import gradio as gr
import requests
import os

HF_TOKEN = os.environ.get("HF_TOKEN", "")
API_URL = "https://api-inference.huggingface.co/models/rahimdzx/AraCode-7B-Full"

SYSTEM = "أنت AraCode، مساعد برمجي ذكي باللغة العربية. تشرح وتكتب وتصحح الكود بدقة."

def chat(message, history, temperature, max_tokens):
    prompt = f"<|im_start|>system\n{SYSTEM}<|im_end|>\n"
    for item in history:
        prompt += f"<|im_start|>{item['role']}\n{item['content']}<|im_end|>\n"
    prompt += f"<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n"
    headers = {"Authorization": f"Bearer {HF_TOKEN}"}
    try:
        r = requests.post(API_URL, headers=headers, json={
            "inputs": prompt,
            "parameters": {"max_new_tokens": int(max_tokens), "temperature": float(temperature), "return_full_text": False}
        }, timeout=60)
        if r.status_code == 200:
            result = r.json()
            yield result[0].get("generated_text", "لا يوجد رد") if isinstance(result, list) else str(result)
        elif r.status_code == 503:
            yield "⏳ النموذج يستيقظ... أعد المحاولة بعد 30 ثانية"
        else:
            yield f"❌ خطأ {r.status_code}"
    except Exception as e:
        yield f"❌ {str(e)}"

with gr.Blocks(title="AraCode-7B", theme=gr.themes.Soft()) as demo:
    gr.HTML("<h1 style='text-align:center'>🐪 AraCode-7B — مساعد البرمجة العربي</h1>")
    with gr.Row():
        with gr.Column(scale=4):
            chatbot = gr.Chatbot(height=480, type="messages", rtl=True, show_copy_button=True)
            msg = gr.Textbox(placeholder="اكتب سؤالك هنا...", lines=2)
            with gr.Row():
                send = gr.Button("إرسال 🚀", variant="primary")
                clear = gr.Button("🗑️ مسح")
        with gr.Column(scale=1):
            temperature = gr.Slider(0.1, 1.0, 0.3, step=0.1, label="الإبداع")
            max_tokens = gr.Slider(128, 1024, 512, step=128, label="طول الرد")
            gr.Markdown("🔗 [النموذج](https://huggingface.co/rahimdzx/AraCode-7B-Full)")
    gr.Examples([
        ["اكتب دالة بايثون للبحث الثنائي"],
        ["ما الفرق بين list و tuple؟"],
        ["اكتب API بسيط بـ Flask"],
    ], inputs=msg)

    def submit(m, h): return "", h + [{"role":"user","content":m}]
    def respond(h, t, k):
        if not h: return h
        h = h + [{"role":"assistant","content":""}]
        for c in chat(h[-2]["content"], h[:-1], t, k):
            h[-1]["content"] = c
            yield h

    msg.submit(submit,[msg,chatbot],[msg,chatbot]).then(respond,[chatbot,temperature,max_tokens],chatbot)
    send.click(submit,[msg,chatbot],[msg,chatbot]).then(respond,[chatbot,temperature,max_tokens],chatbot)
    clear.click(lambda:[],None,chatbot)

demo.launch(server_name="0.0.0.0", server_port=7860)