Pro-Pilot

Preferably use PyCharm/DataSpell (JetBrains) to open the project. These IDEs provide a better experience for working with Python projects. However, they are heavy and may not be suitable for all systems. Alternatively, you can use Visual Studio Code, which is lightweight and provides a good Python development experience. Or you can use any other IDE of your choice.

All non-repository contents in Google Drive: Link

Installation

Make sure you have Python already installed in your machine.

Virtual Environment Setup

Please setup a virtual environment to run the project:

• Option 1: Create using Python: python3.x -m venv {your_virtual_env_name} and activate it: source {your_virtual_env_name}/bin/activate (Mac) / .\venv\Scripts\activate (Windows)
• Option 2: Use Anaconda: conda create --name {your_virtual_env_name} python=3.x and activate it: conda activate {your_virtual_env_name}

You can also execute Option 2 using the Anaconda Navigator GUI. 3.x represents the Python version you want to use. I would recommend using Python 3.11 or higher.

Module Installation

Once you have activated your virtual environment, navigate to the project directory and execute the following commands to install the required modules. If you are in the correct environment your shell should look something like this:

({your_virtual_env_name}) sh-3.2$ pip install -r requirements.txt

Once your access is approved, Create a project.env file in your project directory and add the following environment variables. Replace YOURSECRETKEYGOESHERE with your actual secret keys:

PYTHONUNBUFFERED=YOURSECRETKEYGOESHERE
AZURE_OPENAI_ENDPOINT=YOURSECRETKEYGOESHERE
AZURE_OPENAI_KEY=YOURSECRETKEYGOESHERE
PLATFORM_OPENAI_KEY=YOURSECRETKEYGOESHERE
AZURE_COSMOS_MONGO_ADMIN=YOURSECRETKEYGOESHERE
AZURE_COSMOS_MONGO_PASS=YOURSECRETKEYGOESHERE
AZURE_COSMOS_MONGO_CONNSTRING=YOURSECRETKEYGOESHERE

With all services running and the environment configured, you can start the Flask application by running:

python3.x app.py

Getting started

Accessing Chatbot

• Execute the command python app.py. This will start the local server and the application will be accessible at http://127.0.0.1:8080/.
• To interact with the chatbot, open a web browser and enter the URL http://127.0.0.1:8080/launch/?pwd=491062. (Chrome Recommanded)
  • You will be required to fill out the baseline survey form before proceeding to the chatbot.
  • Responses do not matter during testing
• After this you will be launched into the chatbot interface where you can interact with a simulated client
  • The client queue, their behavior, and how pro-pilot supports you is predetermined based on config.py.
  • You can skip clients by clicking the red dog button on the top left
  • You can end a client conversation forcefully by clicking the yellow cat button on the top left
• The chatbot session can be configured using the following URL format:

  http://127.0.0.1:8080/index/{session_id}?name={client_name}&domain={domain}&category={complaint_type}&grateful={client_grateful}&ranting={client_ranting}&expression={client_expressive}&civil={client_civility}&info={support_info}&emo={support_emo}
  

  http://localhost:8080/session_address/?product=YOURPRODUCT&grateful=YOURGRATEFUL&ranting=YOURRANTING&expression=YOUREXPRESSION

• Replace domain with the specific product name.
• The parameters client_grateful, client_ranting, client_expressive accept values 0 (false) or 1 (true) to determine simulated client behavior.
• The 'client_civility' parameter accepts values 0 (false) or 1 (true) to determine the client's civility.
• The 'support_info' parameter accepts values 0 (false) or 1 (true) to determine the support for informational responses (left panel)
• The 'support_emo' parameter accepts values 0 (false) or 1 (true) to determine the support for emotional responses (right panel)

Managing Data with MongoDB

• Run mongosh in your terminal to start MongoDB Shell.
• Connect to the server and select the database associated with the Flask app by running use flask_db.
• To view the data within a collection, use the command:
  • chat_history_collection — stores chat interaction history.
  • chat_client_info — stores user information.
  • chat_emo_feedback — stores emotional feedback from users.

  db.COLLECTION_NAME.find().pretty()
  

• If you need to reset a collection, type:
  • This command removes all documents from the specified collection and resets it.

  db.COLLECTION_NAME.drop()
  

Project Setup

App.py

Features

• Landing Page: Serves as the initial point of interaction where users are greeted.
• Session Management: Generates unique sessions for users to maintain distinct chat histories and interactions.
• Emotional Support: Offers responses based on the emotional content of user interactions.
• Informational Support: Provides factual support based on user queries and chat context.
• Troubleshooting Guide: Delivers step-by-step assistance for resolving user issues.

Technical Description

• / hello() - Entry route that renders the landing page.
• /chat start_chat() - Creates a new chat session and redirects the user to the chat interface.
• /<session_id>/ index(session_id) - Main chat interface that utilizes the session ID for personalized interactions.
• /<session_id>/get-reply getReply(session_id) - Handles fetching and sending replies within the chat session.
• /<session_id>/get-emo-feedback getEmoFeedback(session_id) - Updates the database with user feedback on emotional interactions.
• /<session_id>/get-emo-support getEmoSupport(session_id) - Provides emotional support responses.
• /<session_id>/get-info-support getInfoSupport(session_id) - Offers informational support based on the ongoing chat.
• /<session_id>/get-trouble-support getTroubleSupport(session_id) - Guides users through troubleshooting processes.

Agents.py

Features

• Emotional Support: Captures and processes the emotional context of customer interactions, providing representatives with insight into the customer's emotional state.
• Informational Support: Dynamically retrieves relevant information based on the ongoing conversation to aid representatives in providing factual and detailed responses.
• Procedural Support: Offers step-by-step troubleshooting assistance, guiding representatives through resolving customer issues effectively.
• Historical Context Management: Maintains and utilizes historical conversation context to ensure responses are relevant and personalized.

Technical Description

• agent_coworker_emo() - Helps service representatives by providing emotional insights into customer complaints.
• agent_coworker_emo_perspective() - Provides the customer's perspective in a summarized form to help the representative understand the customer's view and feelings.
• agent_coworker_info() - Assists service representatives by providing actionable hints based on the customer's product-related complaints.
• agent_coworker_trouble() - Provides procedural support by listing steps to resolve customer complaints.
• get_historical_info_context_chain() - Creates a context chain that reformulates statements to be understood without prior chat history.
• invoke() - Acts as the primary method to trigger the execution of the defined agent chains based on user input.