🧠 AI Calendar Assistant (Agentic, Multi-user)

This project is a fully agentic calendar scheduling assistant using Fetch.ai uAgents, Streamlit for the UI, OpenAI for natural language interpretation, and Google Calendar API for calendar operations.

---

⚙️ Architecture

• streamlit_ui.py – Streamlit app for user login, prompt submission, and confirmation interface
• frontend_agent.py – Formats and sends CalendarIntent to backend
• calendar_agent.py – Handles scheduling logic, conflict checking, and event creation via Google Calendar
• calendar_intent.py – Shared message schema

---

🚀 Features

• 🧠 Natural language to intent conversion using OpenAI
• 🔐 Per-user Google Calendar login (OAuth2)
• 🤖 Agentic backend with automatic conflict checking
• ✅ Human-in-the-loop: event is only booked on user confirmation
• 📨 Multi-turn agent messaging using Fetch.ai's protocol system

---

📦 Setup

1. Clone the repo:

git clone https://github.com/your-username/calendar-agentic-assistant
cd calendar-agentic-assistant

2. Create .streamlit/secrets.toml:

GOOGLE_CLIENT_ID = "your-client-id"
GOOGLE_CLIENT_SECRET = "your-client-secret"

3. Install dependencies:

pip install -r requirements.txt

4. Start the agents:

python agents/calendar_agent.py
python agents/frontend_agent.py

5. Start the frontend:

streamlit run streamlit_ui.py

---

🛡️ Environment Variables (used in secrets.toml)

• GOOGLE_CLIENT_ID – From Google Cloud Console (OAuth credentials)
• GOOGLE_CLIENT_SECRET – From Google Cloud Console

---

💡 Example Prompt

"Schedule a meeting with Alex tomorrow at 3PM"

The LLM will convert this into a structured JSON intent. The calendar agent checks for conflicts and asks for confirmation if the slot is free.

---

🧠 Extendable Ideas

• Add group availability negotiation
• Use persistent storage for confirmed intent history
• Add Slack or SMS notification integration
• Support for recurring events

---

Built with ❤️ using Fetch.ai, Streamlit, OpenAI, and Python.