Home

Awesome

Serverless Messenger Bot Boilerplate by SC5 logo

Facebook has sunset stories in Wit.ai. This boilerplate will no longer function due to this

More at https://wit.ai/blog/2017/07/27/sunsetting-stories

The Objective

Create a Facebook Messenger chatbot using Serverless Framework to build backend for the service.

Objective chart

  1. User sends message from Facebook Messenger to Facebook Messenger API which forwards it to the AWS API Gateway
  2. AWS API Gateway triggers Lambda function facebookBot which sends the message forward to the SNS topic witAiTopic
  3. Lambda function witAiHandler, which subscribes to the witAiTopic SNS topic, receives the message, forwards it to wit.ai, processes actions required by the bot (e.g. fetch data from 3rd party services) and posts the response to the SNS topic fbMessengerTopic that forwards message to wit.ai which processes it and returns context
  4. Lambda function facebookBot subscribes to the fbMessenger topic. It receives the message sent by witAiHandler and forwards it the the Messenger Platform
  5. User receives the response to the message

Boilerplate Installation

  1. Run serverless install --url https://github.com/SC5/serverless-messenger-boilerplate
  2. Run npm install
  3. Rename example.env to .env and fill in FACEBOOK_BOT_VERIFY_TOKEN
  4. Run serverless deploy then copy the GET endpoint url

Facebook App Configuration

  1. In Facebook Developer site create new Facebook application and page
  1. Go to the App Dashboard and under Product Settings and setup webhook by clicking Setup Webhooks
  1. Click Verify and Save
  2. In the Token Generation section, select the page created in step 5 and copy the generated token
  3. In Webhooks section, select page created in step 5 to be the one that subscribes the webhooks
  4. Paste token copied in step 8 to .env file as FACEBOOK_BOT_PAGE_ACCESS_TOKEN

More detailed instructions for Facebook Messenger platform configuration can be found from https://developers.facebook.com/docs/messenger-platform/quickstart/

Wit.ai Configuration

Following steps are slightly modified version of https://wit.ai/docs/quickstart

  1. Register to wit.ai https://wit.ai
  2. Click + icon located in top menu to create a new app
  3. Type in app name and description
  4. If you wish you can change language and data privacy
  5. Select settings
  6. Copy App ID
  7. Paste token to .env file as WIT_AI_TOKEN
  8. Implement wit.ai flow

Testing

The serverless-mocha-plugin module is included in the boilerplate. You can invoke tests locally with

  1. serverless invoke test (for all test)
  2. serverless invoke test -f facebookBot (to run only the facebookBot tests)
  3. serverless invoke test -f witAiHandler (to run only the witAiHandler tests)

Implement the test cases for your bot to test/witAiHandler.js. By default, the tests are run silently (not forwarded to Messenger). If you want to run full tests, comment out the row process.env.SILENT = 1; from the tests test/*.js.

Final Touch

  1. Run serverless deploy
  2. Open Messenger and search you bot
  3. Now you can converse with your bot

License

Copyright (c) 2016 SC5, licensed for users and contributors under MIT license. https://github.com/SC5/serverless-messenger-boilerplate/blob/master/LICENSE