import subprocess
import sys

# Function to install dependencies if missing
def install(package):
    subprocess.check_call([sys.executable, "-m", "pip", "install", package])

# List of required libraries
required_packages = ["transformers", "torch", "accelerate", "gradio", "huggingface_hub"]

# Install any missing packages
for package in required_packages:
    try:
        __import__(package)
    except ImportError:
        install(package)

import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from huggingface_hub import snapshot_download
import re

# 🔹 Set torch num threads to max
torch.set_num_threads(torch.get_num_threads())

# 🔹 Download & load the model from Hugging Face
model_name = "HyperX-Sen/Qwen-2.5-7B-Reasoning"
model_path = snapshot_download(repo_id=model_name, repo_type="model")

# 🔹 Load the model & tokenizer
model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_path)

# 🔹 System prompt
SYSTEM_PROMPT = """
Respond in the following format:
<reasoning>
...
</reasoning>
<answer>
...
</answer>
"""

# 🔹 Function to extract reasoning and answer
def extract_response(full_response):
    reasoning_match = re.search(r"<reasoning>(.*?)</reasoning>", full_response, re.DOTALL)
    answer_match = re.search(r"<answer>(.*?)</answer>", full_response, re.DOTALL)
    reasoning = reasoning_match.group(1).strip() if reasoning_match else ""
    answer = answer_match.group(1).strip() if answer_match else ""
    return f"<reasoning>\n{reasoning}\n</reasoning>\n<answer>\n{answer}\n</answer>"

# 🔹 Function to generate response
def chat_response(user_input, top_p, top_k, temperature, max_length):
    messages = [
        {"role": "system", "content": f"{SYSTEM_PROMPT}"},
        {"role": "user", "content": user_input}
    ]
    
    input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    inputs = tokenizer(input_text, return_tensors="pt").to(model.device)
    
    with torch.no_grad():
        output = model.generate(
            **inputs,
            max_length=max_length,
            do_sample=True,
            top_p=top_p,
            top_k=top_k,
            temperature=temperature
        )
    
    full_response = tokenizer.decode(output[0], skip_special_tokens=True)
    return extract_response(full_response.replace(SYSTEM_PROMPT, ""))

# 🔹 Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("# 🤖 Qwen-2.5-7B-Reasoning Chatbot")
    
    with gr.Row():
        chatbot = gr.Textbox(label="Model Response", lines=8, interactive=False)
    
    with gr.Row():
        user_input = gr.Textbox(label="Your Prompt", placeholder="Ask me anything...", lines=2)
    
    with gr.Accordion("🔧 Advanced Settings", open=False):
        top_p = gr.Slider(0.1, 1.0, value=0.9, label="Top-p")
        top_k = gr.Slider(1, 100, value=50, label="Top-k")
        temperature = gr.Slider(0.1, 1.5, value=0.7, label="Temperature")
        max_length = gr.Slider(128, 1024, value=512, label="Max Length")
    
    with gr.Row():
        submit_button = gr.Button("Generate Response")
    
    submit_button.click(chat_response, inputs=[user_input, top_p, top_k, temperature, max_length], outputs=[chatbot])

# 🔹 Launch the Gradio app
demo.launch()