MITR - An Empathetic AI Companion

MITR is a web-based, multi-persona AI chatbot designed to provide empathetic and context-aware conversations. It leverages a modern microservice architecture and Google's Generative AI to create a persistent and personalized user experience.

Key Features

• Multi-Persona System: Switch between four distinct AI personalities (Lucy, Suzanne, Lexi, and Roxy), each with a unique style.
• Persistent Session History: Conversations are saved per-persona. Users can create new sessions, load past conversations, and delete old ones.
• Client-Side Voice Mode: A fully functional voice-to-text mode that uses the browser's microphone for real-time transcription.
• Shared AI Memory: All personas contribute to and learn from a common memory pool, allowing for richer, context-aware interactions over time.
• Modern, Responsive UI: A stylish and animated user interface featuring a persona selection carousel.

---

Technology Stack

Frontend

• HTML5, CSS3, & JavaScript: Core technologies for the modern and interactive web interface.
• Tailwind CSS: For modern, responsive, and utility-first styling.
• Swiper.js: A powerful library used to create the animated, touch-enabled persona selection carousel.

Backend (Microservice Architecture)

• Docker & Docker Compose: For containerizing the application into portable services and managing the multi-service environment.
• Python: The core language for all server-side logic and API integrations.
• Chatbot Service (emoAI.py): A lightweight Python server that serves the frontend and handles AI and voice transcription logic.
• Data Service (mcp_server.py): A robust Flask & Gunicorn API dedicated to all database operations.

Cloud & AI Services (Google Cloud Platform)

• Google Gemini 1.5 Flash: The core AI model for generating empathetic replies.
• Google Cloud Speech-to-Text: For real-time, high-accuracy voice transcription.
• Google Cloud Firestore: A scalable NoSQL database used to store all user sessions and shared AI memories.
• (Optional) Vertex AI & Natural Language API: The platform is built to easily integrate further Google Cloud AI services.

---

Getting Started

Follow these instructions to get a copy of the project up and running on your local machine.

Prerequisites

You need to have the following software installed:

• Docker & Docker Compose
• Google Cloud SDK (gcloud)

Configuration

1. Clone the repository:

   git clone <your-repository-url>
   cd <your-repository-name>

2. Google Cloud Setup:

   • Create a new project in the Google Cloud Console.
   • Make sure Billing is enabled for your project.
   • Enable the following APIs for your project: Cloud Firestore API, Cloud Speech-to-Text API, and Vertex AI API.

3. Create a Service Account:

   • In your Google Cloud project, navigate to IAM & Admin > Service Accounts.
   • Click + CREATE SERVICE ACCOUNT.
   • Give it a name (e.g., mitr-chatbot-sa) and grant it the Owner role for simplicity during development.
   • Create the account, go to the Keys tab, click Add Key > Create new key, select JSON, and create it.
   • A JSON file will be downloaded. Rename this file to ttsbot-471404-39e4ea704298.json and place it in the root of your project folder.

4. Get Your Gemini API Key:

   • Go to Google AI Studio and create a new API key.
   • Copy this key.

5. Create a Local Environment File:

   • In your project folder, create a new file named .env.
   • Copy the content below into the .env file and paste your Gemini API key.

   # .env
   # Paste your Gemini API key here
   GEMINI_API_KEY=AIzaSyxxxxxxxxxxxxxxxxxxxxxxx

---

Running Locally

1. Make sure Docker Desktop is running.
2. Open a terminal in the project's root directory.
3. Run the following command:

   docker-compose up --build

4. Wait for both services (chatbot-1 and mcp-server-1) to build and start.
5. Open your web browser and navigate to: http://localhost:8080

---

Deployment

This application is configured for easy deployment to Google Cloud Run. The cloudbuild.yaml file and the gcloud deployment scripts used in our development can be adapted for a production CI/CD pipeline.