botkit-middleware-apiai

Middleware for using Api.ai with Botkit-powered bots

Usage no npm install needed!

<script type="module">
  import botkitMiddlewareApiai from 'https://cdn.skypack.dev/botkit-middleware-apiai';
</script>

README

Botkit Middleware Apiai

This middleware plugin for Botkit allows you to utilize Api.ai, a natural language classifier service directly into the Botkit corebot.

The Api.ai platform lets developers seamlessly integrate intelligent voice and text based command systems into their products to create consumer-friendly voice/text-enabled user interfaces.

Setup

In order to utilize api.ai's service you will need to create an account and an agent. An agent will represent your Bot's comprehension skills. Head over to their sign up page to get started. After creating an account you will be able to create your first agent and start creating intents. Grab the developer access token for your local dev and a client access token for production as seen below

Api.ai Tokens

Next you will need to add botkit-middleware-apiai as a dependency to your Botkit bot:

npm install --save botkit-middleware-apiai

Enable the middleware:

var apiai = require('botkit-middleware-apiai')({
    token: <my_apiai_token>,
    skip_bot: true // or false. If true, the middleware don't send the bot reply/says to api.ai
});

controller.middleware.receive.use(apiai.receive);

// apiai.hears for intents. in this example is 'hello' the intent
controller.hears(['hello'],'direct_message',apiai.hears,function(bot, message) {
    // ...
});

// apiai.action for actions. in this example is 'user.setName' the action
controller.hears(['user.setName'],'direct_message',apiai.action,function(bot, message) {
    // ...
});

What it does

Using the Api.ai middleware with Botkit causes every message sent to your bot to be first sent through Api.ai's NLP services for processing. The response from Api.ai is then returned in the incoming messages as message.intent, message.entities for any language entities (dates, places, etc), message.fulfillment for Api.ai specific speech fulfillment, message.confidence for the confidence interval, and finally the message.nlpResponse which represents the raw request as seen below:

    {
      "id": "XXXX",
      "timestamp": "2016-05-31T18:20:38.992Z",
      "result": {
        "source": "agent",
        "resolvedQuery": "hello",
        "action": "",
        "actionIncomplete": false,
        "parameters": {},
        "contexts": [],
        "metadata": {
          "intentId": "XXX",
          "webhookUsed": "false",
          "intentName": "hello"
        },
        "fulfillment": {
          "speech": ""
        },
        "score": 1
      },
      "status": {
        "code": 200,
        "errorType": "success"
      }
    }