TP_Chatbot_Rasa

---

---

Product Screenshots

Current Architecture & Features Implemented

Architeture:

For Roman-Urdu & English, two separate agents are trained, evaluated and tested. Each having its own trainin samples, but same intent_names, actions & domain

Features

• Spell Checker is implemented to correct only delibrate misspelled word through skipping vowels from original words. example ('balance': 'blnc', 'balanc', 'blnce')
• Language Detection is done through an ensemble of custom trained model and Fasttext language detection model.
• Parsers are written to convert text formats to JSON and back to text, for support of UI based drag-n-drop features.
  1- Intnet Parser 2- Stories Parser 3- Domain Parser (json to yml is pending)
• User Authenticatoin, users are verified based on PIN codes as of now
• Small Talk, a limited version of small talk is implemented (for both roman & english agents) alongwith other intents
• Logs, every conversation is being logged (separate file for each session-ID)
• Low Confidence Prompts are implemented using rasa's two stage fallback policy
• UI is currently implemented on LiveZilla (on Evamp & Sanga servers)
• Backend is deployed on redbuffer servers.

---

Future Work

• Integration of Parsers with UI
• FAQs Knowledge Base
• Personalized Content
• Business Logic Prompts
• Low Confidence Prompt needs to be improved using some custom mechanism. Rasa's two stage fallback policy sovles this problem but causes a lot others too.
• One-Click integration modules for different platforms: Slack, Messenger, WhatApp etc.
• Admin Portal with support of self-learning chatbot system

---

Git Repositories

There are 3 branches for project right now. 1- Master is holding the code version that is being used in live demo 2- Development other than live demo, latest stable code will be pushed here. 3- Dev2 all the tests can be conducted on this branch. OR incase multiple developers are working on project.

Instructions for Demo

• Only phone numbers are allowed currently:
  • 923441111111
  • 923451234567
• Enter any 4-digit PIN number
• Enter '/restart' to restart the session
• For now page reload will also reset server sessions (will clear chat history)

---

---

Installation with Docker

$ sudo apt-get update -y

$ sudo apt-get install docker.io

$ sudo curl -L https://github.com/docker/compose/releases/download/1.21.2/docker-compose-`uname -s`-`uname -m` -o /usr/local/bin/docker-compose

$ sudo chmod +x /usr/local/bin/docker-compose

$ cd base_docker; 

$ bash ./base_docker.sh

$ cd ..

$ bash ./run.sh

This installation is only for backend of Chatbot system.

---

---

Installation with Virtual Environment

This installation is only for backend of Chatbot system.

Setup Virtual Environment (Linux Environment):

• Make new python3 environent & activate:

$ virtualenv --python=python3 venv
$ source venv/bin/activate

• and install requirements from 'requirements.txt' file:

pip install -r requirements.txt

Your environment should be ready by now 😃 !

Setup -- RASA Server

1. Clone Project & open terminal in chatbot root directory 'tp_chatbot_rasa'

2. activate virtual environment

   source venv/bin/activate

3. run commands for RASA Action server:

   rasa run actions --port 5005

4. open a new terminal window

5. activate the virtual environment by 2

6. run commands for RASA API server

   rasa run -m models/ --enable-api --log-file out.log --cors "*" --endpoints endpoints.yml --port 8000 --debug

Setup -- Flask API Server

1. open new terminal / linux screen in root directory
2. activate virtual environment
3. run flask api:

   python3 api.py

Client Side

1. open 'index.html' in browser

---

Other Information about Server PORTS:

Setup and Port Numbers for Multi-Lingual Chatbot Redis Tracker Store: 6379 tracker_store: type: redis url: localhost port: 6379 db: 0 password:

Redis Interface: redis-commander (at port 8081)

Roman Urdu chat agent: actions server:
>> rasa run actions -p 5056

shell/api server to listen action server:
(define action endpoint in endpoints.yml)
>> rasa shell --endpoints endpoints.yml -p 5006 --debug

English chat agent: actions server:
>> rasa run actions -p 5055

shell/api server to listen action server:
(define action endpoint in endpoints.yml)
>> rasa shell --endpoints endpoints.yml -p 5005 --debug

>> todo: update docker file for single run, update readme for local UI, dockerized version is a bit slower, 
	clean readme, add directory structure, test cases for complete chatbot system