Skip to main content
The Code connector allows you to integrate an NLU engine that isn’t natively supported by Cognigy.AI or when you need a highly customized integration with a supported NLU engine that goes beyond the default connector’s capabilities. To use this connector, follow these steps:
  1. Go to Build > NLU Connectors and create a Code connector.
  2. Add your JavaScript code to the Transformer section and activate the Enable Code Transformer toggle.
createNluCodeConnector({

  /**
   * This function connects to an external NLU service, for example, Dialogflow, LUIS.
   * and returns the processed result.
   * @param text The message text from the user.
   * @param data Any additional data passed with the message.
   * @param language The language of the input message.
   * @returns nluResult Processed result from the external NLU service.
   */
  getNluResult: async ({ text, data, language }) => {

    // Example: Make an HTTP request to an external NLU API, for example, Dialogflow, LUIS.
    const response = await fetch('https://api.externalnlu.com/analyze', {
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
        'Authorization': 'Bearer YOUR_API_KEY'
      },
      body: JSON.stringify({
        query: text,
        language: language,
      }),
    });

    // Parse the response from the NLU API
    const nluResponse = await response.json();

    // Construct the NLU result based on the response from the NLU service
    let nluResult = {
      intent: nluResponse.intent || null,        // Extracted intent from the NLU response
      type: nluResponse.type || null,            // The type of the Intent (optional)
      slots: nluResponse.entities || {},         // Extracted entities as Slots (parameters)
      intentScore: nluResponse.intentConfidence || 0, // Confidence score of the Intent
      nlu: {
        intentId: nluResponse.intentId || null,  // Unique ID for the detected Intent
        intentFlow: nluResponse.intentFlow || null // Flow associated with the Intent (if available)
      }
    };

    // Return the transformed data and NLU result
    return {
      data,
      nluResult
    };
  }
});
Description
Function and its parameters
createNluCodeTransformerThis function create a transformaer that asynchronously sends the message (text) to an external NLU service for analysis. The function processes the response from the service, populates the nluResult object with values such as intent, type, slots, and intentScore, and returns the data object unchanged, unless explicitly modified.
textThe message received from the endpoint, such as user input, that needs to be analyzed by the NLU service.
dataThe data object associated with the message, containing additional metadata or context about the request. This object can be modified or passed through unchanged.
languageThe language code of the locale, representing the language in which the message is processed. For example, en for English, de for German.
NLU Result InitializationInitial values set for the NLU result object to ensure structure before any processing occurs.
intentInitially null, it will be populated with the Intent detected from the message, for example, book_flight or greet_user, after processing.
typeInitially null, it refers to the type of Intent, for example, action, informational. Will be set based on the NLU service’s response.
slotsInitially {}, this object holds the extracted parameters (Slots) that are associated with the detected intent, for example, destination, date, and class.
intentScoreInitially 0, it represents the confidence score (ranging from 0 to 1) indicating how confident the NLU service is in the detected Intent.
nluAn object containing metadata about the Intent:
- intentId - initially null, the unique identifier for the detected Intent.
- intentFlow - initially null, the Flow associated with the Intent.
Return ObjectThe function returns the data and nluResult.
dataThe unchanged data object passed in, which may contain context or additional information.
nluResultThe object returned after processing, containing the final intent, slots, intentScore, and any other details extracted from the message using the NLU logic.

​
More Information