import torch
from transformers import pipeline

class DummyLLM:
    """A simple dummy LLM for testing."""
    
    def __init__(self):
        """Initialize the dummy LLM."""
        pass
    
    def complete(self, prompt):
        """Complete a prompt with a simple response."""
        class Response:
            def __init__(self, text):
                self.text = text
        
        # For testing only - return a simple response
        return Response("This is a placeholder response. The actual model is not loaded to save resources.")

def setup_llm():
    """Set up a simple LLM for testing."""
    try:
        # Try to load a very small model for text generation
        generator = pipeline(
            "text-generation",
            model="distilgpt2",  # A very small model
            max_length=100
        )
        
        # Create a wrapper class that matches the expected interface
        class SimpleTransformersLLM:
            def complete(self, prompt):
                class Response:
                    def __init__(self, text):
                        self.text = text
                
                try:
                    result = generator(prompt, max_length=len(prompt) + 50, do_sample=True)[0]
                    generated_text = result["generated_text"]
                    response_text = generated_text[len(prompt):].strip()
                    if not response_text:
                        response_text = "I couldn't generate a proper response."
                    return Response(response_text)
                except Exception as e:
                    print(f"Error generating response: {e}")
                    return Response("Error generating response.")
        
        return SimpleTransformersLLM()
    
    except Exception as e:
        print(f"Error setting up model: {e}")
        return DummyLLM()