from transformers import pipeline

class LLMHandler:
    def __init__(self):
        self.pipe = pipeline(
            "text2text-generation",
            model="google/flan-t5-small",
            device=-1,  # Force CPU
            torch_dtype="auto"
        )
    
    def generate_response(self, prompt, context):
        full_prompt = f"Context: {context}\nQuestion: {prompt}\nAnswer:"
        return self.pipe(
            full_prompt,
            max_length=512,
            do_sample=True,
            temperature=0.7
        )[0]['generated_text']