Functions

Functions extend the capabilities of your AI-agent by allowing you to implement custom logic and perform actions that go beyond the built-in features.

Your AI-agent can use them to execute various tasks. For instance, it can can display the AI-agent ID, format dates into readable strings, extract specific categories from user input, parse multi-part FAQ responses into manageable outputs, and much more.

Why use functions for your AI-agent:

Personalized Logic Integration: Functions enable you to add tailored logic that meets specific requirements, which might not be covered by default AI-agent capabilities.
Enhanced flexibility: By writing custom functions, you can handle unique scenarios, perform complex operations, and integrate with external systems.
Improved user experience: Functions can help deliver personalized and dynamic responses, leading to a more engaging and satisfying user interaction.
Seamless integration: Functions facilitate smooth integration with third-party APIs, databases, and other external services, extending the AI-agent's utility.

Create functions in AI-agent builder:

To create a function, follow these steps:

Go to Automation > Build > Functions > + Add new function.
Provide a specific name to your function and click on Add.

When creating functions, it's important to adhere to the following naming rules to ensure validity and consistency within your coding environment. Make sure to adhere to these naming conventions:

- Function names must contain a minimum of 2 characters.
- No special characters (e.g., `@`, `#`, `!`, `$`, etc.) or spaces are allowed.
- The function name should follow camelCase (e.g., `checkAgentAvailability`).
- Function names must be unique, with no duplicates.

Function arguments

In programming, arguments serve as inputs that allow the AI-agent to retrieve precise information or perform specific actions within your code, enabling more tailored and dynamic responses.

The table below provides the list of arguments you can use to access relevant information:

arg	Data type	Use
data.variables.<variable_name>	String	Retrieves the relevant AI-agent variable within the function based on the provided `variable_name`.
data.channel	String	Retrieves the current channel name of the AI-agent user like whatsapp, yellowmessenger, facebook, etc.
data.profile	Object	Retrieves properties of the current user such as name, number, email, city, country, and language.
data.sender	String	Retrieves the sender ID. For example, on WhatsApp, the sender ID is the user's phone number, while for website AI-agents, it's a long string representing the unique session ID.
data.bot	String	Retrieves the AI-agent ID specific to the environment (staging, production, or sandbox).
data.message	String	Retrieves the message inputted by the user during the conversation. This allows the AI-agent to respond to the latest input.
data.event.<event_name>	Object	Retrieves the data related to a specific event, allowing the AI-agent to handle scenarios like ticket creation or resolution.
ymLib.args.apiResponse	any	Retrieves and processes an API response within a parser function before assigning it to a variable.
ymLib.args.logger	Object	Retrieves and pushes the API response into the logs for debugging and tracking purposes.
context.history	Object	Retrieves the history of nodes that the user has passed through on the cloud platform.
prediction.intents	String	Retrieves predicted intents from the user message, based on the confidence level.
prediction.entities	String	Retrieves the list of predicted entities from the user message.

Parameters to access user data

The table below lists all parameters associated with accessing user information:

User property name	arg	Data type	Description
userId	data.user.userId	string	Retrieves the unique identifier of the user as configured in User 360. This could be a mobile number, email, or any other identifier. For more details, visit documentation.
firstName	data.user.firstName	string	Retrieves the first name of the user.
lastName	data.user.lastName	string	Retrieves the last name of the user.
gender	data.user.gender	string	Retrieves the gender of the user.
country	data.user.country	string	Retrieves the country of the user based on their IP address.
city	data.user.city	string	Retrieves the city of the user based on their IP address.
language	data.user.language	string	Retrieves the preferred language of the user.
timezone	data.user.timezone	string	Retrieves the timezone of the user based on their IP address.
email	data.user.email	email	Retrieves the email address of the user.
phone	data.user.phone	phone	Retrieves the phone number of the user with the country code.
lastChannel	data.user.lastChannel	string	Retrieves the most recent channel (set up on yellow.ai) where the user had an active session.
dob	data.user.dob	date	Retrieves the date of birth of the user.
emailOptin	data.user.emailOptin	boolean	Retrieves the subscription status of the user for emails. The value is`true` if subscribed, else `false`.
smsOptin	data.user.smsOptin	boolean	Retrieves the subscription status of the user for SMS. The value is `true` if subscribed, else`false`.
whatsAppOptin	data.user.whatsAppOptin	boolean	Retrieves the subscription status of the user for WhatsApp messages. The value is `true` if subscribed, else `false`.
tags	data.user.tags	list	Retrieves the list of labels associated with the user.

Writing AI-agent function code

You can define a custom function using JavaScript. Here is a sample code snippet. Just replace the placeholder 'Your logic goes here' with the specific functionality you wish to implement.

return new Promise(resolve => {  
      // Your logic goes here
      resolve();
  }); 

How functions work:

This section provides an overview of how functions operate within your AI-agent.

We will begin with basic examples and gradually move to more complex scenarios, allowing you to build up your knowledge and skills.

In the code examples provided below, you'll see some fundamental JavaScript terms such as let, resolve, if/else, and console.log:

let: This keyword is used to declare a variable that can hold data.
resolve: This function signals the completion of a promise by providing the output from the function.
if/else: This construct enables the AI-agent to make decisions based on specified conditions.
console.log: This command is used to display messages or data, which is helpful for debugging.

1. Display AI-agent ID (Basic)

Here is a sample code snippet to display the current bot ID:

return new Promise(resolve => {
    let botId = data.bot;  // Retrieve the bot ID from the data object
    resolve(`The current bot ID is ${botId}`); // Return a message containing the current bot ID
});

Explanation:

Access Bot ID: The botId is extracted from the data object.
Display Message: The function generates a message with the bot ID to be shown as the output.

Output: The current bot ID is [botId]

2. Log current date and time (Intermediate)

The below code snippet captures the current date and time in the Asia/Kolkata time zone. It formats the time accordingly and logs both the date and time. The function then returns the current date.

return new Promise(resolve => {
    let currentdate = new Date();
    let currentTime = new Date().toLocaleTimeString('en-IN', { timeZone: 'Asia/Kolkata' });
    console.log("timestamp#########"); // Outputs a marker for timestamp logging.
    console.log(currentdate);   // Logs the value of currentdate.
    console.log(currentTime);  //  // Logs the formatted current time.
    resolve(currentdate); // // Resolves the promise with currentdate, indicating the operation is complete.
});

Explanation:

Get Current Date: The currentdate variable captures the system's current date and time.
let currentTime: This declares a variable named currentTime using the let keyword, allowing it to be reassigned if necessary.
- new Date(): This creates a new Date object that represents the current date and time.
- .toLocaleTimeString('en-IN', { timeZone: 'Asia/Kolkata' }): This method formats the time according to the specified locale and options:
- 'en-IN': This indicates the desired locale, in this case, English as used in India.
- { timeZone: 'Asia/Kolkata' }: This option specifies the time zone to be used for formatting, ensuring that the returned time corresponds to Indian Standard Time (IST).
Log Information: The complete date and the formatted time are logged for printing both currentdate and currentTime to the console, allowing you to confirm that the correct date and time are being captured and formatted as expected.
Return Date: The function returns the system's current date and time.

Output: Logs the current date and time in 'Asia/Kolkata' timezone and returns the current date.

3. Check email verification status (Intermediate)

This function checks whether the user's email address has been verified. It retrieves the user's email verification status and returns an appropriate message.

return new Promise(resolve => {
    let emailVerified = data.user.emailOptin; // retrieves the email verification status from the data object.
    
    if (emailVerified) { // Checks if the email is verified.
        resolve("Your email address has been successfully verified."); // if verified, resolves the promise with a success message.
    } else {
        resolve("Your email address is not verified. Please verify your email to continue."); // else, resolves the promise with a failure message.
    }
});

Explanation:

Retrieve Email Verification Status: The emailVerified variable gets the email verification status from the user's properties.
Check Verification: The function checks if the email is marked as verified.
Return Message: Depending on the verification status, it returns a confirmation message or a prompt to verify the email.

Output: A message stating whether the user's email is verified or not, based on their stored email verification status.

Complex use cases

4. Check category match (Advanced)

The following code snippet checks if the text provided by the user matches with any of the categories listed in the data.

It processes the categories and the user text, checks if the text includes the word "and" (replacing it with "&"), and then compares the processed text against each category. It returns "yes" if there is a match and "no" otherwise.

return new Promise(resolve => {
  let categories = data.variables.query_result.records[0].main_categories; // Get the list of categories from the AI-agent's data
  let userTextData = data.message; // Capture the user's input message
    
    
  // If the user's input contains 'and', replace it with '&'  
  if (userTextData.includes('and')) { 
    userTextData = userTextData.replace('and', '&');
  }
  console.log("userTextData??", userTextData); // Log the modified user input for debugging
  let cat_arr = categories.split(",");
  const arr_length = cat_arr.length;
    
    
  // Loop through the categories array to find a match with user input
  for (let i = 0; i < arr_length; i++) { /
    if (cat_arr[i].trim().toLowerCase() == userTextData.toLowerCase()) {  // Compare each category (case-insensitive) with the user's input
      return resolve("yes"); // If a match is found, return "yes"
    }
  }
  return resolve("no"); // If no match is found, return "no"
});

Explanation:

Retrieve Categories: The categories variable gets the list of categories from the data.
Process User Text: The userTextData is checked for the word "and", which is replaced with "&" if found.
Split Categories: The categories string is split into an array of category strings (cat_arr).
Check for Match: The function compares the user-provided text with each category in a case-insensitive manner.
Return Result: If a match is found, it resolves with "yes"; otherwise, it resolves with "no".

Output: "yes" or "no" based on whether the user text matches any category.

5. Dynamically add authorization token (Advanced)

This function retrieves the authorization token from an API response and prepares it for use in future requests.

It extracts the auth_token from the API response, allowing for automated token management without the need for manual entry of the token key.

return new Promise(resolve => {
  let { apiResponse } = ymLib.args; // Retrieve API response
  let token = JSON.parse(apiResponse.body); // Parse the response body to extract token

  console.log(token); // Print the token for debugging

  // Return the token in the required format
  return resolve(token.token_type + " " + token.access_token);
});

Explanation:

Retrieve API Response: The apiResponse variable gets the response data from the API call.
Parse Response Body: The response body is parsed using JSON.parse to extract the token_type and access_token.
Log Information: The token is printed to the console for debugging.
Return Token: The function gives back the token in the correct format.

Output: The function will return the extracted authorization token.

6. Checking agent or service availability by time zone (Complex)

Here is a sample code snippet to check whether an agent or a service is available based on the current time in a specific time zone (Asia/Kolkata).

It captures the current time in the Asia/Kolkata time zone and determines if the time falls within the business hours (between 9 AM and 11 PM). Based on this, it either returns "ALLOW" (if within the allowed hours) or "Not ALLOW" (if outside the allowed hours).

return new Promise(resolve => {
  let dates = new Date().toLocaleTimeString('en-IN', { timeZone: 'Asia/Kolkata', hour12: false }); // Get the current time in the Asia/Kolkata time zone in 24-hour format
  let status;  
  let hour = dates.split(':')[0]; // Extract the hour from the time string
    
  // Check if the hour is between 9 AM (09:00) and 11 PM (23:00)
  if (hour >= 9 && hour <= 23) {  
    status = "ALLOW"; // Action/service is allowed
  } else {
    status = "Not ALLOW"; // Action/service is not allowed
  }

  return resolve([status]); // Return the status as an array
});

Explanation:

Fetch Current Time: Retrieves the current hour in the "Asia/Kolkata" timezone using 24-hour format.
Check Time Window: If the time is between 9 AM and 11 PM(business hours), the AI-agent will allow the action or service; otherwise, it will restrict it.
Return Status: Sends a message indicating whether the action/service is allowed or not.

Output: "ALLOW" or "Not ALLOW" based on the current time.

7. Split FAQ response (Complex)

This function splits a FAQ response into two parts and returns them as an array.

It processes the faq_response data, splits it at the $ symbol, and returns the two parts as an array.

return new Promise(resolve => {
    let res = data.variables.faq_response; // Get the FAQ response from the input data
    
    // Split the response at the '$' symbol into two parts
    let responses1 = res.split('$')[0];
    let responses2 = res.split('$')[1];

    // Optional: Log the responses for debugging purposes
    console.log("reeeeeee%%%%%%");
    console.log(responses1);
    console.log(responses2);
    
    let responses = [responses1, responses2];  // Combine both parts into an array
    
    resolve(responses); // Return the array of responses
});

Explanation:

Retrieve FAQ Response: The res variable gets the FAQ response from the data.
Split Response: The res string is split into two parts using $ as the delimiter, resulting in responses1 and responses2.
Log Information: Logs the split parts for debugging purposes.
Return Responses: The function returns the two split responses as an array.

Output:

An array containing the two parts of the FAQ response, such as [responses1, responses2].

Once you've entered your custom logic in the available editor, referring to the arguments, parameters and custom code examples provided above, click on Save.

If there are any errors in the code, you can view them under Warnings at the bottom. Address the displayed warnings/errors if any.

Additional options for functions

This section has some funcionalities common to the entire AI-agent that will help you to adjust your AI-agent's performance.

Tools

This is a global testing and settings tab. It allows you to test your AI-agent's responses by entering user inputs. For example, if you type "speak to support" or "schedule a meeting," it predicts the most relevant intents along with the associated confidence score, helping you understand how effectively your AI-agent can recognize user utterances. Click here to learn more.

Compare (the function between Staging & Production environment)

This option compares the function between staging and production environments depending on the environment the AI-agent is currently in.

Flows

Flows streamline the management of conversation paths, allowing you to make edits without switching back to the build section. For example, while working on the "Book a Test Drive" flow, you can see relevant nodes like text and carousel. By entering/copying the journey name, you can locate all nodes, modify them in functions, and any updates will be automatically reflected in the flow. This makes managing your AI-agent’s conversations much more efficient.

Localization

The Localization feature allows you to easily add translation capabilities to your AI-agent. Instead of creating separate workflows for each language, it offers a unified solution. For instance, the phrase "Hi, how are you?" can be translated into various languages, like English, Hindi, and Tamil, ensuring users receive messages in their preferred language. For more details, visit documentation.

Using functions in flows

Functions are used in flows via the Function node.

The following is an example where the function node is used to convert minutes to seconds.

Create a flow with the Question node to take an input from the user and store the response in a variable, pizzamin.

Go to Functions, pass this variable and write a code to convert minutes to seconds.

For example,

return new Promise(resolve => {
    let minutesTaken = data.variables.pizzamin;
    let secondsTaken = minutesTaken * 60;
    let stringSeconds = secondsTaken.toString();
    resolve(stringSeconds);
  });                          
  

Include a function node, pass the function created in the previous step, and store the response in another variable, pizzasec.

To display the result to the end user, include a text node and choose the pizzasec variable.

Result:

What Next?

Create your own function and use it by adding an API node in the flow.
You can visit our community and share your ideas with other fellow AI-agent builders.

Why use functions for your AI-agent:​

Create functions in AI-agent builder:​

Function arguments​

Parameters to access user data​

Writing AI-agent function code​

How functions work:​

1. Display AI-agent ID (Basic)​

2. Log current date and time (Intermediate)​

3. Check email verification status (Intermediate)​

Complex use cases​

4. Check category match (Advanced)​

5. Dynamically add authorization token (Advanced)​

6. Checking agent or service availability by time zone (Complex)​

7. Split FAQ response (Complex)​

Additional options for functions​

Tools​

Compare (the function between Staging & Production environment)​

Flows​

Localization​

Using functions in flows​

Why use functions for your AI-agent:

Create functions in AI-agent builder:

Function arguments

Parameters to access user data

Writing AI-agent function code

How functions work:

1. Display AI-agent ID (Basic)

2. Log current date and time (Intermediate)

3. Check email verification status (Intermediate)

Complex use cases

4. Check category match (Advanced)

5. Dynamically add authorization token (Advanced)

6. Checking agent or service availability by time zone (Complex)

7. Split FAQ response (Complex)

Additional options for functions

Tools

Compare (the function between Staging & Production environment)

Flows

Localization

Using functions in flows