An SMS-based reminder chatbot using Nexmo, Google Calendars & Gmail, and IBM Watson.
August 25, 2017
When our lives become filled with activity, organization is key. Sometimes there is too much going on to remember, so we must reply on a system to remind ourselves of deadlines, tasks, or events. There are countless applications available for download that set reminders, but these apps demand storage space, take time to set up, and can be inconvenient and frustrating to use. If you want to create a calendar event, you're forced to download an app and use a cluttered interface. What if we could send ourselves reminders or create future events simply by sending a text?
With IBM Watson, Google Calendars, and Gmail, we can build a Nexmo-powered chatbot. The bot works by interpreting our requests to set reminders with Watson and extracting the date, time, and title of the reminder. The chatbot stores your reminder by creating an event in your personal Google Calendar and notifies you via SMS and email when the event occurs.
The program will be stored on AWS Lambda and linked to API Gateway.
cd /path/to/your_repository
git clone https://github.com/kasey-mcfadden/RemindBot
Cloning the repository will create a folder on your local machine containing all the necessary files for this project.
pip install -t /path/to/your_repository -r requirements.txt
This step is needed to store all the required python packages in the repository you have cloned. This way, AWS Lambda will be able to import the packages when we upload the compressed files to our Lambda function.
AWS Lamba is where our program will be stored, and is essentially the center of our chatbot. This is where the HTTP requests sent by Nexmo, Watson, and Google are interpreted. It also controls how we will manage the RDS database, send event information to the user's calendar, and deliver SMS messages to the user.
Create a new Lambda function > Author from scratch
Basic information
Name: Remind-Bot
Runtime: Python 2.7
Lambda function code
Code entry type > Upload a .ZIP file
Upload > Compress and upload this cloned repository: remind-bot.zip
Lambda function handler and role
Handler: interact.lambda_handler
Role > Choose an existing role: IPA-Lambda
Advanced settings
Memory: 128 MB
Timeout: 6 sec
For HTTP requests to be delivered to our Lambda function, we need to establish webhook endpoints. With API Gateway, we can define what happens when these endpoints are called by linking them to our Lambda function.
Before we import our API from the swagger file in GitHub, we must customize it to match our AWS host.
Edit swagger file: open Remind-Bot-API-GW-swagger.json > specify your AWS host ("host", line 7)
Now, we can create the API
Import from Swagger > Select Swagger File: Remind-Bot-API-GW-swagger.json
Import
Set up methods for /auth, /auth/code, /interact, and /notifications:
Integration type: Lambda Function
- Use Lambda Proxy integration
Lambda Region: same region as Lambda Function
Lambda Function: Remind-Bot
Save
Select Actions > Deploy API > Deploy
In the Stage Editor interface, take note of the Invoke URL. This is the base URL that Nexmo, Watson, and Google will use to call our endpoints.
Use the commands in the remindbot.ddl file to create a new database with a table titled 'phone_numbers'.
Your 'phone_numbers' database will have two columns; the first will store users' phone numbers and the second will store their google credentials in a base64-encoded format.
The database will look something like this once users have registered for Remind-Bot:
| id | credentials |
|---|---|
| 12345678999 | google-creds |
| 19876543210 | sample-creds |
Once you've registered for an Nexmo account and have received a virtual number, edit your number.
It's time to specify where we want SMS messages received by Nexmo to be sent. Now that Invoke URL we received in Step 4 will come in handy.
Specify the Webhook URL for your virtual number: https:// INVOKE-URL /interact
Ignore the voice options - these do not matter for our project because it is strictly SMS-based.
Now, whenever your virtual number receives an SMS message, it will invoke the Lambda function with the message data by calling this Webhook URL!
Create a new Watson Conversation Workspace by importing a workspace.
Choose watson_workspace.json and select Import Everything (Intents, Entities, and Dialog).
View the details of your imported workspace and take note of the Workspace ID.
Grab your Watson Conversation Service Credentials and keep track of these values.
Start a Google API Console project
Enable the Calendar API and Gmail API for your project
Create an Oauth client ID for Application type > Web application
Name the application Remind-Bot. Under Authorized redirect URIs, enter https:// INVOKE-URL /auth/code and https://localhost:8080.
Download the JSON file of your Web Client ID as client_secrets.json
Add a Property to your API Gateway Invoke URL in the Google Search Console
Download the user-specific HTML verification file [YOUR-GOOGLE-HTML-VERIFICATION.html] and save it as google_validation_file.html
In the Lambda management console, set google_file_name as an environment variable with a value of YOUR-GOOGLE-HTML-VERIFICATION.html
Navigate to your API Gateway Project
Actions > Create resource
Configure as proxy resource: No
Resource Name: YOUR-GOOGLE-HTML-VERIFICATION.html
Resource Path: /YOUR-GOOGLE-HTML-VERIFICATION.html
Create Resource
Create a GET method for the HTML verification resource that integrates with the Remind-Bot Lambda function.
Return to the Google Search Console and verify the Invoke URL.
Verify your API Gateway Invoke URL for your Google API Console project with Google Domain Verification
In the Google Cloud Platform, create a topic and subscription. For the subscription delivery type, select Push into an endpoint URL and enter https:// INVOKE-URL /notifications as the endpoint URL that will receive POST requests.
Under topic permissions, add gmail-api-push@system.gserviceaccount.com as a Pub/Sub Publisher.
Create a new google account. This will be your master account for this project.
Create your default credentials for the master account:
cd /path/to/your_repository
python creds_test.py
You'll be redirected to a window in your browser and prompted to log in to google. Sign in with your master account. Once the authentication flow completes, you should have a file in default_credentials.json file in /path/to/your_repository.
Compress your repository. Re-upload the .zip file to Lambda.
In your Lambda function console, specify the following environment variables:
- watson_user = username from Watson Conversation service credentials
- watson_pass = password from Watson Conversation service credentials
- watson_workspace = IBM Watson Workspace ID
- redirect_uri = https:// INVOKE-URL /auth/code
- db_host = RDS Database host
- db_name = RDS Database name
- db_user = RDS Database authorized user
- db_pass = RDS Database password
- nexmo_number = your virtual number provided by Nexmo
- nexmo_api_key = your key from the API credentials located on Nexmo's dashboard
- nexmo_api_secret = your secret from the Nexmo API credentials
- google_file_name = YOUR-GOOGLE-HTML-VERIFICATION.html
- master_email = master email address
Save
Test out the Remind-Bot! Send a text to your virtual Nexmo number to get started.
A project by Kasey McFadden, overseen by Murali Ramsunder
MIT