This guide explains in detail how to integrate a neural network into your automation, that will interact with clients, communicate, process data, perform actions, or carry out mathematical calculations.
Note: To use any ChatPlace features involving the neural network, you must have a Pro Plan with the additional ChatPlace AI package, or an Unlimited Plan, which already includes all add-ons.
More about ChatPlace plans can be found here.
How to create an action
Click the "+" button in the top-right corner of the automation builder
Select the "Action" block
Click on the Action block that was just added
Choose the action "ChatPlace AI Request"
How the "ChatPlace AI Request" action works
Role of the neural network
The role helps make the request to the neural network more precise and the generated response more specific.
By setting a role, you define who or what the neural network should "see itself as." This parameter is optional, but we recommend specifying a role for any type of request.
Important: When specifying a role, limit it to a profession, character, or tool, and briefly describe the responsibilities or functions the neural network should perform. Any detailed instructions should be placed only in the neural network’s knowledge base.
Example 1: You are a calculator. Perform mathematical calculations.
Example 2: Take on the role of a professional copywriter who creates content for different platforms such as Instagram, Twitter, or YouTube.
Example 3: Take on the role of an AI assistant for the Instagram account @kendalljenner. Your main task is to respond to Direct messages on my behalf (as if you were Zendaya). Reply in the same language and tone as the message you receive.
Neural network knowledge base
The Knowledge base is a set of instructions for the neural network. This is where you write the full prompt. In simple terms, it’s the guide for how the AI should work, communicate, and any important information about the company, product, person, etc. The knowledge base should include all the data the neural network might need when interacting with clients.
Important: Try to keep the knowledge base within 5,000 characters. This will help the AI respond more quickly and accurately.
If your knowledge base is large, write it in English. This will help the AI better understand user queries and reduce the resources needed to generate responses. You can simply include a note in the instructions asking the AI to reply in the language the client is using.
If the base is very large, you can use another neural network to compress the text, remove unnecessary parts, and translate it into English if needed.
Make sure to specify the variable that contains the client’s message or data in the neural network knowledge base.
The structure of the knowledge base might look something like this:
Write your prompt as detailed as possible, including all necessary information.
Example 1: Write a sales-style Instagram post on the topic “How to sell through Reels?” The main audience of the blog is men aged 20 to 30 who are interested in video production. The post should be no more than 100 words. Use emojis and include relevant hashtags. Use the following keywords: money, online, courses, scale, video.
Example 2: Use the FAQ to find the answer to the question. If the question is not in the FAQ, respond with “Unknown.” Don’t hesitate to use emojis.
FAQ:
Q: What’s your name?
A: My name is Dima Torgov.
Q: Where do you currently live?
A: I currently live in Lisbon.
Q: When did you move to Portugal?
A: In September 2022.
Question: {{ question }}
How to save the neural network’s response
After setting up the neural network's knowledge base, you need to choose the variable where the neural network’s response will be saved:
You can then use this variable in a message to display the neural network’s response in the conversation with the client. Alternatively, you can reference the variable in the next "ChatPlace AI request" block if additional processing is needed, or check the neural network’s response using a condition (for example, to move the client to a different branch of the scenario).
Thus, the "ChatPlace AI Request" action uses at least two variables: one variable, for example {{ question }}, is specified in the knowledge base and contains the data the neural network will work with; the second variable, for example {{ answer }}, is used to store the neural network’s response.
Additional parameters that allow for better interaction between the client and the neural network.
Without message history
This parameter disables the conversation history between the client and the neural network. Each client message will be treated as a new, standalone request, with no connection to previous messages.
It is not recommended to disable history for chatbots that interact with users, as the bot will not remember the conversation history with the client. This may lead to communication issues and make it harder to work with potential customers.
Optimize all requests
This is an internal feature developed by ChatPlace. When enabled, the neural network will rely only on the internal knowledge base and will not use ChatGPT’s broader capabilities. This parameter helps optimize requests to the neural network and focus them specifically on your company. It’s recommended to enable this option in all cases where the client is expected to interact only with the knowledge base, without needing external information.
How to use the "ChatPlace AI request" action in automations
Save all the necessary data that the neural network will work with, for example, this could be the client’s message:
Or a click on the button:
Write the request to the neural network, specifying the task and the necessary variables:
Save the neural network’s response in a variable that will be used to display the answer in a message to the client:
An example of how to use the neural network can be found in the Company Assistant template.
You might find useful:
Working with variables
Payment processing and verification