selenium_cli.py

#!/usr/bin/env python3
"""
Natural Selenium Agent CLI - Feels like chatting with AI directly
"""

import requests
import time
import json

def check_server():
    """Check if the selenium server is running"""
    try:
        response = requests.get("http://localhost:8001/health")
        return response.status_code == 200
    except:
        return False

def start_chat():
    """Start a natural chat session"""
    base_url = "http://localhost:8001"
    current_task_id = None
    
    print("🤖 **AI Browser Assistant**")
    print("=" * 50)
    print("Hi! I'm your AI browser automation assistant.")
    print("Tell me what you'd like me to do, and I'll handle it step by step.")
    print("(Type 'quit' to exit)")
    print("=" * 50)
    
    while True:
        try:
            # Get user input
            user_input = input("\n💬 You: ").strip()
            
            if user_input.lower() in ['quit', 'exit', 'bye']:
                if current_task_id:
                    print("🔄 Cleaning up...")
                    try:
                        requests.delete(f"{base_url}/api/selenium/task/{current_task_id}")
                    except:
                        pass
                print("👋 See you later!")
                break
            
            if not user_input:
                continue
            
            # If no active task, start a new one
            if not current_task_id:
                print("🤖 AI: Let me help you with that! Starting up my browser...")
                
                # Create new task
                task_resp = requests.post(f"{base_url}/api/selenium/task", 
                                         json={"task": user_input})
                
                if task_resp.status_code == 200:
                    task_data = task_resp.json()
                    current_task_id = task_data["task_id"]
                    
                    # Get initial AI response by checking task status
                    status_resp = requests.get(f"{base_url}/api/selenium/task/{current_task_id}/status")
                    if status_resp.status_code == 200:
                        status_data = status_resp.json()
                        if status_data.get('chat_history'):
                            last_message = status_data['chat_history'][-1]
                            if last_message['role'] == 'assistant':
                                print(f"🤖 AI: {last_message['content']}")
                            else:
                                print("🤖 AI: I'm ready to help! What would you like me to do?")
                        else:
                            print("🤖 AI: I'm ready to help! What would you like me to do?")
                    else:
                        print("🤖 AI: Task started! How can I help you?")
                    
                else:
                    print(f"❌ Sorry, I had trouble starting: {task_resp.text}")
                    continue
            
            else:
                # Send message to existing task
                print("🤖 AI: Let me work on that...")
                
                msg_resp = requests.post(f"{base_url}/api/selenium/task/{current_task_id}/message",
                                        json={"message": user_input})
                
                if msg_resp.status_code == 200:
                    msg_data = msg_resp.json()
                    ai_response = msg_data.get("response", "Working on it...")
                    
                    # Show AI response naturally
                    print(f"🤖 AI: {ai_response}")
                    
                    # Show current status naturally
                    status = msg_data.get("status", "")
                    if status == "completed":
                        print("\n✅ Task completed! Want me to help with anything else?")
                        current_task_id = None  # Ready for new task
                    elif status == "failed":
                        print("\n❌ I ran into some trouble. Let me know what else you'd like to try!")
                        current_task_id = None  # Ready for new task
                    elif status == "user_intervention":
                        print("\n🤔 I could use some guidance - what should I do next?")
                    
                    # Optionally show URL if it's interesting
                    current_url = msg_data.get("current_url", "")
                    if current_url and current_url != "data:," and "google.com" in current_url or "youtube.com" in current_url:
                        print(f"🌐 Currently viewing: {current_url[:80]}...")
                    
                else:
                    print(f"❌ I had trouble processing that: {msg_resp.text}")
        
        except KeyboardInterrupt:
            print("\n👋 Goodbye!")
            if current_task_id:
                try:
                    requests.delete(f"{base_url}/api/selenium/task/{current_task_id}")
                except:
                    pass
            break
        except Exception as e:
            print(f"❌ Something went wrong: {e}")

def main():
    if not check_server():
        print("❌ Selenium server not running!")
        print("Please start it with: python selenium_server.py")
        return
    
    start_chat()

if __name__ == "__main__":
    main()