import gradio as gr
import model
import os

FEATURES = [
    {"name": "Administrative", "desc": "Number of administrative pages visited", "type": "number", "min": 0, "max": 30},
    {"name": "Administrative_Duration", "desc": "Total time spent on administrative pages (seconds)", "type": "number", "min": 0, "max": 40000},
    {"name": "Informational", "desc": "Number of informational pages visited", "type": "number", "min": 0, "max": 24},
    {"name": "Informational_Duration", "desc": "Total time spent on informational pages (seconds)", "type": "number", "min": 0, "max": 25480},
    {"name": "ProductRelated", "desc": "Number of product-related pages visited", "type": "number", "min": 0, "max": 705},
    {"name": "ProductRelated_Duration", "desc": "Total time spent on product-related pages (seconds)", "type": "number", "min": 0, "max": 63973},
    {"name": "BounceRates", "desc": "Bounce rate value (0-1)", "type": "number", "step": 0.001, "min": 0, "max": 1},
    {"name": "ExitRates", "desc": "Exit rate value (0-1)", "type": "number", "step": 0.001, "min": 0, "max": 1},
    {"name": "PageValues", "desc": "Page value (0-361)", "type": "number", "step": 0.01, "min": 0, "max": 361},
    {"name": "SpecialDay", "desc": "Special day proximity (0-1)", "type": "number", "step": 0.01, "min": 0, "max": 1},
    {"name": "Month", "desc": "Month of the visit", "type": "select", "options": ["Jan", "Feb", "Mar", "Apr", "May", "June", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"]},
    {"name": "OperatingSystems", "desc": "Operating system (1-8)", "type": "number", "min": 1, "max": 8},
    {"name": "Browser", "desc": "Browser (1-13)", "type": "number", "min": 1, "max": 13},
    {"name": "Region", "desc": "Region (1-9)", "type": "number", "min": 1, "max": 9},
    {"name": "TrafficType", "desc": "Traffic type (1-20)", "type": "number", "min": 1, "max": 20},
    {"name": "VisitorType", "desc": "Type of visitor", "type": "select", "options": ["New_Visitor", "Returning_Visitor", "Other"]},
    {"name": "Weekend", "desc": "Is weekend?", "type": "select", "options": ["True", "False"]},
]

# 1. Main Tab: Show metrics and images
def main_tab():
    metrics = """
    **Model Metrics:**
    - Accuracy: 0.8619
    - Precision: 0.5335
    - Recall: 0.5497
    - F1 Score: 0.5415
    - ROC AUC Score: 0.7330
    """
    img_dir = "imgs"
    img_files = [
        ("Classification Metrics", os.path.join(img_dir, "classification_metrics_bar_plot.png")),
        ("Confusion Matrix", os.path.join(img_dir, "confusion_matrix.png")),
        ("ROC Curve", os.path.join(img_dir, "roc_curve.png")),
    ]
    # Only return label and path for existing files
    img_comps = [(label, img) for label, img in img_files if os.path.exists(img)]
    return metrics, img_comps

# 2. Dataset Tab: Show info and column descriptions
def dataset_tab():
    dataset_info = """
    **Dataset:** [Online Shoppers Purchasing Intention Dataset](https://archive.ics.uci.edu/dataset/468/online+shoppers+purchasing+intention+dataset)
    - 18,000+ sessions, 17 features + target
    - Binary classification: Revenue (True/False)
    """
    columns_html = """
    <table>
      <thead>
      <tr>
        <th>Column</th>
        <th>Description</th>
        <th>Range/Options</th>
      </tr>
      </thead>
      <tbody>
    """
    for f in FEATURES:
      rng = f.get("options", f"{f.get('min', '')} - {f.get('max', '')}")
      if isinstance(rng, list):
        rng = ", ".join(str(x) for x in rng)
      columns_html += f"<tr><td>{f['name']}</td><td>{f['desc']}</td><td>{rng}</td></tr>\n"
    columns_html += "</tbody></table>"
    return gr.Markdown(dataset_info), gr.HTML(columns_html)

# 3. Model Tab: Prediction UI
def predict_gradio(*inputs):
    row = {}
    for i, f in enumerate(FEATURES):
        row[f["name"]] = inputs[i]
        
    pred, proba = model.predict_from_row(row)
    return f"Prediction: {pred}\nProbabilities: {proba}"

inputs = []
example_row = {
    'Administrative': 3,
    'Administrative_Duration': 80.0,
    'Informational': 0,
    'Informational_Duration': 0.0,
    'ProductRelated': 20,
    'ProductRelated_Duration': 500.0,
    'BounceRates': 0.02,
    'ExitRates': 0.04,
    'PageValues': 0.0,
    'SpecialDay': 0.0,
    'Month': 'Nov',
    'OperatingSystems': 2,
    'Browser': 2,
    'Region': 1,
    'TrafficType': 2,
    'VisitorType': 'Returning_Visitor',
    'Weekend': 'False',
}
for f in FEATURES:
    label = f"{f['name']} ({f['desc']})"
    default = example_row.get(f['name'])
    if f["type"] == "select":
        inputs.append(gr.Dropdown(choices=f["options"], label=label, value=default))
    else:
        step = f.get("step", 1)
        inputs.append(gr.Number(label=label, minimum=f["min"], maximum=f["max"], step=step, value=default))

model_tab = gr.Interface(
    fn=predict_gradio,
    inputs=inputs,
    outputs="text",
    title="Model Prediction",
    description="Fill in all fields. Ranges and descriptions are shown for each input.",
)

with gr.Blocks() as demo:
    with gr.Tab("Main"):
        metrics, imgs = main_tab()
        gr.Markdown(metrics)
        for label, img_path in imgs:
            gr.Image(value=img_path, label=label)
    with gr.Tab("Dataset"):
        dataset_info, columns_md = dataset_tab()
        # gr.Markdown(dataset_info.value)
        # gr.Markdown(columns_md.value)
    with gr.Tab("Model"):
        model_tab.render()

demo.launch()