enhanced operational efficiency.

AirOps

Hi I'm Matt from AirOps. By now you've probably heard about open ai's GPT 4 Vision model and some of the mind-blowing ways people are using it. Today I'm going to show you an incredibly powerful way to use gp4 Vision within AirOps that lets you take any image and extract structure data directly into a Google sheet. You'll never have to copy and paste and reformat tables from PDFs or manually transcribe information from a website or other document again. 

Let's dive in.

All right here we are in the AirOps platform. Now time to show you how to use the image to data template that will help us get data from images into Google Sheets. So we'll click explore templates. We'll scroll to the bottom here where our image templates are located. We'll select this one which is image to data. This will drop us into the AirOps Studio where we can configure our workflow exactly to our needs. We're going to go with the kind of standard template as it.

Is here. It's pretty straightforward. It accepts either an image URL or an image file. What we'll do is we'll have an LLM step. This will use the GP4 Turbo Vision model. We'll have a prompt that basically gives the instructions of, you know, find the tabular data in this, uh, image and then turn it into a list of objects. That list of objects is really important to get it into a format that we can then easily flatten into a Google Sheets table to then be able to analyze. 

Give it a user assistant pair here so kind of simulating what we're going to ask the llm to do and then give it some instructions on the format of any quantitative data that we pull through into the table. 

Lastly, conditionally ask for either the image URL or the image file to be analyzed. The last step here is to actually integrate and have the Google sheet that we're creating go to the right folder. 

So parent folder we'll click image to data test this will be ordered.

By the most recently edited folder ERS in your Google Drive here will refer to step four output. So the output is what the array that we're going to turn into a table here. 

And then we'll save changes. 

And then in order to test this I'm going to use the income statement of the meta most recent 10K. So here just to make the image a little smaller I'm going to shrink this a bit and then take a screenshot. 

Use an image URL you want of this and now let's try testing.

The app so I will say test all image URL. I'm going to choose a file and then I'm going to the recent screenshot and then hit execute. Give it a second here. This should be able to create a formatted Google sheet in the folder that we selected. There we go. 

All right moment of truth. There you have it. So we have the income statement of meta broken down just as we saw it on the image. Last thing we'll do, we'll go ahead and publish this app and then we can run it either from.

The main screen or we can connect this in any number of ways. You know we could run this in a batch if we had a bunch of images that we wanted to process. We could run this via an API call as well. So we'll do another video on how to use image with an API step or do in bulk as well as those are very frequently asked for use cases. Thanks for watching. Please share AI Agent Store with colleagues.

Building with AI. If you like this video in particular, I'd suggest watching our video on using AI to generate and iterate on prompts within the AirOps platform. This can help you speed up your time to great outcomes with llms. Thank you.

Note, that there may be slight differences in text and video. Watch video for the original content.

AirOps Video