import gradio as gr
import joblib

# Load your trained artifacts
model = joblib.load("model.pkl")
vectorizer = joblib.load("vectorizer.pkl")

labels = {0: "Negative", 1: "Positive"}  # adjust to your classes

def classify(text):
    probs = model.predict_proba(vectorizer.transform([text]))[0]
    pred_idx = probs.argmax()
    confidence = float(probs[pred_idx])
    label = labels[pred_idx]

    color = "#4CAF50" if pred_idx == 1 else "#F44336"  # green/red
    styled_label = f"<span style='color:{color}; font-weight:bold'>{label}</span>"

    return styled_label, f"{confidence:.2%}"

with gr.Blocks() as demo:
    gr.Markdown("## 🌱 Agricultural Text Classifier")
    gr.Markdown("Enter a description and see the prediction with confidence.")

    with gr.Row():
        with gr.Column():
            text_in = gr.Textbox(lines=3, placeholder="Type here...")
            btn = gr.Button("Classify")
            samples = gr.Row()
            for sample in [
                "Healthy maize after seasonal rains",
                "Coffee plants showing signs of leaf rust",
                "Pest infestation on cassava leaves"
            ]:
                gr.Button(sample).click(fn=lambda s=sample: s, outputs=text_in)

        with gr.Column():
            label_out = gr.HTML()
            conf_out = gr.Textbox(label="Confidence", interactive=False)

    btn.click(classify, inputs=text_in, outputs=[label_out, conf_out])

demo.launch()