# import spaces
import gradio as gr
from phi3_instruct_graph import Phi3InstructGraph
import rapidjson
from pyvis.network import Network
import networkx as nx
import spacy
from spacy import displacy
from spacy.tokens import Span
import random
import os
import pickle

# Constants
TITLE = "🌐 GraphMind: Phi-3 Instruct Graph Explorer"
SUBTITLE = "✨ Extract and visualize knowledge graphs from any text in multiple languages"

# Basic CSS for styling
CUSTOM_CSS = """
.gradio-container {
    font-family: 'Segoe UI', Roboto, sans-serif;
}
"""

# Cache directory and file paths
CACHE_DIR = "cache"
EXAMPLE_CACHE_FILE = os.path.join(CACHE_DIR, "first_example_cache.pkl")

# Create cache directory if it doesn't exist
os.makedirs(CACHE_DIR, exist_ok=True)

# Color utilities
def get_random_light_color():
    r = random.randint(140, 255)
    g = random.randint(140, 255)
    b = random.randint(140, 255)
    return f"#{r:02x}{g:02x}{b:02x}"

# Text preprocessing
def handle_text(text):
    return " ".join(text.split())

# Main processing functions
# @spaces.GPU
def extract(text):
    try:
        model = Phi3InstructGraph()    
        result = model.extract(text)
        return rapidjson.loads(result)
    except Exception as e:
        raise gr.Error(f"Extraction error: {str(e)}")

def find_token_indices(doc, substring, text):
    result = []
    start_index = text.find(substring)
    
    while start_index != -1:
        end_index = start_index + len(substring)
        start_token = None
        end_token = None

        for token in doc:
            if token.idx == start_index:
                start_token = token.i
            if token.idx + len(token) == end_index:
                end_token = token.i + 1

        if start_token is not None and end_token is not None:
            result.append({
                "start": start_token,
                "end": end_token
            })
        
        # Search for next occurrence
        start_index = text.find(substring, end_index)

    return result

def create_custom_entity_viz(data, full_text):
    nlp = spacy.blank("xx")
    doc = nlp(full_text)

    spans = []
    colors = {}
    for node in data["nodes"]:
        entity_spans = find_token_indices(doc, node["id"], full_text)
        for dataentity in entity_spans:
            start = dataentity["start"]
            end = dataentity["end"]
            
            if start < len(doc) and end <= len(doc):
                # Check for overlapping spans
                overlapping = any(s.start < end and start < s.end for s in spans)
                if not overlapping:                
                    node_type = node.get("type", "Entity")
                    span = Span(doc, start, end, label=node_type)
                    spans.append(span)
                    if node_type not in colors:
                        colors[node_type] = get_random_light_color()

    doc.set_ents(spans, default="unmodified")
    doc.spans["sc"] = spans

    options = {
        "colors": colors,
        "ents": list(colors.keys()),
        "style": "ent",
        "manual": True
    }

    html = displacy.render(doc, style="span", options=options)
    # Add custom styling to the entity visualization
    styled_html = f"""
    <div style="padding: 20px; border-radius: 12px; background-color: white; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);">
        {html}
    </div>
    """
    return styled_html

def create_graph(json_data):
    G = nx.Graph()

    # Add nodes with tooltips - with error handling for missing keys
    for node in json_data['nodes']:
        # Get node type with fallback
        node_type = node.get("type", "Entity")
        # Get detailed type with fallback
        detailed_type = node.get("detailed_type", node_type)
        
        # Use node ID and type info for the tooltip
        G.add_node(node['id'], title=f"{node_type}: {detailed_type}")

    # Add edges with labels
    for edge in json_data['edges']:
        # Check if the required keys exist
        if 'from' in edge and 'to' in edge:
            label = edge.get('label', 'related')
            G.add_edge(edge['from'], edge['to'], title=label, label=label)

    # Create network visualization
    nt = Network(
        width="100%",
        height="700px",
        directed=True,
        notebook=False,
        bgcolor="#f8fafc", 
        font_color="#1e293b"
    )
    
    # Configure network display
    nt.from_nx(G)
    nt.barnes_hut(
        gravity=-3000,
        central_gravity=0.3,
        spring_length=50,
        spring_strength=0.001,
        damping=0.09,
        overlap=0,
    )
    
    # Customize edge appearance
    for edge in nt.edges:
        edge['width'] = 2
        edge['arrows'] = {'to': {'enabled': True, 'type': 'arrow'}}
        edge['color'] = {'color': '#6366f1', 'highlight': '#4f46e5'}
        edge['font'] = {'size': 12, 'color': '#4b5563', 'face': 'Arial'}

    # Customize node appearance
    for node in nt.nodes:
        node['color'] = {'background': '#e0e7ff', 'border': '#6366f1', 'highlight': {'background': '#c7d2fe', 'border': '#4f46e5'}}
        node['font'] = {'size': 14, 'color': '#1e293b'}
        node['shape'] = 'dot'
        node['size'] = 25

    # Generate HTML with iframe to isolate styles
    html = nt.generate_html()
    html = html.replace("'", '"')

    return f"""<iframe style="width: 100%; height: 700px; margin: 0 auto; border-radius: 12px; box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.1), 0 4px 6px -4px rgba(0, 0, 0, 0.1);" 
        name="result" allow="midi; geolocation; microphone; camera; display-capture; encrypted-media;" 
        sandbox="allow-modals allow-forms allow-scripts allow-same-origin allow-popups 
        allow-top-navigation-by-user-activation allow-downloads" allowfullscreen="" 
        allowpaymentrequest="" frameborder="0" srcdoc='{html}'></iframe>"""

def process_and_visualize(text, progress=gr.Progress()):
    if not text:
        raise gr.Error("⚠️ Text must be provided!")
    
    # Check if we're processing the first example for caching
    is_first_example = text == EXAMPLES[0][0]
    
    # Try to load from cache if it's the first example
    if is_first_example and os.path.exists(EXAMPLE_CACHE_FILE):
        try:
            progress(0.3, desc="Loading from cache...")
            with open(EXAMPLE_CACHE_FILE, 'rb') as f:
                cache_data = pickle.load(f)
                
            progress(1.0, desc="Loaded from cache!")
            return cache_data["graph_html"], cache_data["entities_viz"], cache_data["json_data"], cache_data["stats"]
        except Exception as e:
            print(f"Cache loading error: {str(e)}")
            # Continue with normal processing if cache fails
    
    progress(0, desc="Starting extraction...")
    json_data = extract(text)
    
    progress(0.5, desc="Creating entity visualization...")
    entities_viz = create_custom_entity_viz(json_data, text)
    
    progress(0.8, desc="Building knowledge graph...")
    graph_html = create_graph(json_data)
    
    node_count = len(json_data["nodes"])
    edge_count = len(json_data["edges"])
    stats = f"📊 Extracted {node_count} entities and {edge_count} relationships"
    
    # Save to cache if it's the first example
    if is_first_example:
        try:
            cache_data = {
                "graph_html": graph_html,
                "entities_viz": entities_viz,
                "json_data": json_data,
                "stats": stats
            }
            with open(EXAMPLE_CACHE_FILE, 'wb') as f:
                pickle.dump(cache_data, f)
        except Exception as e:
            print(f"Cache saving error: {str(e)}")
    
    progress(1.0, desc="Complete!")
    return graph_html, entities_viz, json_data, stats

# Example texts in different languages
EXAMPLES = [
    [handle_text("""The family of Azerbaijan President Ilham Aliyev leads a charmed, glamorous life, thanks in part to financial interests in almost every sector of the economy. His wife, Mehriban, comes from the privileged and powerful Pashayev family that owns banks, insurance and construction companies, a television station and a line of cosmetics. She has led the Heydar Aliyev Foundation, Azerbaijan’s pre-eminent charity behind the construction of schools, hospitals and the country’s major sports complex. Their eldest daughter, Leyla, editor of Baku magazine, and her sister, Arzu, have financial stakes in a firm that won rights to mine for gold in the western village of Chovdar and Azerfon, the country’s largest mobile phone business. Arzu is also a significant shareholder in SW Holding, which controls nearly every operation related to Azerbaijan Airlines (“Azal”), from meals to airport taxis. Both sisters and brother Heydar own property in Dubai valued at roughly $75 million in 2010; Heydar is the legal owner of nine luxury mansions in Dubai purchased for some $44 million.""")],
    
    [handle_text("""Legendary rock band Aerosmith has officially announced their retirement from touring after 54 years, citing 
    lead singer Steven Tyler's unrecoverable vocal cord injury. 
    The decision comes after months of unsuccessful treatment for Tyler's fractured larynx, 
    which he suffered in September 2023.""")],
    
    [handle_text("""Pop star Justin Timberlake, 43, had his driver's license suspended by a New York judge during a virtual 
    court hearing on August 2, 2024. The suspension follows Timberlake's arrest for driving while intoxicated (DWI) 
    in Sag Harbor on June 18. Timberlake, who is currently on tour in Europe, 
    pleaded not guilty to the charges.""")],
]

# Function to preprocess the first example when the app starts
def generate_first_example_cache():
    """Generate cache for the first example if it doesn't exist"""
    if not os.path.exists(EXAMPLE_CACHE_FILE):
        print("Generating cache for first example...")
        try:
            text = EXAMPLES[0][0]
            # model = MODEL_LIST[0] if MODEL_LIST else None
            
            # if model:
            # Extract data
            json_data = extract(text, model)
            entities_viz = create_custom_entity_viz(json_data, text)
            graph_html = create_graph(json_data)
            
            node_count = len(json_data["nodes"])
            edge_count = len(json_data["edges"])
            stats = f"📊 Extracted {node_count} entities and {edge_count} relationships"
            
            # Save to cache
            cache_data = {
                "graph_html": graph_html,
                "entities_viz": entities_viz,
                "json_data": json_data,
                "stats": stats
            }
            with open(EXAMPLE_CACHE_FILE, 'wb') as f:
                pickle.dump(cache_data, f)
            
            print("First example cache generated successfully")
            return cache_data
        except Exception as e:
            print(f"Error generating first example cache: {str(e)}")
    else:
        print("First example cache already exists")
        try:
            with open(EXAMPLE_CACHE_FILE, 'rb') as f:
                return pickle.load(f)
        except Exception as e:
            print(f"Error loading existing cache: {str(e)}")
    
    return None

def create_ui():
    # Try to generate/load the first example cache
    first_example_cache = generate_first_example_cache()
    
    with gr.Blocks(css=CUSTOM_CSS, title=TITLE) as demo:
        # Header 
        gr.Markdown(f"# {TITLE}")
        gr.Markdown(f"{SUBTITLE}")
        
        with gr.Row():
            gr.Markdown("🌍 **Multilingual Support Available**")
        
        # Main content area - redesigned layout
        with gr.Row():
            # Left panel - Input controls
            with gr.Column(scale=1):
                # input_model = gr.Dropdown(
                #     MODEL_LIST, 
                #     label="🤖 Select Model",
                #     info="Choose a model to process your text",
                #     value=MODEL_LIST[0] if MODEL_LIST else None
                # )
                
                input_text = gr.TextArea(
                    label="📝 Input Text", 
                    info="Enter text in any language to extract a knowledge graph",
                    placeholder="Enter text here...",
                    lines=8,
                    value=EXAMPLES[0][0]  # Pre-fill with first example
                )
                
                with gr.Row():
                    submit_button = gr.Button("🚀 Extract & Visualize", variant="primary", scale=2)
                    clear_button = gr.Button("🔄 Clear", variant="secondary", scale=1)
                
                # Statistics will appear here
                stats_output = gr.Markdown("", label="🔍 Analysis Results")
            
            # Right panel - Examples moved to right side
            with gr.Column(scale=1):
                gr.Markdown("## 📚 Example Texts")
                gr.Examples(
                    examples=EXAMPLES,
                    inputs=input_text,
                    label=""
                )
                
                # JSON output moved to right side as well
                with gr.Accordion("📊 JSON Data", open=False):
                    output_json = gr.JSON(label="")
        
        # Full width visualization area at the bottom
        with gr.Row():
            # Full width visualization area
            with gr.Tabs():
                with gr.TabItem("🧩 Knowledge Graph"):
                    output_graph = gr.HTML(label="")
                
                with gr.TabItem("🏷️ Entity Recognition"):
                    output_entity_viz = gr.HTML(label="")
        
        # Functionality
        submit_button.click(
            fn=process_and_visualize, 
            # inputs=[input_text, input_model],
            inputs=[input_text],
            outputs=[output_graph, output_entity_viz, output_json, stats_output]
        )
        
        clear_button.click(
            fn=lambda: [None, None, None, ""],
            inputs=[],
            outputs=[output_graph, output_entity_viz, output_json, stats_output]
        )
        
        # Set initial values from cache if available
        if first_example_cache:
            # Use this to set initial values when the app loads
            demo.load(
                lambda: [
                    first_example_cache["graph_html"], 
                    first_example_cache["entities_viz"], 
                    first_example_cache["json_data"], 
                    first_example_cache["stats"]
                ],
                inputs=None,
                outputs=[output_graph, output_entity_viz, output_json, stats_output]
            )
        
        # Footer
        gr.Markdown("---")
        gr.Markdown("📋 **Instructions:** Enter text in any language, select a model, and click 'Extract & Visualize' to generate a knowledge graph.")
        gr.Markdown("🛠️ Powered by Phi-3 Instruct Graph | Emergent Methods")
        
    return demo

demo = create_ui()
demo.launch(share=False)