Code samples related to Building Better Bots published on the AWS ML Blog

CoffeeBot is a transactional chat bot that can help one order a mocha (relies on AWS Mobile Hub and Android).

Consider this conversation:

User: May I have a mocha?
CoffeeBot: What size? small, medium, large?
User: small
CoffeeBot: Would you like that iced or hot?
User: hot
CoffeeBot: You'd like me to order a small mocha. Is that right?
User: Make it a large mocha
CoffeeBot: You'd like me to order a large mocha. Is that right?
User: yeah
CoffeeBot: Great! Your mocha will be available for pickup soon. Thanks for using CoffeeBot!

Let's build this voice bot, an Android App that talks to you using Amazon Polly and Amazon Lex. You can use the AWS Console for your account to start testing the bot, but you can also build a mobile app using:

• Android development environment (download)
• To test voice (you can use the simulator for text)
  • An Android device
  • A USB cable for USB debugging (more info for Amazon Fire tablets)

First, we'll create the Amazon Lex bot. Then, we'll add some Lambda Functions to bring it to life. Finally, we'll put it all together with Mobile Hub and the Lex Android SDK.

1. From the Amazon Lex console, create a Custom bot with these settings (you can see these in the "Settings" tab later)
   • Bot name: CoffeeBot
     • To work independently in a shared environment, use your initials in the name (e.g., CoffeeBotXXX)
   • Output voice: Salli
   • Session timeout: 5 min
   • IAM role: (accept the default) AWSServiceRoleForLexBots
   • COPPA: (our bot is not directed at children) No
2. Review the Error handling settings
   • Prompts: (one prompt) Sorry, but I didn't understand that. Would you try again, please?
   • Maximum number of retries: 2
   • Hang-up phrase: (one phrase) Sorry, I could not understand. Goodbye.

From the left, add a new Intent called cafeOrderBeverageIntent with the following settings and click "Save Intent" to save the Intent.
To work independently in a shared environment, use your initials in the Intent name (e.g., cafeOrderBeverageIntentXXX).

1. Lambda initialization and validation (leave unchecked)
2. Fulfillment: choose "Return parameters to client" for now
3. Confirmation prompt: You'd like me to order a {BeverageSize} {BeverageType}. Is that right? to confirm and Okay. Nothing to order this time. See you next time! to cancel.
4. Sample Utterances: add these to the list of sample utterances so the bot recognizes similar phrases (each entry on a separate line)

I would like a {BeverageSize} {BeverageType}
Can I get a {BeverageType}
May I have a {BeverageSize} {Creamer} {BeverageType}
Can I get a {BeverageSize} {BeverageTemp} {Creamer} {BeverageType}
Let me get a {BeverageSize} {Creamer} {BeverageType}

Add the following Slot types (each value should be a separate entry); remember to "Save slot type" as you go along. To work independently in a shared environment, use your initials in the names (e.g., cafeBeverageTypeXXX).
Note: Although they are saved with the AWS Account, Slot Types will only show up in the list when they are associated in the next step.

Add the following entries to the list of Slots, choosing the Slot Types created above. Click "Save Intent".

Build the app and test some of the Utterances in the Test Bot dialog at the bottom right of the Amazon Lex Console. For example, if you say May I have a chai?, does Lex correctly map chai to the BeverageType slot?

1. Create the cafeOrderCoffee function by saving cafeOrderCoffee_lambda.js as a Node.js 8.10 function
   • To work independently in a shared environment, use your initials in the function name (e.g., cafeOrderCoffeeXXX)
   • You can get the function source here
   • (No need to set up a trigger; you can accept default values for most of the configuration)
   • Choose an IAM role that includes the AWSLambdaBasicExecutionRole Managed Policy. If no such role exists, you can create a new IAM Role using one of these approaches:
     • Choose "Create new role from template(s)", provide a role name, and choose Simple Microservice permissions from the "Policy templates" dropdown
     • Choose "Create a Custom role", which should open up a new tab where an IAM role is shown; review the policy document and click "Allow"
2. Configure the Test event and test to confirm the function works as expected (see cafeOrderCoffee_test.json)
   • you can get the event source here
3. You'll notice that the function checks the bot name it receives (if (event.bot.name !== 'CoffeeBot')); remember to change this value in the function and in the test event to match the name you used for your bot

1. From the Lex Console, select the CoffeeBot bot and choose Latest from the version drop down to make changes
2. Modify the cafeOrderBeverageIntent Intent
   • Associate it with the new cafeOrderCoffee Lambda function (select "Lambda function" in the "Lambda initialization and validation" area)
     • When prompted, allow Amazon Lex to call your new function
   • Associate it with the new cafeOrderCoffee Lambda function for (select "Lambda function" in the "Fulfillment" area); remember to click "Save Intent"
3. Build the bot
4. Test using the Amazon Lex Console; do you see any responses when you ask May I have a mocha?

1. From the Mobile Hub console, create a new project called CoffeeBot.
2. Add the "Conversational Bots" feature to the project. When prompted, import CoffeeBot. Mobile Hub takes care of a number of important details behind the scenes. A new Amazon Cognito Federated Identity Pool is created for this new app along with roles so that the users can interact with Lex (using voice and text).
3. Source code for the new app is immediately available for download.
4. Follow the instructions in the READ_ME/index.html file to setup, compile, and run the app.

When you're ready, try out AWS Amplify for bringing your chatbot to a mobile or web environment.