Contents
- 1 What is GPT1?
- 2 How Does the System Work?
- 3 The Automation Process
- 4 Monetizing the Co-Pilot
- 5 Example Use Case: LinkedIn Genie
- 6 Building the Co-Pilot
- 7 Conclusion
- 8 Creating a Webhook
- 9 Creating a Chrome Extension with ChatGPT
- 10 Creating the Extension Files
- 11 Creating the Extension Files Manually
- 12 Loading the Extension into Chrome
- 13 Testing the Extension
- 14 Styling the Chat Window
- 15 📊 Testing the Automation
- 16 📝 Adding a Search Record Module
- 17 Verifying API Keys
- 18 Using a Router to Determine Outcomes
- 19 Setting up a Filter
- 20 Sending Web Hook Responses
- 21 Example Use Case
- 22 Code Snippet
- 23 Additional Tips
What is GPT1?
GPT1 is a leaping capability that uses Advanced TR of thought processes. It is outperforming GBT-40 on many benchmarks and is being used by PhD students in physics, chemistry, and biology.
„GPT1 is a preview of a new AI system that is significantly better than GBT-40. It takes some time to think, but the quality of its responses is impressive.“
How Does the System Work?
The system consists of the following components:
| Component | Description |
|---|---|
| GPT1 | The AI system that powers the co-pilot |
| Airtable | A database that stores user information and API keys |
| Stripe | A payment gateway that handles subscriptions |
| Chrome Extension | The co-pilot that users interact with |
The Automation Process
The automation process consists of the following steps:
- User Installs Chrome Extension: The user installs the Chrome extension and is prompted to enter an API key or purchase a subscription.
- User Purchases Subscription: The user purchases a subscription through Stripe and receives an email with their API key.
- API Key is Stored: The API key is stored in Airtable and linked to the user's information.
- User Enters API Key: The user enters their API key in the Chrome extension and gains access to the co-pilot.
- Automation Checks API Key: The automation checks the API key to ensure it is valid and linked to a paying subscriber.
Monetizing the Co-Pilot
The co-pilot can be monetized by charging users a subscription fee to access its features. The subscription fee can be set to any amount, and the automation will handle the payment processing and API key validation.
Example Use Case: LinkedIn Genie
The LinkedIn Genie is a Chrome extension that uses GPT1 to generate responses to comments and posts. The user can interact with the extension by clicking on a button and entering a prompt.
| Feature | Description |
|---|---|
| Comment Generation | The extension generates responses to comments and posts |
| Post Summarization | The extension summarizes posts and provides bullet points |
| Chat with Videos | The extension allows users to chat with videos and ask questions |
Building the Co-Pilot
The co-pilot can be built by following these steps:
- Create a Chrome Extension: Create a Chrome extension that interacts with GPT1 and Airtable.
- Set up Stripe: Set up a Stripe account to handle payment processing.
- Create an Airtable Database: Create an Airtable database to store user information and API keys.
- Configure the Automation: Configure the automation to handle API key validation and payment processing.
Conclusion
Building an AI-powered co-pilot with GPT1 is a powerful way to monetize AI technology. By following the steps outlined above, you can create a co-pilot that provides value to users and generates revenue through subscription fees.## Building a Chrome Extension with ChatGPT 🤖
Creating a Webhook
A webhook is a callback function that is triggered by a specific event. In this case, we will use a webhook to send the conversation history to a specific URL.
To create a webhook, we need to:
- Go to me.com and create a new automation builder
- Name the automation „YouTube Conversation AI“
- Create a new webhook and name it „Custom Webhook“
- Copy the webhook URL and save it for later use
Creating a Chrome Extension with ChatGPT
To create a Chrome extension with ChatGPT, we need to:
- Open ChatGPT and create a new prompt
- Ask ChatGPT to develop a Chrome extension that enables us to chat with any YouTube video
- Provide the webhook URL and ask ChatGPT to send the conversation history to the webhook URL
- Ask ChatGPT to create all the necessary files separately
Creating the Extension Files
ChatGPT will create the following files:
| File Name | Description |
|---|---|
manifest.json |
The manifest file that contains metadata about the extension |
content.js |
The content script that interacts with the YouTube video |
styles.css |
The stylesheet that styles the chat window |
popup.html |
The HTML file that contains the chat window |
Creating the Extension Files Manually
If we want to create the extension files manually, we need to:
- Create a new folder on the desktop and name it „YouTube AI“
- Create a new file called
manifest.jsonand paste the code into it - Create a new file called
content.jsand paste the code into it - Create a new file called
styles.cssand paste the code into it - Create a new file called
popup.htmland paste the code into it
Loading the Extension into Chrome
To load the extension into Chrome, we need to:
- Go to the Chrome extensions page and toggle developer mode
- Click on „Load unpacked“ and select the „YouTube AI“ folder
- The extension should now be loaded into Chrome
Testing the Extension
To test the extension, we need to:
- Go to a YouTube video and click on the chat button
- Type a message and send it
- The conversation history should be sent to the webhook URL
Styling the Chat Window
To style the chat window, we can ask ChatGPT to make the following changes:
- Make the window larger
- Make the gradient larger
- Increase the height and size of the chat and send message button
- Make the send message button circular
// Example code for styling the chat window
const chatWindow = document.getElementById('chat-window');
chatWindow.style.height = '500px';
chatWindow.style.width = '300px';
const sendButton = document.getElementById('send-button');
sendButton.style.borderRadius = '50%';
```## 📹 Integrating a GPT Bot with a YouTube Video Transcript
### Setting Up the GPT Bot
To integrate a GPT bot with a YouTube video transcript, we need to send the transcript to the web hook along with the messages. This will enable the bot to analyze the conversation and provide relevant responses.
### Creating a Prompt Completion
To create a prompt completion, we need to:
* Set the **max tokens** to 0
* Choose a **model**, such as GPT-3
* Add a **message** to the prompt
The prompt should include the following information:
> "You're an AI designed to engage in conversations about YouTube videos. You will receive two pieces of information: a transcript of the YouTube video and the conversation history. Your task is to analyze both the transcript and the conversation history and provide the next appropriate response in the chat."
### Adding Conversation History
To add conversation history, we need to:
* Copy the conversation history and paste it into the prompt
* Repeat this process for each conversation turn
This will enable the bot to keep track of the conversation and provide more accurate responses.
### Setting Up Stripe for Payment
To set up Stripe for payment, we need to:
* Create a new account on Stripe
* Create a new product for the subscription
* Set the price for the subscription
* Create a payment link
### Creating a Payment Link
To create a payment link, we need to:
* Go to the product catalog on Stripe
* Click on the product we created
* Click on "Create link"
* Copy the payment link
### Testing the Payment Link
To test the payment link, we need to:
* Enter fake payment information
* Click on "Subscribe"
* Check if the payment is successful on the Stripe dashboard
### Retrieving the API Key
To retrieve the API key, we need to:
* Go to the developers section on Stripe
* Click on "API Keys"
* Create a restricted key
* Choose the permissions for the key
* Create the key
### Setting Up the Web Hook
To set up the web hook, we need to:
* Go to the web hook section on Mixar
* Click on "Add a new web hook connection"
* Choose the Stripe API key
* Enter the API key
* Save the web hook
### Retrieving a Customer
To retrieve a customer, we need to:
* Go to the Stripe module on Mixar
* Click on "Retrieve a customer"
* Enter the customer ID
* Save the customer information
| **Stripe Module** | **Description** |
| --- | --- |
| Create a product | Create a new product for the subscription |
| Create a payment link | Create a payment link for the product |
| Retrieve a customer | Retrieve a customer's information using their ID |
| **Mixar Module** | **Description** |
| --- | --- |
| Watch events | Watch events for the Stripe API key |
| Web hook | Set up a web hook for the Stripe API key |
| Retrieve a customer | Retrieve a customer's information using their ID |## 📝 Creating a Customer Database in Airtable
To create a customer database in Airtable, we need to set up a new table with the following fields:
| Field Name | Description |
| --- | --- |
| Name | The customer's name |
| Email | The customer's email address |
| Unique ID | A unique identifier for the customer |
| Membership Status | The customer's membership status (e.g. paying subscriber, declined, churned) |
We can add these fields to our Airtable table by clicking on the "Add a field" button and selecting the corresponding field type.
## 📝 Creating a Formula for the Unique ID
To create a unique ID for each customer, we can use a formula in Airtable. The formula is simply `RECORD_ID()`, which returns the unique ID of the record.
## 📧 Creating a Custom Email Automation
To send a custom email to each customer, we need to create a new automation in Airtable. We can do this by clicking on the "Create a new automation" button and selecting the "Send an email" action.
The automation should be triggered when a new record is created in the customer database, and the email should be sent to the customer's email address.
## 📝 Creating the Email Content
The email content should include a welcome message and the customer's unique ID, which will serve as their API key.
> "Your API adventure begins here! Your API key is: [Unique ID]"
## 📊 Adding the API Key to the Chrome Extension
To add the API key to the Chrome extension, we need to modify the code to include a mini portal that asks the user to enter their API key. The API key should be stored in the Chrome browser and sent with every conversation to validate the user.
## 📝 Updating the Chrome Extension Code
To update the Chrome extension code, we need to add the following functionality:
* Show a mini portal that asks the user to enter their API key
* Store the API key in the Chrome browser
* Send the API key with every conversation to validate the user
```javascript
// Add a mini portal to ask the user to enter their API key
// Store the API key in the Chrome browser
// Send the API key with every conversation to validate the user📊 Testing the Automation
To test the automation, we can create a new record in the customer database and verify that the email is sent with the correct API key.
📝 Adding a Search Record Module
To add a search record module, we need to create a new automation in Airtable that searches for records in the customer database based on the API key.
| Field Name | Description |
|---|---|
| API Key | The API key to search for |
| Formula | The formula to use for the search (e.g. =API Key) |
We can add this module to our automation to search for records in the customer database and validate the user's API key.## API Key Verification 🚀
API Key: A unique string used to authenticate and authorize access to a specific service or application.
„An API key is like a special password that allows a user to access a particular service or application.“
Verifying API Keys
To verify an API key, we need to:
- Check if the API key exists in the database
- Check if the API key is valid and active
Using a Router to Determine Outcomes
We can use a router to determine two different outcomes:
| Outcome | Condition |
|---|---|
| Paying Customer 💸 | Unique ID exists |
| Error Response 🚫 | Unique ID does not exist |
Setting up a Filter
To set up a filter, we can:
- Right-click and select „Set up a filter“
- Name the filter (e.g. „Paying Customer“)
- Set the condition to „Unique ID exists“
Sending Web Hook Responses
We can send different web hook responses based on the outcome:
| Outcome | Web Hook Response |
|---|---|
| Paying Customer 💸 | Send a response with the requested data |
| Error Response 🚫 | Send an error message with a link to purchase a subscription |
Example Use Case
Here's an example of how the API key verification process works:
- A user enters an invalid API key
- The system checks the API key and determines that it is invalid
- The system sends an error message with a link to purchase a subscription
- The user purchases a subscription and updates their API key
- The system checks the new API key and determines that it is valid
- The system sends a response with the requested data
Code Snippet
// Example of a web hook response
const webhookResponse = {
"text": "You must have an active subscription. Please purchase here: <link>",
"status": 403
};Additional Tips
- Use GPT-1 or GPT-1 Mini for faster and more efficient processing
- Check the community resources for blueprints, agents, and guides
- Experiment with different configurations to customize the user experience
