import gradio as gr
import requests, json, textwrap, datetime

API_URL   = "http://localhost:30000/generate"      # 原生 generate
API_KEY   = "token-abc123"
MODEL_NAME = "Qwen3-14b-base"

def now():
    return datetime.datetime.now().strftime("%H:%M:%S")

# ----------------------------------------
# 主对话函数
# ----------------------------------------
def chat(user_message, history, max_tokens, temperature):
    headers = {
        "Authorization": f"Bearer {API_KEY}",
        "Content-Type":  "application/json"
    }
    payload = {
        "model": MODEL_NAME,
        "text":  user_message,       # generate 使用 text
        "max_tokens": max_tokens,
        "temperature": temperature
    }

    print(f"\n🟡 [{now()} chat] payload:\n{json.dumps(payload, ensure_ascii=False, indent=2)}")

    try:
        resp = requests.post(API_URL, headers=headers, json=payload, timeout=60)
        print(f"🟢 [{now()} chat] status={resp.status_code}")
        print(f"🟢 [{now()} chat] headers={resp.headers}")
        print(f"🟢 [{now()} chat] body前400字={resp.text[:400]}")

        # —— 解析 —— #
        if resp.status_code != 200:
            return f"[⚠️ HTTP {resp.status_code}] {resp.text[:300]}"

        if not resp.text.strip():          # 空 body
            return "[⚠️ 后端返回空响应]"

        try:
            data = json.loads(resp.text)
            return data.get("text", "").strip() or "[⚠️ JSON 中无 text 字段]"
        except json.JSONDecodeError as je:
            snippet = textwrap.shorten(resp.text, width=400, placeholder=" ...")
            return f"[⚠️ JSON 解析失败] {je}\n原始片段:\n{snippet}"

    except Exception as e:
        return f"[❌ 请求异常] {e}"

# ----------------------------------------
# API 测试函数
# ----------------------------------------
def test_api(max_tokens, temperature):
    headers = {
        "Authorization": f"Bearer {API_KEY}",
        "Content-Type":  "application/json"
    }
    payload = {
        "model": MODEL_NAME,
        "text":  "Ping?",
        "max_tokens": max_tokens,
        "temperature": temperature
    }

    print(f"\n🔵 [{now()} test] payload:\n{json.dumps(payload, ensure_ascii=False, indent=2)}")

    try:
        resp = requests.post(API_URL, headers=headers, json=payload, timeout=10)
        print(f"🟢 [{now()} test] status={resp.status_code}")
        print(f"🟢 [{now()} test] body={resp.text}")

        if resp.status_code != 200 or not resp.text.strip():
            return f"⚠️ 状态{resp.status_code} or 空响应: {resp.text[:200]}"

        try:
            out = json.loads(resp.text).get("text", "").strip()
            return f"✅ API 正常，返回: {out or '[空文本]'}"
        except json.JSONDecodeError as je:
            return f"⚠️ JSON 解析失败: {je} 片段:{resp.text[:200]}"

    except Exception as e:
        return f"❌ 请求异常: {e}"

# ----------------------------------------
# Gradio UI
# ----------------------------------------
with gr.Blocks(title="Base 模型测试 UI") as demo:
    gr.Markdown("# 💬 Base 模型对话界面")

    with gr.Row():
        max_tokens  = gr.Slider(32, 32768, value=2048, label="max_tokens")  # 先别直接 8k+
        temperature = gr.Slider(0.0, 1.5, value=0.7, step=0.05, label="temperature")
        test_btn    = gr.Button("🔁 测试 API 可用性")
        test_output = gr.Textbox(label="API 测试结果", interactive=False)

    chatbot = gr.ChatInterface(
        fn=lambda m, h: chat(m, h, max_tokens.value, temperature.value),
        title=None
    )

    test_btn.click(fn=test_api, inputs=[max_tokens, temperature], outputs=test_output)

demo.launch(server_name="0.0.0.0", server_port=30001)