app.py

import gradio as gr
from model_loader import load_model
from assistant import get_assistant_response

# Load model pipeline
pipe = load_model()

# Wrapper to handle dynamic output length
def smart_response(command, execute):
    # Dynamically adjust max_new_tokens based on command complexity
    simple_keywords = ["hi", "hello", "open", "close", "shutdown", "restart", "lock", "screenshot"]
    is_simple = any(keyword in command.lower() for keyword in simple_keywords)
    max_tokens = 64 if is_simple else 512  # fast replies for simple commands

    return get_assistant_response(pipe, command, execute, max_new_tokens=max_tokens)

# Gradio UI
demo = gr.Interface(
    fn=smart_response,
    inputs=[
        gr.Textbox(
            lines=2,
            placeholder="e.g. Open Chrome or Take a screenshot",
            label="🧾 Command"
        ),
        gr.Checkbox(label="🛠️ Execute command (if possible)")
    ],
    outputs="text",
    title="🧠 Smart AI Assistant",
    description=(
        "AI assistant powered by **Qwen/Qwen3-1.7B** 🧠\n\n"
        "Try commands like:\n"
        "• 'Open Chrome'\n"
        "• 'Take a screenshot'\n"
        "• 'What is AGI?'\n"
        "• 'Summarize the history of AI'\n\n"
        "✅ Simple commands → fast replies\n"
        "🧠 Complex questions → thoughtful answers\n"
        "⚠️ Check the box to allow real OS command execution"
    ),
    flagging_mode="never"
)

if __name__ == "__main__":
    demo.launch(share=False, server_name="0.0.0.0", server_port=7860)