Using AI to Generate Code (Ai Node) – Latenode

An AI assistant is available inside the JavaScript node. It helps you generate, modify, and debug code directly inside your scenario — just by describing what you want in natural language.

The assistant works contextually: it understands the current node, available data, and even connections to other blocks. You can ask it to build logic, create inputs, work with global variables, integrate with APIs, and more.

Where the AI Assistant Is Available

You can launch AI Assistant in two ways:

1.Via the "AI Node" button at the bottom of the editor. This opens a dedicated chat window with assistant.

2.Inside any JavaScript node. Open the node and enter your message to start a conversation.

Table of contents: What Assistant Can Do

Create custom input parameters

Process data from other nodes

Work with global variables

Integrate with services like Telegram (example of creating custom api calls)

Debug and explain existing code

And more — just describe your task!

1. Custom Parameters

Assistant can automatically create custom input parameters for your nodes.

Instead of hardcoding a value:

const message = "Hello world";

Copilot will rewrite it to:

const message = input.text_input;

And it will automatically add the text_input field to the node UI — as shown in the screenshot above.

More info: Custom JS Parameters

2. Request to process data from other nodes

❓

Generate code to process data from the Trigger on Webhook node so that based on the Name and Count parameters, text is generated based on the following pattern: Hi, Name! Count of your scenarios has been executed in X on the Latenode platform! Instead of X, substitute the name of the current month.

The assistant will provide the answer:

To generate the required text based on the parameters from the "Trigger on Webhook" node, you can follow these steps inside the provided run function. First, we'll extract the Name and Count values from the data provided by the webhook node with ID 4. Then, we'll use JavaScript's Date object to get the current month and format it as a name. Finally, we'll construct the message according to the pattern you've requested.

Here is how you can do it:

export default async function run({execution_id, input, data, store}) {
    // Extracting Name and Count from the webhook data
    const name = data["{{4.`body`.`Name`}}"];
    const count = data["{{4.`body`.`Count`}}"];

    // Getting the current month as a name
    const monthNames = ["January", "February", "March", "April", "May", "June", "July", "August", "September", "October", "November", "December"];
    const currentMonthName = monthNames[new Date().getMonth()];

    // Generating the message
    const message = `Hi, ${name}! ${count} of your scenarios has been executed in ${currentMonthName} on the Latenode platform!`;

    // Returning the generated message
    return {
        message
    };
}

Explanation:

data["{{4.body.Name}}"] and data["{{4.body.Count}}"] access the Name and Count values from the webhook's body.

We use JavaScript's Date object to get the current month as an index (getMonth() returns a zero-based index for the months), and then we use this index to get the month's name from the monthNames array.

The message is constructed using template literals, embedding the name, count, and currentMonthName variables into the predefined text pattern.

Finally, the function returns an object containing the generated message, which can be used by subsequent nodes in your scenario.

3. Request for processing global variables

💡

It is assumed that global variables GlobalVar_number1 and GlobalVar_number2 exist in the account. See the section on global variables.

❓

Generate code that adds up the values of two global variables GlobalVar_number1 and GlobalVar_number2 and return an array of the values: sum of variables, GlobalVar_number1, GlobalVar_number2. Get global variables using store.getGlobalVariable

The assistant will provide the answer:

To accomplish the task you've described, we will use the store.getGlobalVariable method to retrieve the values of the global variables GlobalVar_number1 and GlobalVar_number2. Then, we'll add these values together, and return an array that includes the sum, GlobalVar_number1, and GlobalVar_number2 in that order. Here's how you can do it:

export default async function run({execution_id, input, data, store}) {
    // Retrieve the values of the global variables
    const number1 = await store.getGlobalVariable('GlobalVar_number1');
    const number2 = await store.getGlobalVariable('GlobalVar_number2');
    
    // Convert the retrieved values to numbers, if they are not already
    const num1 = parseFloat(number1);
    const num2 = parseFloat(number2);

    // Calculate the sum of the two numbers
    const sum = num1 + num2;

    // Return an array with the sum, GlobalVar_number1, and GlobalVar_number2
    return {
        result: [sum, num1, num2]
    };
}

This code snippet does the following:

It uses store.getGlobalVariable to fetch the values of GlobalVar_number1 and GlobalVar_number2. These methods are asynchronous, so we use await to wait for the promises to resolve.

It then converts these values to numbers using parseFloat to ensure that arithmetic operations can be performed correctly. This is a safeguard in case the values are stored as strings.

It calculates the sum of the two numbers.

Finally, it returns an object with a result key that maps to an array containing the sum, GlobalVar_number1, and GlobalVar_number2.

Ensure that the global variables GlobalVar_number1 and GlobalVar_number2 are set and contain numeric values. If these variables contain non-numeric values, parseFloat will return NaN, and the sum calculation will not work as expected.

4. Request to connect to Telegram to send messages (example of creating custom api calls)

❓

Generate a code with a Telegram connection to send a message to the desired chat room with a given ID

The assistant will provide the answer:

To send a message to a Telegram chat room, you will need to use the Telegram Bot API. First, ensure you have created a bot and obtained your bot's API token from BotFather in Telegram. Then, you can use the axios library to make HTTP requests to the Telegram API. Here’s how you can do it:

Install Axios: Since your environment automatically installs npm packages, you just need to import it.

Use the Telegram Bot API: You'll need to use the sendMessage method of the Telegram Bot API to send a message to a chat.

Here's an example code snippet that sends a message to a specified chat ID using a Telegram bot:

import axios from 'axios';

export default async function run({execution_id, input, data, store}) {
    // Your Telegram Bot Token
    const botToken = 'YOUR_BOT_TOKEN_HERE';
    
    // The chat ID you want to send a message to
    const chatId = 'YOUR_CHAT_ID_HERE';
    
    // The message you want to send
    const message = 'Hello, this is a message from my bot!';
    
    // Telegram API URL to send messages
    const apiUrl = `https://api.telegram.org/bot${botToken}/sendMessage`;
    
    try {
        // Making a POST request to the Telegram API to send the message
        const response = await axios.post(apiUrl, {
            chat_id: chatId,
            text: message,
        });
        
        // Returning the response from the Telegram API
        return {
            success: true,
            response: response.data,
        };
    } catch (error) {
        // Returning the error if the request failed
        return {
            success: false,
            error: error.message,
        };
    }
}

Replace 'YOUR_BOT_TOKEN_HERE' with your actual Telegram bot token and 'YOUR_CHAT_ID_HERE' with the chat ID you want to send a message to.

Important Notes:

Ensure your bot is added to the chat room you want to send a message to.

The chat ID can be a user ID, group ID, or channel username (prepend with @ for channels).

This function runs asynchronously, leveraging axios for HTTP requests.

The code handles basic success and error cases, returning the API response or error message.

5. Debugging & Error Fixes

If the assistant generates code that doesn't work as expected — no problem. Just send the error message or the broken code back to the assistant, and it will:

Analyze what went wrong

Suggest and generate a corrected version

Explain what was fixed (if needed)

You don't need to debug manually — just say something like:

“This code throws an error: Cannot read property 'xyz' of undefined — fix it”

Or:

“This returns undefined, but I expect the message to be shown — what's wrong?”

The assistant is designed to iteratively improve the output, so feel free to go back and forth. It’s part of the workflow.

Tips & Notes

The assistant understands your context and node configuration.

You can ask complex multi-step queries (e.g. "get data from webhook, filter it, and send to Telegram").

If something goes wrong, try rephrasing your request — the assistant will adapt.

Responses include explanations — great for learning and debugging.