Ask AI To Generate a LinkedIn Connection Request
Last updated
Last updated
🍎 Before you begin this tutorial, make sure you've completed the Developer Quick Start Guide.
This tutorial guides you through building a PixieBrix mod using OpenAI to create a Quick Bar action that reads a user's LinkedIn profile, sends that data in a prompt to OpenAI, and then returns messages for a customized connection request to that user.
Don’t worry— this mod only generates the content of a message. It will not automatically send a message. This means you can test this and view responses without actually firing off any accidental connection requests!
We’ll build this mod to search a user’s page, so it’s best to start there. Go to any user’s page except your own. If you need an idea, you’re welcome to use mine 💁♀️.
From the page, open the PixieBrix Page Editor. If you’ve never done this, right-click on the page, then click Inspect at the bottom of the menu. Go to the PixieBrix tab.
You’re ready to move to the next step when your screen looks like this:
To trigger our mod, we need to select a starter brick. These are found via the New Mod button next to the PixieBrix logo on the left sidenav. Click it and select Quick Bar Action.
You’ve just started your mod and are ready to configure the first brick!
In the Name field at the very top, set the value to something descriptive about the mod. I named mine the verbose yet descriptive Generate a personalized LinkedIn connection request message with AI.
The Action Title is the text that appears on the Quick Bar - this should be shorter but still clear about the action. I chose to name mine Get personalized connect message.
By default, the Contexts and Sites fields will suit this tutorial and can be ignored. If desired, you can configure an icon before the Action Title on the Command bar. I prefer to do this as it adds a bit of style and helps distinguish the actions, plus there are a lot of icons to choose from.
✅ Checkpoint! 🏁
To confirm your Quick Bar action brick is set up the way you’d like, click the keyboard shortcut you configured to toggle your Quick Bar. View your new Quick Bar action - it should be at the top! If it’s ready to go, move to the next step. (Don’t worry—you don’t have to save anything!)
Next, we need to scrape some info from the LinkedIn profile page to help OpenAI to make a personalized message.
To add another brick to our mod, click the + button below the Quick Bar Action brick in the Brick Actions pipeline.
This opens a modal where you can search for bricks, like steps, to add to our mod. Search for the Extract from Page brick and click + Add to add it into our mod.
Click the mouse button next to the Selector field and click the part of the page you want to scrape info from. PixieBrix will automatically apply a selector.
We will need to add multiple properties to get a comprehensive profile. In this case, we want to grab the following:
the name of the person
their description under their name
the current company they’re at
To do this, when you click the selector, the screen turns purple as you mouse over parts. Select the area, then rename the property value to the corresponding info. To make another property, click the Add Property button.
Clicking the mouse turns on the selector grabber and highlights elements on hover over a page to help you better select the right data to scrape
If the selector doesn’t grab the right elements, copy and paste these values into the value field for each property.
You will need to manually copy and paste the value for the company as the default selector value from clicking the company element does not work consitently.
✅ Checkpoint! 🏁 You’re ready to move on if your Extract from Page brick looks like this.
You can also test it out by running your mod and previewing the output on the right-side panel. It should show values for each of those properties. To run your mod, use the keyboard shortcut you configured to toggle the Quick Bar, and select the action.
We’ll add another brick, just like we did before, by clicking the + icon below the Extract from Page brick in the Brick Actions panel. Search for the Create Text Completion with OpenAI brick and add it to the mod. There’s a lot you can customize and play around with here, and I encourage you to do that, but for the sake of the tutorial, start with these configurations and then tweak as you go.
The first two options (OpenAI integration and the model) will be configured by default so you can safely ignore these.
The next part is where it gets fun. This is where we’ll pass a prompt asking OpenAI to write a connection request. Feel free to customize the prompt however you’d like, but here’s what I’ve been using and found to work pretty well.
Change the second to last paragraph with information about yourself!
Notice that inside the prompt, we’re referencing the information we scraped from that user’s profile, allowing OpenAI to get to know the person and make a personalized note. To reference pieces from another brick, you simply need to wrap it like {{@data}}.
How do you find @data.description? Easy! Return to the brick you want to reference information from and go to the right side to view the Output Panel. This shows everything you can reference from the brick. Click the copy button to copy the name of the variable you want to reference.
⚡ Pro tip: This is also useful for debugging and ensuring you get the expected information from each brick in your mod.
After setting the prompt, I recommend using these settings to help fine-tune the results and give you a couple of options in case you don’t like the first one that’s generated. (Turns out even robots sometimes need a couple of tries to get things right! 🤖)
You can read more about these different settings in the descriptions below each field.
✅ Checkpoint! 🏁
You’re ready to move to the next step when you have set the prompt and updated the fields to values in the above screenshot.
We are nearly finished, but first, we need to create a way to view the response from OpenAI. To do this, we’ll add one more brick to our mod. Click the + button and search for the Display Temporary Information brick. Add it to the mod.
This brick has a few nested steps. The top step is primarily descriptive. You can change the Title field to anything you’d like. I chose something simple, like AI Assist.
Click into the brick inside this brick called Render Document.
This is where we style our data! 🎉
Click the text that says Example Document on the right side, and you’ll see a Header box appear. Type something else in the Title field to see it updated in the preview on the right side. I named mine 💌 Customized Connection Requests for {{@data.name}} and changed it to an H3 heading type. You can use the options below to customize the text further by centering it, changing the color, adding margin, and more.
Next, we’ll need to add a row containing a List for displaying our responses. Click the three dots in the preview panel. See a lot of dots? Click the second set from the right side, as the screenshot shows. Click Row. Then click the three dots in the newly created box, and select List.
That's a lot - here’s a visual walkthrough of adding a list element.
Click the newly created List in the preview, and you can configure the data you want to show.
The Array value should be set to loop through the info we got from the Open AI brick. You already know how to find the output of a brick, but ill save you a minute - copy @transformed.completions and leave the Element Key value as is, set to element.
Select the Text type of item, which creates a new text box inside your List. Select the text that says “paragraph text…”
Replace the paragraph text with {{@element}} . This tells it to show the element text for each item so that it will show the actual response.
✅ Final Checkpoint! 🏁
Are you ready? Time to test it out! Use your keyboard shortcut to toggle your PixieBrix Quick Bar. Select the action, and watch a sidebar show personalized connection request messages for you and the user on the page!
Demo of generating a personalized connection request from OpenAI
Not working? Don’t worry—it happens to the best of us. Here are some tips for debugging.
Not getting any information? Check that each step of the brick has an output. Using the last panel on the right, you should be able to preview the output of each brick. See if any brick isn’t returning output, and then check if the selectors or configuration on that brick matches what we have in the tutorial.
Sidebar pops up, but there’s no content? The Render Document brick is a bit tricky. We’re working on optimizing it to be a better experience for users, but it’s a lot to click through and can take a bit to get used to navigating in there. If you want to see how it should be configured, I recommend checking out the “source code”, aka this published mod in the marketplace. You can activate that mod, and then in your Page Editor, you’ll see precisely what that brick should look like to see what’s different in your setup. You can even copy the Display Temporary Information brick, paste it into your mod, and customize it.
Once you have completed these steps, you will have created a "mod" that uses PixieBrix and OpenAI to generate a customized LinkedIn connection request.
🙋Need some help with building? Head over to the Slack Community, and we’ll gladly help you!
Property
Selector Value
name
.text-heading-xlarge
description
#main .text-body-medium
company
#main ul:nth-child(2) button:first
