app.py

import os
import spacy
import pickle
import random
import logging
import asyncio
import rapidjson

import gradio as gr
import networkx as nx

# from dotenv import load_dotenv
from llm_graph import LLMGraph, MODEL_LIST

from pyvis.network import Network
from spacy import displacy
from spacy.tokens import Span

logging.basicConfig(level=logging.INFO)
# load_dotenv()

# Constants
TITLE = "🌐 Text2Graph: Extract Knowledge Graphs from Natural Language"
SUBTITLE = "✨ Extract and visualize knowledge graphs from texts in any language!"

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

# Cache directory and file paths
CACHE_DIR = "./cache"
WORKING_DIR = "./sample"
EXAMPLE_CACHE_FILE = os.path.join(CACHE_DIR, "first_example_cache.pkl")
GRAPHML_FILE = WORKING_DIR + "/graph_chunk_entity_relation.graphml"

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

# Initialize the LLMGraph model
model = LLMGraph()

def get_random_light_color():
    """
    Color utilities
    """

    r = random.randint(140, 255)
    g = random.randint(140, 255)
    b = random.randint(140, 255)

    return f"#{r:02x}{g:02x}{b:02x}"

def handle_text(text=""):
    """
    Text preprocessing
    """

    # Catch empty text
    if not text:
        return ""
    
    return " ".join(text.split())

def extract_kg(text="", model_name=MODEL_LIST[0]):
    """
    Extract knowledge graph from text
    """

    # Catch empty text
    if not text or not model_name:
        raise gr.Error("⚠️ Both text and model must be provided!")
    try:
        result = model.extract(text, model_name)

        if isinstance(result, dict):
            return result
        else: # convert string to dict
            return rapidjson.loads(result)
    except Exception as e:
        raise gr.Error(f"❌ Extraction error: {str(e)}")

def find_token_indices(doc, substring, text):
    """
    Find token indices for a given substring in the text
    based on the provided spaCy doc.
    """

    result = []
    start_idx = text.find(substring)
    
    while start_idx != -1:
        end_idx = start_idx + len(substring)
        start_token = None
        end_token = None

        for token in doc:
            if token.idx == start_idx:
                start_token = token.i
            if token.idx + len(token) == end_idx:
                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_idx = text.find(substring, end_idx)

    return result

def create_custom_entity_viz(data, full_text, type_col="type"):
    """
    Create custom entity visualization using spaCy's displacy
    """

    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 entity in entity_spans:
            start = entity["start"]
            end = entity["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_col, "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: gray; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);">
        {html}
    </div>
    """

    return styled_html

def create_graph(json_data, model_name=MODEL_LIST[0]):
    """
    Create interactive knowledge graph using pyvis
    """

    if model_name == MODEL_LIST[0]:
        G = nx.Graph()

        # Add nodes with tooltips and error handling for missing keys
        for node in json_data['nodes']:
            # Get node type with fallback
            type = node.get("type", "Entity")

            # Get detailed type with fallback
            detailed_type = node.get("detailed_type", type)
            
            # Use node ID and type info for the tooltip
            G.add_node(node['id'], title=f"{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)
    else:
        G = nx.read_graphml(GRAPHML_FILE)

    # Create network visualization
    network = Network(
        width="100%",
        # height="700px",
        height="100vh",
        notebook=False,
        bgcolor="#f8fafc", 
        font_color="#1e293b"
    )
    
    # Configure network display
    network.from_nx(G)
    if model_name == MODEL_LIST[0]:
        network.barnes_hut(
            gravity=-3000,
            central_gravity=0.3,
            spring_length=50,
            spring_strength=0.001,
            damping=0.09,
            overlap=0,
        )

    # Customize node appearance
    for node in network.nodes:
        if "description" in node:
            node["title"] = node["description"]

        node['color'] = {'background': '#e0e7ff', 'border': '#6366f1', 'highlight': {'background': '#c7d2fe', 'border': '#4f46e5'}}
        node['font'] = {'size': 14, 'color': '#1e293b'}
        node['shape'] = 'dot'
        node['size'] = 20
    
    # Customize edge appearance
    for edge in network.edges:
        if "description" in edge:
            edge["title"] = edge["description"]

        edge['width'] = 4
        # edge['arrows'] = {'to': {'enabled': False, 'type': 'arrow'}}
        edge['color'] = {'color': '#6366f1', 'highlight': '#4f46e5'}
        edge['font'] = {'size': 12, 'color': '#4b5563', 'face': 'Arial'}
    
    # Generate HTML with iframe to isolate styles
    html = network.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, model_name, progress=gr.Progress()):
    """
    Process text and visualize knowledge graph and entities
    """

    if not text or not model_name:
        raise gr.Error("⚠️ Both text and model must be provided!")
    
    # Check if we're processing the first example for caching
    is_first_example = text == EXAMPLES[0][0]
    
    # Ensure RAG is initialized
    # TODO: Clear all the previous inserted texts
    asyncio.run(model.initialize_rag()) 

    # Try to load from cache if it's the first example
    if is_first_example and model_name == MODEL_LIST[0] and os.path.exists(EXAMPLE_CACHE_FILE):
        try:
            progress(0.3, desc="Loading from cache...")
            with open(EXAMPLE_CACHE_FILE, 'rb') as f:
                cached_data = pickle.load(f)
                
            progress(1.0, desc="Loaded from cache!")
            return cached_data["graph_html"], cached_data["entities_viz"], cached_data["json_data"], cached_data["stats"]
        except Exception as e:
            logging.error(f"Cache loading error: {str(e)}")

    # Continue with normal processing if cache fails
    progress(0, desc="Starting extraction...")
    json_data = extract_kg(text, model_name)
    
    progress(0.5, desc="Creating entity visualization...")
    if model_name == MODEL_LIST[0]:
        entities_viz = create_custom_entity_viz(json_data, text, type_col="type")
    else:
        entities_viz = create_custom_entity_viz(json_data, text, type_col="entity_type")
    
    progress(0.8, desc="Building knowledge graph...")
    graph_html = create_graph(json_data, model_name)
    
    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 and model_name == MODEL_LIST[0]:
        try:
            cached_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(cached_data, f)
        except Exception as e:
            logging.error(f"Cache saving error: {str(e)}")
    
    progress(1.0, desc="Complete!")
    return graph_html, entities_viz, json_data, stats

# Example texts
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("""Les jardins du Luxembourg, situés au cœur du sixième arrondissement de Paris, offrent un véritable havre de paix aux citadins pressés. 
                 Créés au début du dix-septième siècle sur l'initiative de Marie de Médicis, ces jardins à la française s'étendent sur vingt-trois hectares 
                 et abritent le Palais du Luxembourg, siège du Sénat français. Les promeneurs peuvent y admirer les parterres de fleurs soigneusement entretenus, 
                 les bassins ornés de statues mythologiques, et les allées bordées de marronniers centenaires. Chaque matin, les jardiniers s'affairent à tailler 
                 les buis et à arroser les rosiers, perpétuant ainsi une tradition d'excellence horticole qui fait la fierté de la capitale française.""")],
]

def generate_first_example():
    """
    Generate cache for the first example if it doesn't exist when the app starts
    """

    if not os.path.exists(EXAMPLE_CACHE_FILE):
        logging.info("Generating cache for first example...")

        try:
            text = EXAMPLES[0][0]
            model_name = MODEL_LIST[0] if MODEL_LIST else None

            # Extract data
            json_data = extract_kg(text, model_name)
            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
            cached_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(cached_data, f)
            logging.info("First example cache generated successfully")

            return cached_data
        except Exception as e:
            logging.error(f"Error generating first example cache: {str(e)}")
    else:
        logging.info("First example cache already exists")

        # Load existing cache
        try:
            with open(EXAMPLE_CACHE_FILE, 'rb') as f:
                return pickle.load(f)
        except Exception as e:
            logging.error(f"Error loading existing cache: {str(e)}")
    
    return None

def create_ui():
    """
    Create the Gradio UI
    """

    # Try to generate/load the first example cache
    first_example = generate_first_example()
    
    with gr.Blocks(css=CUSTOM_CSS, title=TITLE) as demo:
        # Header 
        gr.Markdown(f"# {TITLE}")
        gr.Markdown(f"{SUBTITLE}")
        
        # Main content area
        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],
            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:
            # Use this to set initial values when the app loads
            demo.load(
                lambda: [
                    first_example["graph_html"], 
                    first_example["entities_viz"], 
                    first_example["json_data"], 
                    first_example["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 [GPT-4.1-mini](https://platform.openai.com/docs/models/gpt-4.1-mini) and [Phi-3-mini-128k-instruct-graph](https://huggingface.co/EmergentMethods/Phi-3-mini-128k-instruct-graph)")
        
    return demo

def main():
    """
    Main function to run the Gradio app
    """

    demo = create_ui()
    demo.launch(share=False)
    
if __name__ == "__main__":
    main()