Chatbots
Architecture Decisions
ADR-0005 - Use Dialogflow for chatbots
Dialogflow
Implement the intent
Guide: Google Documenation
Model your Dialog
Here is the plantuml activity diagram
Create a "Default Welcome Intent" with answers: "Hi I am your personal trainer, let me advice you on your health. I can currently calculate your BMI, store the BMI and advices you on your BMI."
Implement a "mange_bmi" intent
After creating the intent add training phrases and activate webhook fulfilment (this is required to check if allready a user exists for this profile):
Do the same for the update_bmi intent (training phrase: My hight is 177 cm, I weigh 80 kg and my birthday is 19.12.1987). How to handle partial dates (https://miningbusinessdata.com/dialogflow-tutorial-setting-context-from-your-inline-webhook-using-contextout/)
Implement the fulfilment
Using Firebase
To start it makes sense to implement all fulfillments in one webhook on firebases. Later for more complex logic it is adviced to move this logic to it's own webhooks that are then called from the fulfillment webhook. For more details see the later section.
Setup Firebase (https://den.gitbook.io/developerplaybook/services-api/firebase)
mkdir fulfillment-router
firebase login
firebase initGuide: Google Documentation
"we highly recommend that you use the Node.js client library to process requests and return responses"
Using a Webhook with AWS Lambda
First build a webhook with a fullfilment router
https://medium.com/@pallavtrivedi03/how-to-make-a-webhook-for-dialogflow-fulfillment-d02835cc50bf
Test a chatbot - manually & unit test
Connect the Chatbot to Facebook Messanger
Integrate fulfilment with REST Python AWS Lambda (e.g. "analyse my bmi trend")
Sample: https://github.com/miningbusinessdata/apiai-python-webhook/blob/master/app.py
Last updated