Chatbot-core

Core dialogue engine for chatbots

Getting the chatbot running on docker locally (docker-compose)

• Install Docker
• Clone this repo and cd to the cloned repo
• Execute the following command docker-compose -f stack.yml up This command would build the bot and the router images and run them along with redis as a container.

Run the below curl to check if the bot is up and running. A successful setup would return the following response 'Hi there! Please press 0 for menu.'.

    curl -X POST \
        http://localhost:4000/bot \
        -H 'content-type: application/json' \
        -d '{
            "Body": "Hi",
            "From": "123"
        }'

Getting the chatbot running on docker swarm in production (assumes SSL)

• Install Docker and initialise a docker swarm by running docker swarm init
• Clone this repo and cd to the cloned repo
• cd clonedrepo/bot
• docker build --tag rasachatbot:0.0.1 . This would build a docker image for the rasa bot
• cd clonedrepo/router
• docker build --tag rasachatrouter:0.0.1 . This would build a docker image for the router that will integrate with Rasa bot
• Push these images to a public repo so that you can reference the images from the prod.yml. You can test locally without this step, but it is mandatory to push to a public repo if you have a multi-node swarm
• Make the below changes in prod.yml
  • Update the bot and router to images that you built above
  • Update the path to the site.key, crt and cabundle.
• Execute docker stack deploy --compose-file prod.yml bot and this will deploy all the services to the docker swarm. Run docker swarm leave --force to purge the swarm and services

Run the below curl to check if the bot is up and running. A successful setup would return the following response 'Hi there! Please press 0 for menu.'.

    curl -X POST \
        http://IP:4000/bot \
        -H 'content-type: application/json' \
        -d '{
            "Body": "Hi",
            "From": "123"
        }'

Getting the chatbot running on a VM

Router Module (A node js application to interface between the chat client and bot server)

• Installation:

  • install node and npm
  • npm install to install dependencies

• configuration:

  • "config/config.js" can be modified if needed to change ports and other parameters

• Starting the services:

  • node appRest to start the REST endpoint
  • node appSocket to start Socket endpoint

Bot module

• Installation:

  • install latest anaconda environment
  • run conda env create -f environment.yml to install python dependencies. This step creates a virtual environment named rasa with python 3.6 and all dependencies
  • activate the environment conda activate rasa.
  • conda deactivate to de-activate the environment

• configuration:

  • modify config.yml if needed to change NLU components for training or change policies
  • modify endpoints.yml if needed to change action server endpoint

• Starting the services:

  • start RASA core server by running make bot or rasa run -p 5005 --enable-api --cors "*" -vv
  • start RASA action endpoint server by running make action or rasa run actions -vv -p 5056

Integration Instructions

• REST integration Method: Post Endpoint: http://<IP>:<PORT>/bot Body: { "Body": "Hi", "From": "AC6436e7066283ef84c88a05392cc0fcd6" } Header: {'content-type': 'application/json'}

  *Body - User message *From - session id (uuid or some Session/ User identifier)

  • Response: Text based response

  • Curl request for the same:

     curl -X POST \
     http://<IP>:<PORT>/bot \
     -H 'content-type: application/json' \
     -d '{
         "Body": "Hi",
         "From": "AC6436e7066283ef84c88a05392cc0fcd6"
     }'
    

• Socket Integration

  • Socket message body sample(using socket.io):
  • Request: Url: http://52.173.240.27:4005 Body: ["user_uttered",{"message":"what are the timelines for portal?","customData":{"userId":"123"}}]
  • Response: ["bot_uttered",{"text":"please choose from CBSE or State Board, other bords are not handled as of yet. 1. CBSE \n 2. State Board","intent":"template_ans_demo","type":"response"}]