RASA Component for Virtual Receptionist

This codebase are used for processing the input from Webex component and produce the output for Heygen to vocalize. The action.py simulates the backend integration of the virtual receptionist capabilities.

Prerequisites

Ensure you have the following installed:

• Python 3.8+: Rasa requires a modern Python version.

• Rasa Open Source Version: You can install it via pip:

  pip install rasa-sdk

• Virtual Environment (Recommended): This is to manage project dependencies.

  python -m venv .venv
  source .venv/bin/activate  # On macOS/Linux
  .venv\Scripts\activate     # On Windows

Main Project Files

• domain.yml: Defines the intents, entities, slots, forms, and responses of the assistant.
• nlu.yml: Contains Natural Language Understanding (NLU) training data, mapping user inputs to intents and extracting entities.
• rules.yml: Defines short, fixed sequences of turns that the assistant should follow.
• stories.yml: Defines longer, example conversations that the assistant should be able to handle.
• actions.py: Contains custom actions that the assistant can execute, involving integrations and more complex logic. In this code it is used to simulate integration with Google calendar, employee's location checking and webex messaging simulation.

Setup RASA

Follow these steps to get the RASA up and running:

1. Navigate to the Project Directory: Make sure you are in the root directory of your Rasa project where config.yml, credentials.yml, domain.yml, endpoints.yml, and the folders actions, data, models and tests are located.

2. Train the Rasa Model: This command will train the NLU and Core models based on your nlu.yml, rules.yml, and stories.yml files.

   rasa train

3. Run the Actions Server: Open a new terminal window and execute this command:

   rasa run actions

   Keep this terminal window open while you interact with the Rasa assistant.

4. Enable API and Run the Rasa Assistant: In your first terminal window (where you run rasa train), start the Rasa assistant with the API enabled and CORS configured to allow requests from any origin.

   rasa shell --cors "*"

actions.py Overview

The actions.py file simulates interactions with external systems like calendar applications and employee location checking integrations.

Dummy Values for Simulation

The actions.py uses several dummy dictionaries to simulate external integration responses:

• DUMMY_CALENDAR_MEETINGS:

  • Simulates response from a calendar system that stores confirmed meetings.

• DUMMY_EMPLOYEE_LOCATIONS:

  • Simulates response from Cisco Space employee location tracking system.

• DUMMY_EMPLOYEE_REPLY:

  • Simulates the employee's response to a message sent by the receptionist via Webex Messaging.

Functions

• ValidateMeetingDetailsForm:

  • This function ensures the required slots (visitor_name, employee_name, meeting_time) for the meeting_details_form are correctly extracted from the user's input.

• ActionCheckMeetingDetails:

  • Simulates checking a calendar system. It uses DUMMY_CALENDAR_MEETINGS to determine if a meeting exists for the provided details. It sets the meeting_confirmed and room_name slots.

• ActionLocateEmployee:

  • Simulates querying a location tracking system. It uses DUMMY_EMPLOYEE_LOCATIONS to check if the employee is currently in the office at the specified meeting time.

• ActionSendWebexMessage:

  • Simulates sending a message to the employee (e.g., via Webex) to inform them about their visitor. It then retrieves a dummy reply from DUMMY_EMPLOYEE_REPLY and sets the employee_reply slot. This action handles both scenarios where the employee is in or out of the office.

• ActionProcessEmployeeResponse:

  • Interprets the employee_reply slot and provides an appropriate response to the visitor, informing them whether the employee is coming, needs them to wait, or needs to reschedule.

• ActionFinalInstructions:

  • Provides final instructions to the visitor based on the outcome of the meeting check and employee response.

• ActionHandOff:

  • Simulates transferring the conversation to a live human agent.