app.py

# -*- coding: utf-8 -*-
"""
UltraData-Math L3 Generator - Hugging Face Space Demo
"""

import os
import asyncio
import gradio as gr

from openai import AsyncOpenAI

from qa_synthesis import QA_PROMPTS, get_qa_prompt
from conversation_synthesis import CONVERSATION_PROMPTS, get_conversation_prompt
from multistyle_rewrite import MULTISTYLE_PROMPTS, get_multistyle_prompt
from knowledge_textbook import (
    get_knowledge_extraction_prompt,
    get_textbook_exercise_prompt,
    TEXTBOOK_EXERCISE_PROMPTS,
)
from run_synthesis import (
    parse_qa_output,
    parse_conversation_output,
    parse_rewrite_output,
    parse_knowledge_output,
    parse_textbook_output,
)

# API 配置从环境变量读取（通过 HF Secrets 设置）
API_KEY = os.getenv("OPENAI_API_KEY")
BASE_URL = os.getenv("OPENAI_BASE_URL", "https://llm-center.ali.modelbest.cn/llm/openai/v1")
DEFAULT_MODEL = "GEMINI_anxt74"

# 示例数据
EXAMPLE_MATH_CONTENT = """The quadratic formula is a fundamental result in algebra that provides the solutions to any quadratic equation of the form ax² + bx + c = 0, where a ≠ 0.

The formula states that the solutions are:
x = (-b ± √(b² - 4ac)) / (2a)

The term b² - 4ac is called the discriminant. It determines the nature of the roots:
- If b² - 4ac > 0, there are two distinct real roots
- If b² - 4ac = 0, there is exactly one real root (a repeated root)
- If b² - 4ac < 0, there are two complex conjugate roots

This formula was known to ancient mathematicians and remains one of the most important tools in solving polynomial equations."""

EXAMPLE_KNOWLEDGE_POINT = """Definition: A continuous function is a function f: R → R such that for every point x₀ in its domain and every ε > 0, there exists a δ > 0 such that |f(x) - f(x₀)| < ε whenever |x - x₀| < δ.

Key Properties:
1. The sum, difference, and product of continuous functions are continuous
2. The composition of continuous functions is continuous
3. A continuous function on a closed interval attains its maximum and minimum values (Extreme Value Theorem)
4. A continuous function on a closed interval takes on every value between its minimum and maximum (Intermediate Value Theorem)"""


async def call_api(prompt: str, temperature: float = 0.7) -> str:
    """调用 API 生成内容"""
    if not API_KEY:
        return "Error: API Key not configured. Please contact administrator."
    
    client = AsyncOpenAI(api_key=API_KEY, base_url=BASE_URL)
    try:
        response = await client.chat.completions.create(
            model=DEFAULT_MODEL,
            messages=[{"role": "user", "content": prompt}],
            temperature=temperature,
            max_tokens=8192,
        )
        # 处理 reasoning model 的返回格式
        message = response.choices[0].message
        content = message.content
        # 如果 content 为空，尝试获取 reasoning_content
        if not content and hasattr(message, 'reasoning_content') and message.reasoning_content:
            content = message.reasoning_content
        return content or ""
    except Exception as e:
        return f"Error: {str(e)}"


def run_async(coro):
    """运行异步函数"""
    try:
        loop = asyncio.get_event_loop()
    except RuntimeError:
        loop = asyncio.new_event_loop()
        asyncio.set_event_loop(loop)
    return loop.run_until_complete(coro)


# ============================================================================
# Task Handlers
# ============================================================================

def qa_synthesis(text: str, level: str):
    """Q&A 问答对合成"""
    if not text.strip():
        return "", "", ""
    
    prompt_template = get_qa_prompt(level)
    prompt = prompt_template.format(text=text)
    
    response = run_async(call_api(prompt))
    parsed = parse_qa_output(response)
    
    return (
        parsed.get("problem", ""),
        parsed.get("solution", ""),
        response
    )


def conversation_synthesis(text: str, style: str):
    """多轮对话合成"""
    if not text.strip():
        return "", ""
    
    prompt_template = get_conversation_prompt(style)
    prompt = prompt_template.format(text=text)
    
    response = run_async(call_api(prompt))
    parsed = parse_conversation_output(response)
    
    return parsed.get("content", response), response


def rewrite_synthesis(text: str, style: str):
    """多风格改写"""
    if not text.strip():
        return "", ""
    
    prompt_template = get_multistyle_prompt(style)
    prompt = prompt_template.format(text=text)
    
    response = run_async(call_api(prompt))
    parsed = parse_rewrite_output(response)
    
    return parsed.get("rewritten", response), response


def knowledge_extraction(text: str):
    """知识点提取"""
    if not text.strip():
        return "", ""
    
    prompt_template = get_knowledge_extraction_prompt()
    prompt = prompt_template.format(text=text)
    
    response = run_async(call_api(prompt))
    parsed = parse_knowledge_output(response)
    
    knowledge_points = parsed.get("knowledge_points", [])
    formatted = "\n\n---\n\n".join(knowledge_points) if knowledge_points else "No knowledge points extracted."
    
    return formatted, response


def textbook_exercise(knowledge_point: str, difficulty: str):
    """教材练习生成"""
    if not knowledge_point.strip():
        return "", ""
    
    prompt_template = get_textbook_exercise_prompt(difficulty)
    prompt = prompt_template.format(mathematical_knowledge_point=knowledge_point)
    
    response = run_async(call_api(prompt))
    parsed = parse_textbook_output(response)
    
    return parsed.get("material", response), response


# ============================================================================
# Gradio UI
# ============================================================================

custom_css = """
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;600;700&family=JetBrains+Mono:wght@400;500&display=swap');

:root {
    --bg: #f8fafc;
    --surface: #ffffff;
    --surface-2: #f1f5f9;
    --border: #e2e8f0;
    --text: #0f172a;
    --muted: #64748b;
    --accent: #4f46e5;
    --accent-2: #6366f1;
}

body {
    background-color: var(--bg);
    color: var(--text);
}

.gradio-container {
    font-family: 'Inter', sans-serif !important;
    background: linear-gradient(180deg, #f8fafc 0%, #eef2ff 100%) !important;
    max-width: 1200px !important;
}

/* Title & Header */
.main-title {
    font-family: 'Inter', sans-serif !important;
    font-weight: 800 !important;
    font-size: 2.6rem !important;
    background: linear-gradient(90deg, #0f172a, #4f46e5, #7c3aed) !important;
    -webkit-background-clip: text !important;
    -webkit-text-fill-color: transparent !important;
    text-align: center !important;
    margin-bottom: 0.4rem !important;
}

.subtitle {
    text-align: center !important;
    color: var(--muted) !important;
    font-size: 1.05rem !important;
    margin-bottom: 2.5rem !important;
    font-weight: 400 !important;
}

/* Panels */
.glass-panel {
    background: var(--surface) !important;
    border: 1px solid var(--border) !important;
    border-radius: 16px !important;
    padding: 24px !important;
    box-shadow: 0 10px 30px rgba(15, 23, 42, 0.08) !important;
}

/* Labels */
.block > label > span, 
.form > label > span, 
.gr-form > label > span,
.label-wrap > span {
    color: var(--text) !important;
    font-weight: 600 !important;
    font-size: 1rem !important;
    margin-bottom: 0.5rem !important;
    text-shadow: none !important;
}

/* Radio group title */
fieldset legend, fieldset legend span,
.gr-radio > label, .gr-radio > label span,
.gradio-container .label-wrap, .gradio-container .label-wrap span {
    color: var(--text) !important;
    font-weight: 600 !important;
    text-shadow: none !important;
}

/* Info Text (Description) */
span.description, .description {
    color: var(--muted) !important;
    font-weight: 500 !important;
    text-shadow: none !important;
    opacity: 1 !important;
}

/* Radio/Checkbox alignment */
fieldset label span {
    margin-bottom: 0 !important;
    text-shadow: none !important;
    font-weight: 500 !important;
    color: var(--text) !important;
    display: flex !important;
    align-items: center !important;
}

fieldset label.selected span {
    color: #ffffff !important;
}

fieldset label.selected {
    background: var(--accent) !important;
    border-color: var(--accent) !important;
}

/* Inputs & Textareas */
.gr-input, textarea, input, .gr-box, .gr-check-radio, .gr-dropdown {
    font-family: 'JetBrains Mono', monospace !important;
    background-color: var(--surface) !important;
    border: 1px solid var(--border) !important;
    color: var(--text) !important;
    box-shadow: none !important;
}

.gr-input:focus, textarea:focus, input:focus {
    border-color: var(--accent) !important;
    box-shadow: 0 0 0 2px rgba(79, 70, 229, 0.15) !important;
}

/* Dropdown options */
ul.options, .gr-dropdown-options {
    background-color: var(--surface) !important;
    color: var(--text) !important;
    border: 1px solid var(--border) !important;
}

/* Markdown prose */
.prose, .prose p, .prose h1, .prose h2, .prose h3, .prose strong, .prose li {
    color: var(--text) !important;
}

/* Outputs */
.output-textbox textarea {
    background-color: var(--surface-2) !important;
    border: 1px solid var(--border) !important;
    border-radius: 8px !important;
    color: var(--text) !important;
}

.markdown-box {
    background: var(--surface-2) !important;
    border: 1px solid var(--border) !important;
    border-radius: 8px !important;
    padding: 16px !important;
    color: var(--text) !important;
}

.markdown-box * {
    color: var(--text) !important;
}

.markdown-box code, .markdown-box pre {
    background: #e2e8f0 !important;
}

/* Buttons */
.gr-button-primary {
    background: linear-gradient(135deg, var(--accent) 0%, var(--accent-2) 100%) !important;
    border: none !important;
    color: #ffffff !important;
    font-weight: 600 !important;
    box-shadow: 0 6px 14px rgba(79, 70, 229, 0.25) !important;
}

.gr-button-secondary {
    background: #f1f5f9 !important; /* clearer separation from page background */
    border: 1px solid #cbd5e1 !important;
    color: #0f172a !important;
    box-shadow: 0 4px 10px rgba(15, 23, 42, 0.08) !important;
}
.gr-button-secondary:hover {
    background: #e2e8f0 !important;
    border-color: #94a3b8 !important;
}

/* Tabs */
.tabs button {
    color: var(--muted) !important;
}
.tabs button.selected {
    color: var(--accent) !important;
    border-bottom-color: var(--accent) !important;
}

/* Radio buttons */
.gr-radio-label {
    color: var(--text) !important;
}

/* Footer */
.footer-text, .footer-text p {
    color: var(--muted) !important;
}
.footer-text a {
    color: var(--accent) !important;
}
"""

with gr.Blocks(title="UltraData-Math L3 Generator", css=custom_css, theme=gr.themes.Soft()) as demo:
    gr.HTML('<h1 class="main-title">🧮 UltraData-Math L3 Generator</h1>')
    gr.HTML('<p class="subtitle">✨ Next-Gen Mathematical Data Synthesis Powered by LLM ✨</p>')
    
    with gr.Tabs():
        # Q&A Synthesis Tab
        with gr.TabItem("📝 Q&A Synthesis"):
            with gr.Column(elem_classes=["glass-panel"]):
                gr.Markdown("### 💡 Transform Text into Q&A Pairs\nGenerate high-quality question-answer pairs from mathematical content, tailored to different educational levels.")
                with gr.Row():
                    with gr.Column(scale=1):
                        qa_input = gr.Textbox(
                            label="Input Mathematical Content",
                            placeholder="Paste your mathematical text here (e.g., definitions, theorems, proofs)...",
                            lines=10,
                        )
                        qa_level = gr.Radio(
                            choices=list(QA_PROMPTS.keys()),
                            value="high_school",
                            label="Difficulty Level",
                            info="Select the target audience level"
                        )
                        with gr.Row():
                            qa_example_btn = gr.Button("📋 Load Example", variant="secondary")
                            qa_btn = gr.Button("🚀 Generate Q&A", variant="primary")
                    
                    with gr.Column(scale=1):
                        qa_problem = gr.Textbox(label="Generated Problem", lines=5, show_copy_button=True)
                        qa_solution = gr.Textbox(label="Generated Solution", lines=12, show_copy_button=True)
                        qa_raw = gr.Textbox(label="Raw Response", lines=4, visible=False)
            
            qa_example_btn.click(
                lambda: EXAMPLE_MATH_CONTENT,
                outputs=[qa_input],
            )
            qa_btn.click(
                qa_synthesis,
                inputs=[qa_input, qa_level],
                outputs=[qa_problem, qa_solution, qa_raw],
            )
        
        # Conversation Synthesis Tab
        with gr.TabItem("💬 Conversation Synthesis"):
            with gr.Column(elem_classes=["glass-panel"]):
                gr.Markdown("### 🗣️ Create Multi-turn Dialogues\nConvert static mathematical text into dynamic, engaging multi-turn conversations between students and teachers.")
                with gr.Row():
                    with gr.Column(scale=1):
                        conv_input = gr.Textbox(
                            label="Input Mathematical Content",
                            placeholder="Paste your mathematical text here...",
                            lines=10,
                        )
                        conv_style = gr.Radio(
                            choices=list(CONVERSATION_PROMPTS.keys()),
                            value="teacher_student",
                            label="Conversation Style",
                            info="Choose the persona and tone of the conversation"
                        )
                        with gr.Row():
                            conv_example_btn = gr.Button("📋 Load Example", variant="secondary")
                            conv_btn = gr.Button("🚀 Generate Conversation", variant="primary")
                    
                    with gr.Column(scale=1):
                        conv_output = gr.Textbox(label="Generated Conversation", lines=20, show_copy_button=True)
                        conv_raw = gr.Textbox(label="Raw Response", lines=4, visible=False)
            
            conv_example_btn.click(
                lambda: EXAMPLE_MATH_CONTENT,
                outputs=[conv_input],
            )
            conv_btn.click(
                conversation_synthesis,
                inputs=[conv_input, conv_style],
                outputs=[conv_output, conv_raw],
            )
        
        # Rewrite Tab
        with gr.TabItem("✨ Multi-style Rewrite"):
            with gr.Column(elem_classes=["glass-panel"]):
                gr.Markdown("### 🎨 Style Transfer\nRewrite mathematical content into various styles, from rigorous textbooks to engaging blog posts.")
                with gr.Row():
                    with gr.Column(scale=1):
                        rewrite_input = gr.Textbox(
                            label="Input Mathematical Content",
                            placeholder="Paste your mathematical text here...",
                            lines=10,
                        )
                        rewrite_style = gr.Radio(
                            choices=list(MULTISTYLE_PROMPTS.keys()),
                            value="textbook",
                            label="Target Style",
                            info="Select the desired output style"
                        )
                        with gr.Row():
                            rewrite_example_btn = gr.Button("📋 Load Example", variant="secondary")
                            rewrite_btn = gr.Button("🚀 Rewrite Content", variant="primary")
                    
                    with gr.Column(scale=1):
                        rewrite_output = gr.Textbox(label="Rewritten Content", lines=20, show_copy_button=True)
                        rewrite_raw = gr.Textbox(label="Raw Response", lines=4, visible=False)
            
            rewrite_example_btn.click(
                lambda: EXAMPLE_MATH_CONTENT,
                outputs=[rewrite_input],
            )
            rewrite_btn.click(
                rewrite_synthesis,
                inputs=[rewrite_input, rewrite_style],
                outputs=[rewrite_output, rewrite_raw],
            )
        
        # Knowledge Extraction Tab
        with gr.TabItem("📚 Knowledge Extraction"):
            with gr.Column(elem_classes=["glass-panel"]):
                gr.Markdown("### 🧠 Extract Core Knowledge\nAutomatically identify and extract key definitions, theorems, and properties from unstructured text.")
                with gr.Row():
                    with gr.Column(scale=1):
                        know_input = gr.Textbox(
                            label="Input Mathematical Content",
                            placeholder="Paste your mathematical text here...",
                            lines=12,
                        )
                        with gr.Row():
                            know_example_btn = gr.Button("📋 Load Example", variant="secondary")
                            know_btn = gr.Button("🚀 Extract Knowledge", variant="primary")
                    
                    with gr.Column(scale=1):
                        know_output = gr.Textbox(label="Extracted Knowledge Points", lines=20, show_copy_button=True)
                        know_raw = gr.Textbox(label="Raw Response", lines=4, visible=False)
            
            know_example_btn.click(
                lambda: EXAMPLE_MATH_CONTENT,
                outputs=[know_input],
            )
            know_btn.click(
                knowledge_extraction,
                inputs=[know_input],
                outputs=[know_output, know_raw],
            )
        
        # Textbook Exercise Tab
        with gr.TabItem("📖 Textbook Exercise"):
            with gr.Column(elem_classes=["glass-panel"]):
                gr.Markdown("### 📝 Generate Exercises\nCreate comprehensive textbook-style exercises and problems based on specific knowledge points.")
                with gr.Row():
                    with gr.Column(scale=1):
                        textbook_input = gr.Textbox(
                            label="Input Knowledge Point",
                            placeholder="Enter a specific mathematical concept or theorem...",
                            lines=8,
                        )
                        textbook_diff = gr.Radio(
                            choices=list(TEXTBOOK_EXERCISE_PROMPTS.keys()),
                            value="easy",
                            label="Difficulty",
                            info="Select the problem difficulty"
                        )
                        with gr.Row():
                            textbook_example_btn = gr.Button("📋 Load Example", variant="secondary")
                            textbook_btn = gr.Button("🚀 Generate Exercise", variant="primary")
                    
                    with gr.Column(scale=1):
                        textbook_output = gr.Textbox(label="Generated Exercise Material", lines=20, show_copy_button=True)
                        textbook_raw = gr.Textbox(label="Raw Response", lines=4, visible=False)
            
            textbook_example_btn.click(
                lambda: EXAMPLE_KNOWLEDGE_POINT,
                outputs=[textbook_input],
            )
            textbook_btn.click(
                textbook_exercise,
                inputs=[textbook_input, textbook_diff],
                outputs=[textbook_output, textbook_raw],
            )
    
    gr.HTML("""
    <div class="footer-text">
        <p>🔬 <strong>UltraData-Math L3 Generator</strong> - Part of the UltraData-Math Project</p>
        <p>Powered by OpenBMB & ModelBest • <a href="https://huggingface.co/spaces/openbmb/UltraData-Math-L3-Generator" target="_blank" style="color: #818cf8; text-decoration: none;">View on Hugging Face</a></p>
    </div>
    """)


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