Digital Twin AI Personal Assistant

This project is a sophisticated, proactive AI personal assistant built using the Qwen Large Language Model and the Qwen-Agent framework. The "Digital Twin" is designed to learn its user's unique communication style and assist with managing their digital life, including emails and calendar schedules.

Core Features

• Interactive Chat: A command-line interface for real-time, conversational interaction with the agent.
• Persona Learning: The agent ingests the user's sent emails to learn their specific writing style, tone, and phrasing. This is achieved through Retrieval-Augmented Generation (RAG) with a ChromaDB vector store.
• Gmail & Calendar Integration: Securely connects to Google services using OAuth 2.0 to:
  • Read calendar schedules.
  • Summarize unread emails.
  • Draft emails that mimic the user's persona.
  • Send emails and create calendar events with user confirmation.
• Proactive Assistance: A dedicated mode where the agent analyzes the user's current context (new emails and events) to suggest actions, such as scheduling a meeting mentioned in an email.
• Safety First: Implements a strict confirmation protocol. The agent will always ask for user permission before taking any write-action (sending an email, creating an event).

Technical Architecture

• LLM Framework: qwen-agent
• Core LLM: qwen-max (or other powerful model) via Alibaba Cloud
• Vector Database: ChromaDB for local, persistent persona storage
• Embeddings: sentence-transformers
• API Authentication: Google OAuth 2.0 for secure service access

Create a simple diagram showing how the Agent, Tools, and Vector DB interact, and add it here.

Setup & Installation

1. Clone the repository:

   git clone [your-repo-url]
   cd digital_twin_agent

2. Create and activate a virtual environment:

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

3. Install dependencies:

   pip install -r requirements.txt

4. Configure Google Cloud & API Keys:

   • Follow the Google Cloud setup instructions to enable the Gmail and Calendar APIs and download your credentials.json file. Place it in the project root.
   • Create a .env file in the project root and add your Model Studio API credentials:

     MODEL_STUDIO_URL="your_model_studio_api_endpoint_url_here"
     MODEL_STUDIO_API_KEY="your_model_studio_api_key_here"

How to Use the Digital Twin

Step 1: First-Time Authentication

Run the auth.py script once to authorize the application with your Google account. This will open a browser window for you to log in and consent.

python -m core.auth

Step 2: Learn Your Persona

Run the ingestion script once to populate the vector database with your writing style from your 50 most recent sent emails.

python run_ingestion.py

Step 3: Interact with Your Agent

Start the interactive chat loop to talk to your assistant.

python -m core.agent

You can ask things like:

• "What's on my schedule for tomorrow?"
• "Any new emails?"
• "Draft an email to my colleague about the project deadline."
• "Send an email to test@example.com..." (will ask for confirmation)

Step 4: Get Proactive Suggestions

Run the proactive script to have the agent analyze your current situation and suggest actions.

python run_proactive_assistant.py