import gradio as gr
from huggingface_hub import InferenceClient
import os
import re

# --- Configuration ---
API_TOKEN = os.getenv("HF_TOKEN", None)
MODEL = "Qwen/Qwen2.5-Coder-32B-Instruct"

# --- Initialize Inference Client ---
try:
    print(f"Initializing Inference Client for model: {MODEL}")
    client = InferenceClient(model=MODEL, token=API_TOKEN) if API_TOKEN else InferenceClient(model=MODEL)
except Exception as e:
    raise gr.Error(f"Failed to initialize model client. Error: {e}")

# --- Core Code Generation Function ---
def generate_code(
    prompt: str,
    backend_choice: str,
    max_tokens: int,
    temperature: float,
    top_p: float,
):
    print(f"Generating code for: {prompt[:100]}... | Backend: {backend_choice}")

    # --- Dynamically Build System Message ---
    system_message = (
        "you are an ai that is supposed to generate websites, you must not say anything except giving code , "
        "user can select backend like static , flask , nodejs only , you should always keep the website sfw and minimal errors, "
        "you must create an index.html following the user prompt, "
        "if the user asks you create an code that's not about an website you should say "
        "'hey there! am here to create websites for you unfortunately am programmed to not create codes! otherwise I would go on the naughty list :-(', "
        "your code always must have no useless comments you should only add comments where users are required to modify the code."
    )

    user_prompt = f"USER_PROMPT = {prompt}\nUSER_BACKEND = {backend_choice}"

    messages = [
        {"role": "system", "content": system_message},
        {"role": "user", "content": user_prompt}
    ]

    response_stream = ""
    full_response = ""

    try:
        stream = client.chat_completion(
            messages=messages,
            max_tokens=max_tokens,
            stream=True,
            temperature=temperature,
            top_p=top_p,
        )
        for message in stream:
            token = message.choices[0].delta.content
            if isinstance(token, str):
                response_stream += token
                full_response += token
                yield response_stream

        cleaned_response = full_response.strip()
        cleaned_response = re.sub(r"^\s*```[a-z]*\s*\n?", "", cleaned_response)
        cleaned_response = re.sub(r"\n?\s*```\s*$", "", cleaned_response)
        cleaned_response = re.sub(r"<\s*\|?\s*(user|assistant)\s*\|?\s*>", "", cleaned_response, flags=re.IGNORECASE)

        common_phrases = [
            "Here is the code:", "Okay, here is the code:", "Here's the code:",
            "Sure, here is the code you requested:", "Let me know if you need anything else."
        ]
        for phrase in common_phrases:
            if cleaned_response.lower().startswith(phrase.lower()):
                cleaned_response = cleaned_response[len(phrase):].lstrip()

        yield cleaned_response.strip()

    except Exception as e:
        yield f"## Error\n\nFailed to generate code.\n**Reason:** {e}"

# --- Build Gradio Interface ---
with gr.Blocks(css=".gradio-container { max-width: 90% !important; }") as demo:
    gr.Markdown("# ✨ Website Code Generator ✨")
    gr.Markdown(
        "Describe the website you want. The AI will generate a **single-file** `index.html` website.\n\n"
        "**Rules:**\n"
        "- Backend hint (Static / Flask / Node.js).\n"
        "- Always fully SFW and minimal errors.\n"
        "- Only generates websites. No other codes.\n"
        "- Minimal necessary comments only."
    )

    with gr.Row():
        with gr.Column(scale=2):
            prompt_input = gr.Textbox(
                label="Website Description",
                placeholder="e.g., A simple landing page with a hero section and contact form.",
                lines=6,
            )
            backend_radio = gr.Radio(
                ["Static", "Flask", "Node.js"],
                label="Backend Context",
                value="Static",
                info="Hint only. Always generates only index.html."
            )
            generate_button = gr.Button("✨ Generate Website Code", variant="primary")

        with gr.Column(scale=3):
            code_output = gr.Code(
                label="Generated index.html",
                language="html",
                lines=30,
                interactive=False,
            )

    with gr.Accordion("Advanced Settings", open=False):
        max_tokens_slider = gr.Slider(
            minimum=512,
            maximum=4096,
            value=3072,
            step=256,
            label="Max New Tokens"
        )
        temperature_slider = gr.Slider(
            minimum=0.1, maximum=1.2, value=0.7, step=0.1, label="Temperature"
        )
        top_p_slider = gr.Slider(
            minimum=0.1, maximum=1.0, value=0.9, step=0.05, label="Top-P"
        )

    generate_button.click(
        fn=generate_code,
        inputs=[prompt_input, backend_radio, max_tokens_slider, temperature_slider, top_p_slider],
        outputs=code_output,
    )

if __name__ == "__main__":
    demo.queue(max_size=10).launch()