Integrate ChatGPT

By integrating ChatGPT technology, you can build a custom chatbot capable of handling a variety of situations, allowing customers to easily interact with your company. This article provides a series of simple steps explaining how to apply for an OpenAI API Key and integrate ChatGPT into the system so your chatbot can provide smarter and more efficient support and responses, thereby improving customer satisfaction.

Webhook simulation case limitations and commercial application assessment reminder

Please note that the following examples are only technical simulation cases. If you want to apply them to real commercial scenarios, be sure to evaluate the following limitations:

1. Webhook concurrency limits: In high concurrency situations, it may not support all conversation counts.

2. Webhook response timeliness: If you cannot respond within an extremely short time, the request will be directly rejected by the system.

3. No context support: This method is only suitable for single question-and-answer scenarios and cannot handle multi-turn conversations.

4. Webhook parameter character limit:params has a limited number of characters and cannot run in parallel with the full content of the system's built-in knowledge base.

....etc., be sure to evaluate according to your business needs to avoid encountering performance bottlenecks or functional limitations in practical applications.

Apply for an OpenAI key

1. Go to the official OpenAI website (https://openai.com/), click the "Get started for free" button to register and log in.

2. The first step to integrate the OpenAI API is to register an account. Just log in to the OpenAI website, click the account icon in the top right corner, then choose View API keys to enter the API page.

3. On the API page, just click Create new secret key to get a key. Be sure to save this key securely, as it will only appear once; if lost, you will need to request a new one.

Create a Webhook script

Add script

First, we need to use the "Webhook Script" feature to call OpenAI's API. You can find this feature under the Integrations category of Applications. Click "Add" and choose "Conversation Script" to match your usage scenario type, then you can name the script, for example "ChatGPT Integration."

Enter request data

Next, you need to enter the integration settings according to the following conditions:

1. Call URL:https://api.openai.com/v1/chat/completions

2. HTTP request method: POST

3. Headers: choose Authorization and enter Bearer <your API key>

Finally, enter the most critical params:

// You can sequentially adjust the settings needed for chatGPT
{
  "model": "gpt-3.5-turbo",
  "max_tokens": 1000,
  "temperature": 0.5,
  "messages": [
    // Of course, you can also adjust ChatGPT's dialogue content by passing in more specific message content.
    // {
    //  "role": "system",
    //  "content": "You are now a XXX, you should provide suggestions for the questions I ask."
    // },
    {
      "role": "user",
      "content": "{{ question }}"
    }
  ]
}

Set Callback actions

Finally, click the "Advanced" tab to set how to handle the data returned by the API. These settings will be used for variable data in the "Conversation Script."

// You can adjust different variable names according to your needs,
// This indicates storing each returned choice into the answers and content variables.
{
  "answers": "{{ choices | this_is_array }}",
  "content": "{{ choices.0.message.content }}"
}

Create a conversation script

We can easily integrate ChatGPT to add a simple conversation flow to the customer service system. As long as a customer sends any message, ChatGPT will respond and complete the expected flow.

Of course, we can also add different conditions and actions so customers can ask more diverse questions, for example guiding customers with options to ask different questions, or automatically routing to a human agent when a customer mentions specific keywords.

1. Add an "Event" node at the beginning of the script and choose "When customer sends a message," set "Customer message matches" to "*", meaning continue execution whenever any message is received.

2. Next, add an "Action" node and choose "Webhook Script," selecting the "Webhook Script" created in the previous section.

3. Add another "Action" node and choose "Send Message," enter "{{ content }}" in the message content to indicate replying to the customer using the content variable.

4. Finally, simply bind the channels you want to use (for example LINE Official Account, live chat) to this script, and you can start using ChatGPT to provide a better service experience for your customers!

5. 🎉 Start the conversation!