from transformers import AutoModelForCausalLM, AutoTokenizer
import torch


class EndpointHandler:
    def __init__(self, path="/repository"):
        self.tokenizer = AutoTokenizer.from_pretrained(path)
        self.model = AutoModelForCausalLM.from_pretrained(
            path,
            torch_dtype=torch.bfloat16,
            device_map="auto",
        )
        self.model.eval()

    def __call__(self, data):
        inputs = data.get("inputs", data)
        parameters = data.get("parameters", {})

        max_new_tokens = parameters.get("max_new_tokens", 256)
        temperature = parameters.get("temperature", 0.7)

        if isinstance(inputs, dict) and "messages" in inputs:
            messages = inputs["messages"]
            # Step 1: apply template to get a STRING (tokenize=False)
            prompt = self.tokenizer.apply_chat_template(
                messages,
                tokenize=False,
                add_generation_prompt=True,
            )
            # Step 2: tokenize the string
            input_ids = self.tokenizer.encode(
                prompt, return_tensors="pt"
            ).to(self.model.device)
        else:
            text = inputs if isinstance(inputs, str) else str(inputs)
            input_ids = self.tokenizer.encode(
                text, return_tensors="pt"
            ).to(self.model.device)

        with torch.no_grad():
            output = self.model.generate(
                input_ids,
                max_new_tokens=max_new_tokens,
                temperature=temperature,
                do_sample=True,
            )

        result = self.tokenizer.decode(
            output[0][input_ids.shape[-1]:], skip_special_tokens=True
        )
        return [{"generated_text": result}]