Rainbow bot using Microsoft Luis sample

Introduction

This is a sample of how a NodeJS Rainbow bot can easily use the Microsoft Luis capabilities. The user can ask for current weather or forecast in a location with natural language. For example:
"How is the weather in San Francisco ?",
"How will the weather be in the coming days in London?"

Preparing the LUIS application

You need to have an account on https://www.luis.ai

1. App and Intents:
   From there create a new App "luis-rainbow"
   Select the English culture in the Create new app popup
   

• create one intent: queryWeather with the following utterances:
  What is the weather like in paris ?
Weather in Boston next week ?
How's the weather in Sydney today ?
Will it rain in Tokyo in april ?
Is it sunny in Washington this afternoon ?
Forecast of next week in Seattle
Give me the weather forecast for London

• Don't forget to feed the intent None with some samples of unrelated questions (e.g.)
  How far is the moon from earth
Penguins cannot fly
What did people eat during world war two ?
Asia is very large

2. Entities:
   Add the following pre-built entities to your application:

• datetimeV2
• geographyV2

3. Train the application and publish it.
   From the "Keys and Endpoints" part of the "Manage" section, you can retrieve the "Endpoint" url that you will use in the json configuration file.

Installation

To install, just clone this repository, and type:

npm install

Usage

Requirements

To be able to play with this sample bot you need:

• Your Rainbow developer account (you can create one from here: https://hub.openrainbow.com),
• A Microsoft Luis endpoint URL (https://www.luis.ai),
• An OpenWeatherMap API key (https://openweathermap.org).

Configuration

You can now configure the credentials and options for the Rainbow SDK, for Microsoft Luis and OpenWeatherMap. This sample uses the config module (https://www.npmjs.com/package/config).

A config/default.json file is provided. We recommend you make a copy of this file and name it config/local.json. The local.json file is used to locally store your credentials without pushing them in a git (this file is part of .gitignore). There are three parts in this configuration file: one for Rainbow, one for Luis and one for OpenWeatherMap. Replace each value between < and > wih your associated credentials.

If you plan to use the bot in Docker, please refer to the Docker section below for configuration in this case.

Running

To start, just type:

npm start

Logs

Using the default configuration in config/default.json, Rainbow SDK logs will be stored in the logs subdirectory (file with name <date>-rainbowsdk.log):

        "logs": {
            "enableConsoleLogs": false,
            "enableFileLogs": true,
            "file": {
                "path": "logs",
                "level": "debug"
            }
        },    

Bot logs will also be stored in the logs directory, in a file named rainbow-luis-sample.log.

Docker

This sample can also be used in docker.

Configuration

You can configure the credentials (Rainbow SDK, Microsoft Luis and OpenWeatherMap) by editing the file bot_env_file. The environment variables embedded in bot_env_file will be used by config/custom-environment-variables.json file.

Building the docker image

To build the docker image type:

docker build -t rainbow-luis-sample .

Starting the bot service

The bot service can be started by typing:

docker-compose up

Bot architecture

The sample bot is divided in different parts:

• app.js : main file of the application, interfaces with Rainbow
• module/bot.js: handle user messages with the help of luis and openweathermap modules, sends back response (event emitter)
• module/luis.js: this does the job to send a user sentence to Luis, and get back the result (intents, entities)
• module/openweathermap.js: interfaces with OpenWeatherMap service.

The bot subscribes to Rainbow SDK rainbow_onmessagereceived events (https://hub.openrainbow.com/#/documentation/doc/sdk/node/api/events#Events+event_rainbow_onmessagereceived).

When the receive message type is chat message we give it to the bot module (handleMessage) and we mark the message as read (https://hub.openrainbow.com/#/documentation/doc/sdk/node/api/im#IM+markMessageAsRead).
The bot.js handleMessage function sends the message to Luis and gets back intents and entities.
Given the intents / entities combination, it calls the openweathermap module to get weather information.
Response messages are sent back to app.js using the sendIM event. app.js listens to sendIM bot messages.
The sendIM handler calls the Rainbow SDK to send the message to the user (https://hub.openrainbow.com/#/documentation/doc/sdk/node/api/im#IM+sendMessageToJid).