Dialogflow ES

Suggest Edits

Overview

In this guide, we will take a look at how to setup a custom NLU connection between Dialogflow and Voiceflow.

📘

Resources

Detect Intent Function

The below detectIntent function is responsible for sending a user utterance to our Dialogflow agent and responding with the triggered intent and any relevant entity values.

🚧

Intents and Entities must match

For any custom NLU connection to work, it is important that your intents and entities in your NLU match what you have stored in your Voiceflow Interaction Model.

Other data that may be useful in theDetectIntent response body include intentDetectionConfidence and languageCode. For more information of what is available, see here.

const detectIntent = async (languageCode, queryText, sessionId) => {
  let sessionPath = sessionClient.projectAgentSessionPath(PROJECTID, sessionId);

  let request = {
    session: sessionPath,
    queryInput: {
      text: {
        text: queryText,
        languageCode: languageCode,
      },
    },
  };
  const responses = await sessionClient.detectIntent(request);
  const result = responses[0].queryResult;

  const confidence = result.intentDetectionConfidence;

  const obj = result.parameters.fields;
  const createIntent = (name, value) => ({ name, value });
  const intents = Object.entries(obj).map(([key, value]) =>
    createIntent(key, value.stringValue)
  );

  return {
    response: result.intent.displayName,
    entities: intents,
    confidence: confidence,
  };
};

Intent Request Type

Finally, we will pass the intent name and entities object into a type: "intent" request to return the next action in the conversation.

isRunning = await interact(userID, {
  type: "intent",
  payload: {
    intent: {
      name: intent.response,
    },
    entities: intent.entities,
    confidence: intent.confidence,
  },
}

Project file

const dialogflow = require("@google-cloud/dialogflow");
const axios = require("axios");
const { cli } = require("cli-ux");

require("dotenv").config();

const CREDENTIALS = JSON.parse(process.env.CREDENTIALS);
const PROJECTID = CREDENTIALS.project_id;

const CONFIGURATION = {
  credentials: {
    private_key: CREDENTIALS["private_key"],
    client_email: CREDENTIALS["client_email"],
  },
};

const sessionClient = new dialogflow.SessionsClient(CONFIGURATION);

const detectIntent = async (languageCode, queryText, sessionId) => {
  let sessionPath = sessionClient.projectAgentSessionPath(PROJECTID, sessionId);

  let request = {
    session: sessionPath,
    queryInput: {
      text: {
        text: queryText,
        languageCode: languageCode,
      },
    },
  };
  const responses = await sessionClient.detectIntent(request);
  const result = responses[0].queryResult;

  const confidence = result.intentDetectionConfidence;

  const obj = result.parameters.fields;
  const createIntent = (name, value) => ({ name, value });
  const intents = Object.entries(obj).map(([key, value]) =>
    createIntent(key, value.stringValue)
  );

  return {
    response: result.intent.displayName,
    entities: intents,
    confidence: confidence,
  };
};

async function interact(userID, request) {
  // call the Voiceflow API with the user's name & request, get back a response
  const response = await axios({
    method: "POST",
    url: `https://general-runtime.voiceflow.com/state/user/${userID}/interact`,
    headers: {
      Authorization: process.env.API_KEY,
    },
    data: {
      request,
    },
  });

  // loop through the response
  for (const trace of response.data) {
    switch (trace.type) {
      case "text":
      case "speak": {
        console.log(trace.payload.message);
        break;
      }
      case "end": {
        // an end trace means the the Voiceflow dialog has ended
        return false;
      }
    }
  }

  return true;
}

async function main() {
  const userID = await cli.prompt("> What is your name?");
  // send a simple launch request starting the dialog
  let isRunning = await interact(userID, { type: "launch" });

  while (isRunning) {
    const nextInput = await cli.prompt("> Say something");
    // send a simple text type request with the user input
    let intent = await detectIntent("en", nextInput, userID);
    isRunning = await interact(userID, {
      type: "intent",
      payload: {
        intent: {
          name: intent.response,
        },
        entities: intent.entities,
        confidence: intent.confidence,
      },
    });
  }
  console.log("The end! Start me again with `npm start`");
}

main();

Updated about 1 year ago