app.py

import gradio as gr
import torch
from gliner import GLiNER
import pandas as pd
import warnings
import random
import re
import time
warnings.filterwarnings('ignore')

# Common NER entity types
STANDARD_ENTITIES = [
    'DATE', 'EVENT', 'FAC', 'GPE', 'LANG', 'LOC',
    'MISC', 'NORP', 'ORG', 'PER', 'PRODUCT', 'Work of Art'
]

# Colour schemes
STANDARD_COLORS = {
    'DATE': '#FF6B6B',      # Red
    'EVENT': '#4ECDC4',     # Teal
    'FAC': '#45B7D1',       # Blue
    'GPE': '#F9CA24',       # Yellow
    'LANG': '#6C5CE7',      # Purple
    'LOC': '#A0E7E5',       # Light Cyan
    'MISC': '#FD79A8',      # Pink
    'NORP': '#8E8E93',      # Grey
    'ORG': '#55A3FF',       # Light Blue
    'PER': '#00B894',       # Green
    'PRODUCT': '#E17055',   # Orange-Red
    'WORK OF ART': '#DDA0DD'  # Plum
}

# Additional colours for custom entities
CUSTOM_COLOR_PALETTE = [
    '#FF9F43', '#10AC84', '#EE5A24', '#0FBC89', '#5F27CD',
    '#FF3838', '#2F3640', '#3742FA', '#2ED573', '#FFA502',
    '#FF6348', '#1E90FF', '#FF1493', '#32CD32', '#FFD700',
    '#FF4500', '#DA70D6', '#00CED1', '#FF69B4', '#7B68EE'
]

class HybridNERManager:
    def __init__(self):
        self.gliner_model = None
        self.spacy_model = None
        self.flair_models = {}
        self.all_entity_colors = {}
        self.model_names = [
            'entities_flair_ner-large',
            'entities_spacy_en_core_web_trf', 
            'entities_flair_ner-ontonotes-large',
            'entities_gliner_knowledgator/modern-gliner-bi-large-v1.0'
        ]

    def load_model(self, model_name):
        """Load the specified model"""
        try:
            if 'spacy' in model_name:
                return self.load_spacy_model()
            elif 'flair' in model_name:
                return self.load_flair_model(model_name)
            elif 'gliner' in model_name:
                return self.load_gliner_model()
        except Exception as e:
            print(f"Error loading {model_name}: {str(e)}")
            return None

    def load_spacy_model(self):
        """Load spaCy model for common NER"""
        if self.spacy_model is None:
            try:
                import spacy
                try:
                    # Try transformer model first, fallback to small model
                    self.spacy_model = spacy.load("en_core_web_trf")
                    print("✓ spaCy transformer model loaded successfully")
                except OSError:
                    try:
                        self.spacy_model = spacy.load("en_core_web_sm")
                        print("✓ spaCy common model loaded successfully") 
                    except OSError:
                        print("spaCy model not found. Using GLiNER for all entity types.")
                        return None
            except Exception as e:
                print(f"Error loading spaCy model: {str(e)}")
                return None
        return self.spacy_model

    def load_flair_model(self, model_name):
        """Load Flair models"""
        if model_name not in self.flair_models:
            try:
                from flair.models import SequenceTagger
                if 'ontonotes' in model_name:
                    model = SequenceTagger.load("flair/ner-english-ontonotes-large")
                    print("✓ Flair OntoNotes model loaded successfully")
                else:
                    model = SequenceTagger.load("flair/ner-english-large") 
                    print("✓ Flair large model loaded successfully")
                self.flair_models[model_name] = model
            except Exception as e:
                print(f"Error loading {model_name}: {str(e)}")
                # Fallback to GLiNER
                return self.load_gliner_model()
        return self.flair_models[model_name]

    def load_gliner_model(self):
        """Load GLiNER model for custom entities"""
        if self.gliner_model is None:
            try:
                # Try the modern GLiNER model first, fallback to stable model
                self.gliner_model = GLiNER.from_pretrained("knowledgator/gliner-bi-large-v1.0")
                print("✓ GLiNER knowledgator model loaded successfully")
            except Exception as e:
                print(f"Primary GLiNER model failed: {str(e)}")
                try:
                    # Fallback to stable model
                    self.gliner_model = GLiNER.from_pretrained("urchade/gliner_medium-v2.1")
                    print("✓ GLiNER fallback model loaded successfully")
                except Exception as e2:
                    print(f"Error loading GLiNER model: {str(e2)}")
                    return None
        return self.gliner_model

    def assign_colours(self, standard_entities, custom_entities):
        """Assign colours to all entity types"""
        self.all_entity_colors = {}

        # Assign common colours
        for entity in standard_entities:
            # Handle the special case of "Work of Art"
            colour_key = "WORK OF ART" if entity == "Work of Art" else entity.upper()
            self.all_entity_colors[entity.upper()] = STANDARD_COLORS.get(colour_key, '#CCCCCC')

        # Assign custom colours
        for i, entity in enumerate(custom_entities):
            if i < len(CUSTOM_COLOR_PALETTE):
                self.all_entity_colors[entity.upper()] = CUSTOM_COLOR_PALETTE[i]
            else:
                # Generate random colour if we run out
                self.all_entity_colors[entity.upper()] = f"#{random.randint(0, 0xFFFFFF):06x}"

        return self.all_entity_colors

    def extract_entities_by_model(self, text, entity_types, model_name, threshold=0.3):
        """Extract entities using the specified model"""
        if 'spacy' in model_name:
            return self.extract_spacy_entities(text, entity_types)
        elif 'flair' in model_name:
            return self.extract_flair_entities(text, entity_types, model_name)
        elif 'gliner' in model_name:
            return self.extract_gliner_entities(text, entity_types, threshold, is_custom=False)
        else:
            return []

    def extract_spacy_entities(self, text, entity_types):
        """Extract entities using spaCy"""
        model = self.load_spacy_model()
        if model is None:
            return []

        try:
            doc = model(text)
            entities = []
            for ent in doc.ents:
                if ent.label_ in entity_types:
                    entities.append({
                        'text': ent.text,
                        'label': ent.label_,
                        'start': ent.start_char,
                        'end': ent.end_char,
                        'confidence': 1.0,  # spaCy doesn't provide confidence scores
                        'source': 'spaCy'
                    })
            return entities
        except Exception as e:
            print(f"Error with spaCy extraction: {str(e)}")
            return []

    def extract_flair_entities(self, text, entity_types, model_name):
        """Extract entities using Flair"""
        model = self.load_flair_model(model_name)
        if model is None:
            return []

        try:
            from flair.data import Sentence
            sentence = Sentence(text)
            model.predict(sentence)
            entities = []
            for entity in sentence.get_spans('ner'):
                # Map Flair labels to our common set
                label = entity.tag
                if label == 'PERSON':
                    label = 'PER'
                elif label == 'ORGANIZATION':
                    label = 'ORG'
                elif label == 'LOCATION':
                    label = 'LOC'
                elif label == 'MISCELLANEOUS':
                    label = 'MISC'

                if label in entity_types:
                    entities.append({
                        'text': entity.text,
                        'label': label,
                        'start': entity.start_position,
                        'end': entity.end_position,
                        'confidence': entity.score,
                        'source': f'Flair-{model_name.split("-")[-1]}'
                    })
            return entities
        except Exception as e:
            print(f"Error with Flair extraction: {str(e)}")
            return []

    def extract_gliner_entities(self, text, entity_types, threshold=0.3, is_custom=True):
        """Extract entities using GLiNER"""
        model = self.load_gliner_model()
        if model is None:
            return []

        try:
            entities = model.predict_entities(text, entity_types, threshold=threshold)
            result = []
            for entity in entities:
                result.append({
                    'text': entity['text'],
                    'label': entity['label'].upper(),
                    'start': entity['start'],
                    'end': entity['end'],
                    'confidence': entity.get('score', 0.0),
                    'source': 'GLiNER-Custom' if is_custom else 'GLiNER-Common'
                })
            return result
        except Exception as e:
            print(f"Error with GLiNER extraction: {str(e)}")
            return []

def find_overlapping_entities(entities):
    """Find and share overlapping entities - specifically entities found by BOTH common NER models AND custom entities"""
    if not entities:
        return []

    # Sort entities by start position
    sorted_entities = sorted(entities, key=lambda x: x['start'])
    shared_entities = []

    i = 0
    while i < len(sorted_entities):
        current_entity = sorted_entities[i]
        overlapping_entities = [current_entity]

        # Find all entities that overlap with current entity
        j = i + 1
        while j < len(sorted_entities):
            next_entity = sorted_entities[j]

            # Check if entities overlap (same text span or overlapping positions)
            if (current_entity['start'] <= next_entity['start'] < current_entity['end'] or
                next_entity['start'] <= current_entity['start'] < current_entity['end'] or
                current_entity['text'].lower() == next_entity['text'].lower()):
                overlapping_entities.append(next_entity)
                sorted_entities.pop(j)
            else:
                j += 1

        # Create shared entity only if we have BOTH common and custom entities
        if len(overlapping_entities) == 1:
            shared_entities.append(overlapping_entities[0])
        else:
            # Check if this is a true "shared" entity (common + custom)
            has_common = False
            has_custom = False
            
            for entity in overlapping_entities:
                source = entity.get('source', '')
                if source in ['spaCy', 'GLiNER-Common'] or source.startswith('Flair-'):
                    has_common = True
                elif source == 'GLiNER-Custom':
                    has_custom = True
            
            if has_common and has_custom:
                # This is a true shared entity (common + custom)
                shared_entity = share_entities(overlapping_entities)
                shared_entities.append(shared_entity)
            else:
                # These are just overlapping entities from the same source type, keep separate
                shared_entities.extend(overlapping_entities)

        i += 1

    return shared_entities

def share_entities(entity_list):
    """Share multiple overlapping entities into one"""
    if len(entity_list) == 1:
        return entity_list[0]

    # Use the entity with the longest text span as the base
    base_entity = max(entity_list, key=lambda x: len(x['text']))

    # Collect all labels and sources
    labels = [entity['label'] for entity in entity_list]
    sources = [entity['source'] for entity in entity_list]
    confidences = [entity['confidence'] for entity in entity_list]

    return {
        'text': base_entity['text'],
        'start': base_entity['start'],
        'end': base_entity['end'],
        'labels': labels,
        'sources': sources,
        'confidences': confidences,
        'is_shared': True,
        'entity_count': len(entity_list)
    }

def create_highlighted_html(text, entities, entity_colors):
    """Create HTML with highlighted entities"""
    if not entities:
        return f"<div style='padding: 15px; border: 1px solid #ddd; border-radius: 5px; background-color: #fafafa;'><p>{text}</p></div>"

    # Find and share overlapping entities
    shared_entities = find_overlapping_entities(entities)

    # Sort by start position
    sorted_entities = sorted(shared_entities, key=lambda x: x['start'])

    # Create HTML with highlighting
    html_parts = []
    last_end = 0

    for entity in sorted_entities:
        # Add text before entity
        html_parts.append(text[last_end:entity['start']])

        if entity.get('is_shared', False):
            # Handle shared entity with multiple colours
            html_parts.append(create_shared_entity_html(entity, entity_colors))
        else:
            # Handle single entity
            html_parts.append(create_single_entity_html(entity, entity_colors))

        last_end = entity['end']

    # Add remaining text
    html_parts.append(text[last_end:])

    highlighted_text = ''.join(html_parts)

    return f"""
    <div style='padding: 15px; border: 2px solid #ddd; border-radius: 8px; background-color: #fafafa; margin: 10px 0;'>
        <h4 style='margin: 0 0 15px 0; color: #333;'>📝 Text with Highlighted Entities</h4>
        <div style='line-height: 1.8; font-size: 16px; background-color: white; padding: 15px; border-radius: 5px;'>{highlighted_text}</div>
    </div>
    """

def create_single_entity_html(entity, entity_colors):
    """Create HTML for a single entity"""
    label = entity['label']
    colour = entity_colors.get(label.upper(), '#CCCCCC')
    confidence = entity.get('confidence', 0.0)
    source = entity.get('source', 'Unknown')

    return (f'<span style="background-color: {colour}; padding: 2px 4px; '
            f'border-radius: 3px; margin: 0 1px; '
            f'border: 1px solid {colour}; color: white; font-weight: bold;" '
            f'title="{label} ({source}) - confidence: {confidence:.2f}">'
            f'{entity["text"]}</span>')

def create_shared_entity_html(entity, entity_colors):
    """Create HTML for a shared entity with multiple colours"""
    labels = entity['labels']
    sources = entity['sources']
    confidences = entity['confidences']

    # Get colours for each label
    colours = []
    for label in labels:
        colour = entity_colors.get(label.upper(), '#CCCCCC')
        colours.append(colour)

    # Create gradient background
    if len(colours) == 2:
        gradient = f"linear-gradient(to right, {colours[0]} 50%, {colours[1]} 50%)"
    else:
        # For more colours, create equal segments
        segment_size = 100 / len(colours)
        gradient_parts = []
        for i, colour in enumerate(colours):
            start = i * segment_size
            end = (i + 1) * segment_size
            gradient_parts.append(f"{colour} {start}%, {colour} {end}%")
        gradient = f"linear-gradient(to right, {', '.join(gradient_parts)})"

    # Create tooltip
    tooltip_parts = []
    for i, label in enumerate(labels):
        tooltip_parts.append(f"{label} ({sources[i]}) - {confidences[i]:.2f}")
    tooltip = " | ".join(tooltip_parts)

    return (f'<span style="background: {gradient}; padding: 2px 4px; '
            f'border-radius: 3px; margin: 0 1px; '
            f'border: 2px solid #333; color: white; font-weight: bold;" '
            f'title="SHARED: {tooltip}">'
            f'{entity["text"]} 🤝</span>')

def create_entity_table_html(entities_of_type, entity_type, colour, is_shared=False):
    """Create HTML table for a specific entity type"""
    if is_shared:
        header = f"🤝 Shared Entities ({len(entities_of_type)} found)"
        
        table_html = f"""
        <div style="margin: 15px 0;">
            <h4 style="color: {colour}; margin-bottom: 15px;">{header}</h4>
            <table style="width: 100%; border-collapse: collapse; border: 1px solid #ddd;">
                <thead>
                    <tr style="background-color: {colour}; color: white;">
                        <th style="padding: 12px; text-align: left; border: 1px solid #ddd;">Entity Text</th>
                        <th style="padding: 12px; text-align: left; border: 1px solid #ddd;">All Labels</th>
                        <th style="padding: 12px; text-align: left; border: 1px solid #ddd;">Sources</th>
                        <th style="padding: 12px; text-align: left; border: 1px solid #ddd;">Count</th>
                    </tr>
                </thead>
                <tbody>
        """
        
        for entity in entities_of_type:
            labels_text = " | ".join(entity['labels'])
            sources_text = " | ".join(entity['sources'])
            
            table_html += f"""
                    <tr style="background-color: #fff;">
                        <td style="padding: 10px; border: 1px solid #ddd; font-weight: bold;">{entity['text']}</td>
                        <td style="padding: 10px; border: 1px solid #ddd;">{labels_text}</td>
                        <td style="padding: 10px; border: 1px solid #ddd;">{sources_text}</td>
                        <td style="padding: 10px; border: 1px solid #ddd; text-align: center;">
                            <span style='background-color: #28a745; color: white; padding: 2px 6px; border-radius: 10px; font-size: 11px;'>
                                {entity['entity_count']}
                            </span>
                        </td>
                    </tr>
            """
    else:
        # Determine if it's common or custom
        is_standard = entity_type in STANDARD_ENTITIES
        icon = "🎯" if is_standard else "✨"
        source_text = "Common NER" if is_standard else "Custom GLiNER"
        header = f"{icon} {source_text} - {entity_type} ({len(entities_of_type)} found)"
        
        table_html = f"""
        <div style="margin: 15px 0;">
            <h4 style="color: {colour}; margin-bottom: 15px;">{header}</h4>
            <table style="width: 100%; border-collapse: collapse; border: 1px solid #ddd;">
                <thead>
                    <tr style="background-color: {colour}; color: white;">
                        <th style="padding: 12px; text-align: left; border: 1px solid #ddd;">Entity Text</th>
                        <th style="padding: 12px; text-align: left; border: 1px solid #ddd;">Confidence</th>
                        <th style="padding: 12px; text-align: left; border: 1px solid #ddd;">Type</th>
                        <th style="padding: 12px; text-align: left; border: 1px solid #ddd;">Source</th>
                    </tr>
                </thead>
                <tbody>
        """
        
        # Sort by confidence score
        entities_of_type.sort(key=lambda x: x.get('confidence', 0), reverse=True)
        
        for entity in entities_of_type:
            confidence = entity.get('confidence', 0.0)
            confidence_colour = "#28a745" if confidence > 0.7 else "#ffc107" if confidence > 0.4 else "#dc3545"
            source = entity.get('source', 'Unknown')
            source_badge = f"<span style='background-color: #007bff; color: white; padding: 2px 6px; border-radius: 10px; font-size: 11px;'>{source}</span>"
            
            table_html += f"""
                    <tr style="background-color: #fff;">
                        <td style="padding: 10px; border: 1px solid #ddd; font-weight: bold;">{entity['text']}</td>
                        <td style="padding: 10px; border: 1px solid #ddd;">
                            <span style="color: {confidence_colour}; font-weight: bold;">
                                {confidence:.3f}
                            </span>
                        </td>
                        <td style="padding: 10px; border: 1px solid #ddd;">{entity['label']}</td>
                        <td style="padding: 10px; border: 1px solid #ddd;">{source_badge}</td>
                    </tr>
            """
    
    table_html += "</tbody></table></div>"
    return table_html

def create_legend_html(entity_colors, standard_entities, custom_entities):
    """Create a legend showing entity colours"""
    if not entity_colors:
        return ""

    html = "<div style='margin: 15px 0; padding: 15px; background-color: #f8f9fa; border-radius: 8px;'>"
    html += "<h4 style='margin: 0 0 15px 0;'>🎨 Entity Type Legend</h4>"

    if standard_entities:
        html += "<div style='margin-bottom: 15px;'>"
        html += "<h5 style='margin: 0 0 8px 0;'>🎯 Common Entities:</h5>"
        html += "<div style='display: flex; flex-wrap: wrap; gap: 8px;'>"
        for entity_type in standard_entities:
            colour = entity_colors.get(entity_type.upper(), '#ccc')
            html += f"<span style='background-color: {colour}; padding: 4px 8px; border-radius: 15px; color: white; font-weight: bold; font-size: 12px;'>{entity_type}</span>"
        html += "</div></div>"

    if custom_entities:
        html += "<div>"
        html += "<h5 style='margin: 0 0 8px 0;'>✨ Custom Entities:</h5>"
        html += "<div style='display: flex; flex-wrap: wrap; gap: 8px;'>"
        for entity_type in custom_entities:
            colour = entity_colors.get(entity_type.upper(), '#ccc')
            html += f"<span style='background-color: {colour}; padding: 4px 8px; border-radius: 15px; color: white; font-weight: bold; font-size: 12px;'>{entity_type}</span>"
        html += "</div></div>"

    html += "</div>"
    return html

# Initialize the NER manager
ner_manager = HybridNERManager()

def process_text(text, standard_entities, custom_entities_str, confidence_threshold, selected_model, progress=gr.Progress()):
    """Main processing function for Gradio interface with progress tracking"""
    if not text.strip():
        return "❌ Please enter some text to analyse", "", None, None, None, None, None, None, None, None

    progress(0.1, desc="Initialising...")
    
    # Parse custom entities
    custom_entities = []
    if custom_entities_str.strip():
        custom_entities = [entity.strip() for entity in custom_entities_str.split(',') if entity.strip()]

    # Parse common entities
    selected_standard = [entity for entity in standard_entities if entity]

    if not selected_standard and not custom_entities:
        return "❌ Please select at least one common entity type OR enter custom entity types", "", None, None, None, None, None, None, None, None

    progress(0.2, desc="Loading models...")
    
    all_entities = []

    # Extract common entities using selected model
    if selected_standard and selected_model:
        progress(0.4, desc="Extracting common entities...")
        standard_entities_results = ner_manager.extract_entities_by_model(text, selected_standard, selected_model, confidence_threshold)
        all_entities.extend(standard_entities_results)

    # Extract custom entities using GLiNER
    if custom_entities:
        progress(0.6, desc="Extracting custom entities...")
        custom_entity_results = ner_manager.extract_gliner_entities(text, custom_entities, confidence_threshold, is_custom=True)
        all_entities.extend(custom_entity_results)

    if not all_entities:
        return "❌ No entities found. Try lowering the confidence threshold or using different entity types.", "", None, None, None, None, None, None, None, None

    progress(0.8, desc="Processing results...")
    
    # Assign colours
    entity_colors = ner_manager.assign_colours(selected_standard, custom_entities)

    # Create outputs
    legend_html = create_legend_html(entity_colors, selected_standard, custom_entities)
    highlighted_html = create_highlighted_html(text, all_entities, entity_colors)
    
    # Share overlapping entities
    shared_entities = find_overlapping_entities(all_entities)
    
    # Group entities by type
    entity_groups = {}
    for entity in shared_entities:
        if entity.get('is_shared', False):
            key = 'SHARED_ENTITIES'
        else:
            key = entity['label']
        
        if key not in entity_groups:
            entity_groups[key] = []
        entity_groups[key].append(entity)

    progress(0.9, desc="Creating summary...")
    
    # Create summary
    total_entities = len(all_entities)
    final_count = len(shared_entities)
    shared_count = sum(1 for e in shared_entities if e.get('is_shared', False))

    summary = f"""
    ## 📊 Analysis Summary
    - **Total entities found:** {total_entities}
    - **Final entities displayed:** {final_count}
    - **Shared entities:** {shared_count}
    - **Average confidence:** {sum(e.get('confidence', 0) for e in all_entities) / total_entities:.3f}
    """

    progress(1.0, desc="Complete!")
    
    # Create HTML tables for each entity type
    tables = {}
    
    # Create table for shared entities if any
    if 'SHARED_ENTITIES' in entity_groups:
        tables['shared'] = create_entity_table_html(entity_groups['SHARED_ENTITIES'], 'SHARED_ENTITIES', '#666666', is_shared=True)
    else:
        tables['shared'] = None
    
    # Create tables for other entity types (up to 7 for the interface)
    entity_types = [k for k in entity_groups.keys() if k != 'SHARED_ENTITIES']
    for i in range(7):
        if i < len(entity_types):
            entity_type = entity_types[i]
            colour = entity_colors.get(entity_type.upper(), '#f0f0f0')
            tables[f'tab{i+1}'] = create_entity_table_html(entity_groups[entity_type], entity_type, colour)
        else:
            tables[f'tab{i+1}'] = None
    
    return (summary, legend_html + highlighted_html, 
            tables.get('shared'), tables.get('tab1'), tables.get('tab2'), 
            tables.get('tab3'), tables.get('tab4'), tables.get('tab5'), 
            tables.get('tab6'), tables.get('tab7'))

# Create Gradio interface
def create_interface():
    with gr.Blocks(title="Hybrid NER + GLiNER Tool", theme=gr.themes.Soft()) as demo:
        gr.Markdown("""
        # 🎯 Hybrid NER + Custom GLiNER Entity Recognition Tool
        
        Combine common NER categories with your own custom entity types! This tool uses both traditional NER models and GLiNER for comprehensive entity extraction.
        
        ## 🤝 NEW: Overlapping entities are automatically shared with split-colour highlighting!
        
        ### How to use:
        1. **📝 Enter your text** in the text area below
        2. **🎯 Select a model** from the dropdown for common entities
        3. **☑️ Select common entities** you want to find (PER, ORG, LOC, etc.)
        4. **✨ Add custom entities** (comma-separated) like "relationships, occupations, skills"
        5. **⚙️ Adjust confidence threshold** 
        6. **🔍 Click "Analyse Text"** to see results with tabbed output
        """)
        
        with gr.Row():
            with gr.Column(scale=2):
                text_input = gr.Textbox(
                    label="📝 Text to Analyse",
                    placeholder="Enter your text here...",
                    lines=6,
                    max_lines=10
                )
            
            with gr.Column(scale=1):
                confidence_threshold = gr.Slider(
                    minimum=0.1,
                    maximum=0.9,
                    value=0.3,
                    step=0.1,
                    label="🎚️ Confidence Threshold"
                )
        
        with gr.Row():
            with gr.Column():
                gr.Markdown("### 🎯 Common Entity Types")
                
                # Model selector
                model_dropdown = gr.Dropdown(
                    choices=ner_manager.model_names,
                    value=ner_manager.model_names[0],
                    label="Select Model for Common Entities",
                    info="Choose which model to use for common NER"
                )
                
                # Common entities with select all functionality
                standard_entities = gr.CheckboxGroup(
                    choices=STANDARD_ENTITIES,
                    value=['PER', 'ORG', 'LOC', 'MISC'],  # Default selection
                    label="Select Common Entities"
                )
                
                # Select/Deselect All button
                with gr.Row():
                    select_all_btn = gr.Button("🔘 Deselect All", size="sm")
                    
                # Function for select/deselect all
                def toggle_all_entities(current_selection):
                    if len(current_selection) > 0:
                        # If any are selected, deselect all
                        return [], "☑️ Select All"
                    else:
                        # If none selected, select all
                        return STANDARD_ENTITIES, "🔘 Deselect All"
                
                select_all_btn.click(
                    fn=toggle_all_entities,
                    inputs=[standard_entities],
                    outputs=[standard_entities, select_all_btn]
                )
            
            with gr.Column():
                gr.Markdown("### ✨ Custom Entity Types")
                custom_entities = gr.Textbox(
                    label="Custom Entities (comma-separated)",
                    placeholder="e.g. relationships, occupations, skills, emotions",
                    lines=3
                )
                gr.Markdown("""
                **Examples:**
                - relationships, occupations, skills
                - emotions, actions, objects  
                - medical conditions, treatments
                - financial terms, business roles
                """)
        
        analyse_btn = gr.Button("🔍 Analyse Text", variant="primary", size="lg")
        
        # Output sections
        with gr.Row():
            summary_output = gr.Markdown(label="Summary")
        
        with gr.Row():
            highlighted_output = gr.HTML(label="Highlighted Text")
        
        # Results section with native Gradio tabs
        with gr.Row():
            with gr.Column():
                gr.Markdown("### 📋 Detailed Results")
                
                with gr.Tabs() as results_tabs:
                    # Pre-create tabs for different entity types
                    with gr.Tab("🤝 Shared", visible=False) as shared_tab:
                        shared_output = gr.HTML()
                    
                    with gr.Tab("Entity Type 1", visible=False) as tab1:
                        tab1_output = gr.HTML()
                    
                    with gr.Tab("Entity Type 2", visible=False) as tab2:
                        tab2_output = gr.HTML()
                    
                    with gr.Tab("Entity Type 3", visible=False) as tab3:
                        tab3_output = gr.HTML()
                    
                    with gr.Tab("Entity Type 4", visible=False) as tab4:
                        tab4_output = gr.HTML()
                    
                    with gr.Tab("Entity Type 5", visible=False) as tab5:
                        tab5_output = gr.HTML()
                    
                    with gr.Tab("Entity Type 6", visible=False) as tab6:
                        tab6_output = gr.HTML()
                    
                    with gr.Tab("Entity Type 7", visible=False) as tab7:
                        tab7_output = gr.HTML()
        
        # Connect the button to the processing function
        analyse_btn.click(
            fn=process_text,
            inputs=[
                text_input,
                standard_entities,
                custom_entities,
                confidence_threshold,
                model_dropdown
            ],
            outputs=[
                summary_output, 
                highlighted_output, 
                shared_output,
                tab1_output,
                tab2_output,
                tab3_output,
                tab4_output,
                tab5_output,
                tab6_output,
                tab7_output
            ]
        )
        
        # Add examples
        gr.Examples(
            examples=[
                [
                    "John Smith works at Google in New York. He graduated from Stanford University in 2015 and specialises in artificial intelligence research. His wife Sarah is a doctor at Mount Sinai Hospital.",
                    ["PER", "ORG", "LOC", "DATE"],
                    "relationships, occupations, educational background",
                    0.3,
                    "entities_spacy_en_core_web_trf"
                ],
                [
                    "The meeting between CEO Jane Doe and the board of directors at Microsoft headquarters in Seattle discussed the Q4 financial results and the new AI strategy for 2024.",
                    ["PER", "ORG", "LOC", "DATE"],
                    "corporate roles, business events, financial terms",
                    0.4,
                    "entities_flair_ner-ontonotes-large"
                ],
                [
                    "Dr. Emily Watson published a research paper on machine learning algorithms at MIT. She collaborates with her colleague Prof. David Chen on natural language processing projects.",
                    ["PER", "ORG", "Work of Art"],
                    "academic titles, research topics, collaborations",
                    0.3,
                    "entities_gliner_knowledgator/modern-gliner-bi-large-v1.0"
                ]
            ],
            inputs=[
                text_input,
                standard_entities,
                custom_entities,
                confidence_threshold,
                model_dropdown
            ]
        )

    return demo

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