README.md

(Archived) Triage AI (Archived)

Disclaimer : This is for educational purposes only. DO NOT USE IN PRODUCTION

This is a Romanian language triage chatbot

Credits

• https://github.com/RasaHQ/rasa for Rasa obviously
• https://github.com/JiteshGaikwad/Chatbot-Widget for the chatbot widget

How to use

Run the start.sh script. You need to have an OpenAI token in order to run the explanation tasks

./start.sh

After everything is set up, open a browser and navigate to: http://localhost:8080

How to test

Run the test.sh script

./test.sh

Implementation details

Docker

It uses 3 containers:

• rasa_actions for Rasa custom actions
  • the Dockerfile installs openai with pip in order to use the API
  • uses .env file that can be generated from the start.sh script
• rasa for the Rasa model
  • the Dockerfile installs ro_core_news_lg language package for spacy
• webapp for the Nginx server that serves a static page with the widget

Start.sh

Can be used with the cleanup parameter to remove old models. It will train Rasa if no models are found, then it will run docker-compose for the docker-compose.yml file.

Test.sh

Executes docker-compose for the docker-compose.test.yml file that contains 2 containers:

• rasa_actions for Rasa custom actions
• rasa with the test command

All containers used in the docker-compose.test.yml are used in the production environment as well

Check stories

In order to check if all the stories result in a color code, run stories_graph.py and check for red nodes. All the nodes with a path to the leaves will be in green and all the leaves in blue

python3 stories_graph.py

OpenAI API

The start.sh script will generate the .env file used by the rasa_actions container

Enjoy!!!