Alex: Prototype Chatbot for Tet.ai

Introduction

Alex is a prototype patient chatbot, allowing medical students - and other would-be doctors - to practice patient-doctor interactions and their diagnostic skills.

There are two main parts: the dialogue model and the NLU model. The NLU model can take natural language and convert it into structured data that a computer can understand. The dialogue model takes this structured data and chooses a suitable reply to it.

It can be run either locally, using the terminal, or running on Facebook Messenger using a websocket hosted on Heroku.

Setup

The All-in-One

1. Run this command in terminal

   pip install rasa_core
   pip install rasa_nlu[spacy]
   cd ~/
   git clone https://github.com/tet-ai/alex
   cd ~/alex/
   python -m rasa_core.train -d domain.yml -s stories.md -o models/dialogue
   python -m rasa_nlu.train -c nlu_config.yml --data nlu.md -o models --fixed_model_name nlu --project current --verbose
   python -m rasa_core.run -d models/dialogue -u models/current/nlu

   and you should be able to talk to the chatbot.

In Baby Steps

1. Install Rasa

   pip install rasa_core
   pip install rasa_nlu[spacy]

   For more info or alternative installation methods, go here.

2. Clone the repository

   cd ~/
   git clone https://github.com/tet-ai/alex

3. Move to the repository

   cd ~/alex/

4. Train the dialogue model

   Run

   python -m rasa_core.train -d domain.yml -s stories.md -o models/dialogue

   This will train the dialogue model and store it into models/dialogue.

5. (Optional) Talk to bot with structured data

   We have only trained the dialogue model, not the NLU. But we can send structured data to the bot now.

   Run

   python -m rasa_core.run -d models/dialogue

   Then send it a message such as /greet.

6. Train NLU

   Run

   python -m rasa_nlu.train -c nlu_config.yml --data nlu.md -o models --fixed_model_name nlu --project current --verbose

   This creates the NLU model in models/current/nlu

7. Talk to Bot

   Now the bot is all trained up. To speak to it, run

   python -m rasa_core.run -d models/dialogue -u models/current/nlu

Retraining the chatbot

If you make changes to the chatbot, you will need to retrain it.

python3 -m rasa_core.train -d domain.yml -s stories.md -o models/dialogue
python3 -m rasa_nlu.train -c nlu_config.yml --data nlu.md -o models --fixed_model_name nlu --project current --verbose
python3 -m rasa_core.run -d models/dialogue -u models/current/nlu

Running for facebook

If setting up with facebook, follow these instructions.

python -m rasa_core.run -d models/dialogue -u models/current/nlu \
   --port 5002 --connector facebook --credentials credentials.yml

python -m rasa_core.run -d models/dialogue -u models/current/nlu \
   --port 5002 --credentials credentials.yml

To run locally for facebook

1. Expose local link

   ngrok http 5002

   see ngrok.

   This makes random_name.ngrok.io point to localhost:5002

2. • Run virtual environment

     source .env/bin/activate

   • Run Rasa

     python -m rasa_core.run -d models/dialogue -u models/current/nlu \
     --port 5002 --credentials credentials.yml

3. Create webhook

   node index.js

4. Point facebook in the right direction

   1. Go to application page on 'Facebook for developers'. Mine is here.
   2. Webhooks
   3. Edit subscription
   4. Callback URL: 'random_name.ngrok.io/webhooks/facebook/webhook'

5. Talk on messenger!