app.py

#!/usr/bin/env python3

import os
import gradio as gr
from datasets import load_dataset
from dataclasses import dataclass
from typing import Any

# HF token should be set as environment variable HF_TOKEN

@dataclass
class DatasetViewer:
    dataset: Any = None
    current_index: int = 0
    total_items: int = 0
    show_answer: bool = False
    
    def load_dataset(self, dataset_name: str = "yourbench/reachy_mini_info_benchmark", 
                    subset: str = "prepared_lighteval"):
        try:
            self.dataset = load_dataset(dataset_name, subset)
            
            # Try to get the first available split
            if isinstance(self.dataset, dict):
                first_split = next(iter(self.dataset.keys()))
                self.dataset = self.dataset[first_split]
            
            self.total_items = len(self.dataset)
            self.current_index = 0
            self.show_answer = False
            return True
        except Exception as e:
            print(f"Error loading dataset: {e}")
            # Create mock data for demonstration
            self.create_mock_data()
            return False
    
    def create_mock_data(self):
        """Create mock data that simulates yourbench dataset structure"""
        mock_data = [
            {
                "question": "What is the capital of France?",
                "ground_truth_answer": "Paris",
                "document_summary": "France is a country in Western Europe.",
                "question_category": "geography",
                "estimated_difficulty": 2,
                "kind": "single_shot",
                "additional_instructions": "Ask generalizable questions."
            },
            {
                "question": "Explain the concept of object-oriented programming.",
                "ground_truth_answer": "Object-oriented programming (OOP) is a programming paradigm based on the concept of 'objects', which can contain data and code: data in the form of fields, and code in the form of procedures. Key principles include encapsulation, inheritance, and polymorphism.",
                "document_summary": "Programming concepts",
                "question_category": "computer_science",
                "estimated_difficulty": 4,
                "kind": "single_shot",
                "additional_instructions": "Provide detailed explanations."
            },
        ]
        
        # Create a mock dataset-like structure
        class MockDataset:
            def __init__(self, data):
                self.data = data
            
            def __len__(self):
                return len(self.data)
            
            def __getitem__(self, idx):
                return self.data[idx]
        
        self.dataset = MockDataset(mock_data)
        self.total_items = len(mock_data)
        self.current_index = 0
        self.show_answer = False

viewer = DatasetViewer()
viewer.load_dataset()

def get_current_item():
    if viewer.dataset is None or viewer.total_items == 0:
        return {"question": "No data available", "answer": "", "context": "", "task": "", "difficulty": ""}
    
    item = viewer.dataset[viewer.current_index]
    
    # Map actual dataset fields to expected interface fields
    mapped_item = {
        "question": item.get("question", "No question available"),
        "answer": item.get("ground_truth_answer", "No answer available"),
        "context": item.get("document_summary", "No context available"),
        "task": item.get("question_category", "Unknown category"),
        "difficulty": str(item.get("estimated_difficulty", "Unknown")),
        "additional_instructions": item.get("additional_instructions", ""),
        "kind": item.get("kind", ""),
        "citations": item.get("citations", []),
        "document_id": item.get("document_id", ""),
    }
    
    return mapped_item

def get_card_content():
    item = get_current_item()
    
    if viewer.show_answer:
        # Show answer side
        content = f"""
        <div style="text-align: center; padding: 40px; min-height: 400px; display: flex; flex-direction: column; justify-content: center;">
            <h1 style="color: #2d5aa0; font-size: 2.5em; margin-bottom: 30px;">Answer</h1>
            <div style="font-size: 1.4em; line-height: 1.6; color: #333; max-width: 800px; margin: 0 auto;">
                {item['answer']}
            </div>
        </div>
        """
        button_text = "Show Question"
    else:
        # Show question side
        content = f"""
        <div style="text-align: center; padding: 40px; min-height: 400px; display: flex; flex-direction: column; justify-content: center;">
            <h1 style="color: #2d5aa0; font-size: 2.5em; margin-bottom: 30px;">Question</h1>
            <div style="font-size: 1.4em; line-height: 1.6; color: #333; max-width: 800px; margin: 0 auto;">
                {item['question']}
            </div>
        </div>
        """
        button_text = "Show Answer"
    
    return content, button_text

def get_metadata():
    item = get_current_item()
    
    context = item.get('context', 'No context')
    task = item.get('task', 'Unknown task')
    difficulty = item.get('difficulty', 'Unknown')
    kind = item.get('kind', 'Unknown')
    additional_instructions = item.get('additional_instructions', '')
    
    metadata = f"Category: {task} | Difficulty: {difficulty}/5 | Type: {kind}"
    if additional_instructions:
        metadata += f" | Instructions: {additional_instructions}"
    
    # Truncate context if too long
    if len(context) > 150:
        context = context[:150] + "..."
    
    return metadata, context

def toggle_answer():
    viewer.show_answer = not viewer.show_answer
    content, button_text = get_card_content()
    return content, button_text

def next_item():
    if viewer.total_items > 0:
        viewer.current_index = (viewer.current_index + 1) % viewer.total_items
        viewer.show_answer = False
    
    content, button_text = get_card_content()
    metadata, context = get_metadata()
    progress = f"{viewer.current_index + 1} of {viewer.total_items}"
    
    return content, button_text, metadata, context, progress

def previous_item():
    if viewer.total_items > 0:
        viewer.current_index = (viewer.current_index - 1) % viewer.total_items
        viewer.show_answer = False
    
    content, button_text = get_card_content()
    metadata, context = get_metadata()
    progress = f"{viewer.current_index + 1} of {viewer.total_items}"
    
    return content, button_text, metadata, context, progress

def jump_to_item(item_number):
    if viewer.total_items > 0 and 1 <= item_number <= viewer.total_items:
        viewer.current_index = item_number - 1
        viewer.show_answer = False
    
    content, button_text = get_card_content()
    metadata, context = get_metadata()
    progress = f"{viewer.current_index + 1} of {viewer.total_items}"
    
    return content, button_text, metadata, context, progress

# Initialize
initial_content, initial_button_text = get_card_content()
initial_metadata, initial_context = get_metadata()
initial_progress = f"{viewer.current_index + 1} of {viewer.total_items}"

# Custom CSS for flashcard styling
css = """
.flashcard {
    border: 2px solid #ddd;
    border-radius: 15px;
    box-shadow: 0 4px 8px rgba(0,0,0,0.1);
    background: white;
    margin: 20px 0;
    min-height: 500px;
}

.controls {
    text-align: center;
    padding: 20px;
}

.metadata {
    background: #f8f9fa;
    padding: 15px;
    border-radius: 8px;
    margin: 10px 0;
    font-size: 0.9em;
    color: #666;
}

.context {
    background: #fff3cd;
    padding: 15px;
    border-radius: 8px;
    margin: 10px 0;
    font-size: 0.9em;
    color: #856404;
    border: 1px solid #ffeaa7;
}

.progress {
    text-align: center;
    font-size: 1.2em;
    font-weight: bold;
    color: #2d5aa0;
    margin: 20px 0;
}
"""

# Create Gradio interface
with gr.Blocks(css=css, title="Reachy Trivia") as demo:
    gr.Markdown("# 🤖 Reachy Trivia")
    gr.Markdown("""
    **Test your knowledge about Reachy Mini and Reachy 2!**
    
    Reachy Mini is the $299 open-source desktop robot designed for AI developers, educators, and hobbyists. 
    With its expressive antennas, screen-based eyes, and full Python programmability, it's the perfect companion 
    for exploring human-robot interaction and AI development.
    
    *Questions generated by Yourbench - HuggingFace's Dynamic Evaluation Generation Platform*
    """)
    
    # Main flashcard area
    with gr.Row():
        with gr.Column(scale=1):
            flashcard = gr.HTML(initial_content, elem_classes=["flashcard"])
            
            # Controls
            with gr.Row(elem_classes=["controls"]):
                prev_btn = gr.Button("Previous", size="lg", variant="secondary")
                toggle_btn = gr.Button(initial_button_text, size="lg", variant="primary")
                next_btn = gr.Button("Next", size="lg", variant="secondary")
            
            # Progress
            progress_display = gr.HTML(f'<div class="progress">{initial_progress}</div>')
    
    # Metadata and navigation sidebar
    with gr.Row():
        with gr.Column(scale=2):
            metadata_display = gr.HTML(f'<div class="metadata">{initial_metadata}</div>')
            context_display = gr.HTML(f'<div class="context">Context: {initial_context}</div>')
        
        with gr.Column(scale=1):
            gr.Markdown("### Jump to Item")
            jump_input = gr.Number(
                label="Item Number",
                value=1,
                minimum=1,
                maximum=viewer.total_items,
                precision=0
            )
            jump_btn = gr.Button("Jump", variant="secondary")
            
            gr.Markdown("---")
            gr.Markdown("### About Reachy Mini")
            gr.Markdown("""
            **Key Features:**
            - 🎯 $299 open-source desktop robot
            - 📱 Expressive screen-based eyes
            - 🔧 Full Python programmability  
            - 🤝 Human-robot interaction focus
            - 🧠 Hugging Face AI integration
            - 📦 DIY assembly kit
            """)
            
            gr.Markdown("*Questions powered by [YouBench](https://huggingface.co/yourbench)*")
    
    # Event handlers
    toggle_btn.click(
        toggle_answer,
        outputs=[flashcard, toggle_btn]
    )
    
    next_btn.click(
        next_item,
        outputs=[flashcard, toggle_btn, metadata_display, context_display, progress_display]
    )
    
    prev_btn.click(
        previous_item,
        outputs=[flashcard, toggle_btn, metadata_display, context_display, progress_display]
    )
    
    jump_btn.click(
        jump_to_item,
        inputs=[jump_input],
        outputs=[flashcard, toggle_btn, metadata_display, context_display, progress_display]
    )

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