main.py

"""
AI Email Agent — Built with AgentMail + OpenAI

This template creates an AI-powered email agent that can:
1. Create its own email inbox
2. Send emails
3. Receive and process incoming emails
4. Reply intelligently using GPT-4
5. Handle webhooks for real-time email processing

Get your API keys:
- AgentMail: https://console.agentmail.to (free)
- OpenAI: https://platform.openai.com/api-keys

Set AGENTMAIL_API_KEY and OPENAI_API_KEY in Replit Secrets.
"""

import os
import json
from agentmail import AgentMail
from openai import OpenAI
from flask import Flask, request

# Initialize clients
agentmail = AgentMail(api_key=os.environ.get("AGENTMAIL_API_KEY"))
openai_client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))

app = Flask(__name__)

# System prompt for your AI agent
SYSTEM_PROMPT = """You are a helpful AI email assistant. 
Reply to emails concisely and helpfully.
Keep replies under 3 paragraphs.
Be professional but friendly."""


def create_agent_inbox():
    """Create a new email inbox for the agent."""
    inbox = agentmail.inboxes.create(display_name="AI Email Agent")
    print(f"✅ Agent inbox created: {inbox.inbox_id}")
    return inbox


def send_email(inbox_id: str, to: str, subject: str, text: str):
    """Send an email from the agent's inbox."""
    message = agentmail.inboxes.messages.send(
        inbox_id=inbox_id,
        to=to,
        subject=subject,
        text=text,
    )
    print(f"📤 Sent email to {to}: {subject}")
    return message


def process_incoming_email(inbox_id: str, message_id: str):
    """Process an incoming email and generate an AI reply."""
    
    # Get the message
    msg = agentmail.inboxes.messages.get(
        inbox_id=inbox_id, message_id=message_id
    )
    
    # Use extracted_text to get clean content (strips quoted text from replies)
    customer_message = msg.extracted_text or msg.text
    print(f"📥 Received from {msg.from_}: {customer_message[:100]}...")
    
    # Get thread for conversation context
    thread = agentmail.inboxes.threads.get(
        inbox_id=inbox_id, thread_id=msg.thread_id
    )
    
    # Build conversation history
    conversation = []
    for thread_msg in thread.messages:
        role = "assistant" if thread_msg.from_ == inbox_id else "user"
        content = thread_msg.extracted_text or thread_msg.text
        if content:
            conversation.append({"role": role, "content": content})
    
    # Generate reply with GPT-4
    response = openai_client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": SYSTEM_PROMPT},
            *conversation,
        ],
    )
    
    reply_text = response.choices[0].message.content
    
    # Send reply
    reply = agentmail.inboxes.messages.reply(
        inbox_id=inbox_id, message_id=message_id, text=reply_text
    )
    
    print(f"💬 Replied: {reply_text[:100]}...")
    return reply


# Webhook endpoint for real-time email processing
@app.route("/webhook", methods=["POST"])
def webhook():
    """Handle incoming email events from AgentMail."""
    event = request.json
    
    if event.get("type") == "message.received":
        data = event["data"]
        process_incoming_email(
            inbox_id=data["inbox_id"],
            message_id=data["message_id"],
        )
    
    return "", 200


@app.route("/", methods=["GET"])
def index():
    return """
    <h1>🤖 AI Email Agent</h1>
    <p>Built with <a href="https://agentmail.to">AgentMail</a> + OpenAI</p>
    <p>Webhook endpoint: <code>/webhook</code></p>
    """


if __name__ == "__main__":
    # Create an inbox on startup
    inbox = create_agent_inbox()
    
    print(f"\n🤖 AI Email Agent is running!")
    print(f"📧 Send emails to: {inbox.inbox_id}")
    print(f"🔗 Webhook URL: Set your Replit URL + /webhook in AgentMail dashboard\n")
    
    # Start the webhook server
    app.run(host="0.0.0.0", port=5000)