import gradio as gr

import os, sys
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig, pipeline
import torch

# Define the model repository
REPO_NAME = 'schuler/experimental-JP47D20'
# REPO_NAME = 'schuler/experimental-JP47D21-KPhi-3-micro-4k-instruct'

# How to cache?
def load_model(repo_name):
    tokenizer = AutoTokenizer.from_pretrained(repo_name, trust_remote_code=True)
    generator_conf = GenerationConfig.from_pretrained(repo_name)
    model = AutoModelForCausalLM.from_pretrained(repo_name, trust_remote_code=True, torch_dtype=torch.bfloat16)
    return tokenizer, generator_conf, model

tokenizer, generator_conf, model = load_model(REPO_NAME)

global_error = ''
try:
    generator = pipeline("text-generation", model=model, tokenizer=tokenizer)
except Exception as e:
    global_error =  f"Failed to load model: {str(e)}"

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    result = 'none'
    try:
        # Build the conversation prompt
        prompt = ''
        messages = []
        if (len(system_message)>0):
            prompt = "<|assistant|>"+system_message+f"<|end|>\n"    
        for val in history:
            if val[0]:
                messages.append({"role": "user", "content": val[0]})
            if val[1]:
                messages.append({"role": "assistant", "content": val[1]})
    
        messages.append({"role": "user", "content": message})
    
        for hmessage in messages:
            role = "<|assistant|>" if hmessage['role'] == 'assistant' else "<|user|>"
            prompt += f"{role}{hmessage['content']}<|end|>"        
        # prompt += f"<|user|>{message}<|end|><|assistant|>"
        prompt += f"<|assistant|>"

        # """
        # Generate the response
        response_output = generator(
            prompt,
            generation_config=generator_conf,
            max_new_tokens=max_tokens,
            do_sample=True,
            top_p=top_p,
            repetition_penalty=1.2
        )
    
        generated_text = response_output[0]['generated_text']
    
        # Extract the assistant's response
        result = generated_text[len(prompt):].strip()
        # """
        # result = prompt +':'+result
    except Exception as error:
        exc_type, exc_obj, exc_tb = sys.exc_info()
        fname = os.path.split(exc_tb.tb_frame.f_code.co_filename)[1]
        result = str(error) +':'+ exc_type +':'+ fname +':'+ exc_tb.tb_lineno

    yield result

    
    """
    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content

        response += token
        yield response
    """


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot." + global_error, label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=64, step=1, label="Max new tokens"),
        # gr.Slider(minimum=0.1, maximum=4.0, value=1.0, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.25,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)


if __name__ == "__main__":
    demo.launch()