app.py

import pandas as pd
import numpy as np
import gradio as gr
from xgboost import XGBRegressor
from sklearn.preprocessing import LabelEncoder

EXCEL_PATH = "microalgae_pot_experiment_corrected_doses.xlsx"
MODEL_PATH = "EcoGrowAI_yield_model.json"

df = pd.read_excel(EXCEL_PATH)

# Encode categorical columns same as training
for col in ["Crop", "Microalgae_Strain"]:
    le = LabelEncoder()
    df[col] = le.fit_transform(df[col])

drop_cols = ["Yield_g_per_pot", "Relative_Yield_%"]
feature_cols = [c for c in df.columns if c not in drop_cols]

model = XGBRegressor()
model.load_model(MODEL_PATH)


def make_row_label(i):
    return f"Row {i+1} (index={i})"


def parse_index(label):
    return int(label.split("index=")[1].strip(")"))


def on_row_select(row_label):
    idx = parse_index(row_label)
    row_df = df.iloc[[idx]][feature_cols]
    return "### Selected Row (Input Features)\n\n" + row_df.T.to_markdown()


def on_predict(row_label):
    idx = parse_index(row_label)
    row = df.iloc[idx][feature_cols]
    X_row = row.astype(float).values.reshape(1, -1)

    # Prediction
    pred = model.predict(X_row)[0]
    pred_rounded = float(np.round(pred, 3))

    # Ground truth
    actual = float(df.iloc[idx]["Yield_g_per_pot"])

    # Error (always positive)
    error = abs(pred_rounded - actual)
    error_rounded = float(np.round(error, 3))

    # Markdown display
    features_md = "### Input Features\n\n" + row.to_frame().to_markdown()
    result_md = (
        f"### Results\n\n"
        f"**Predicted Yield (g per pot):** {pred_rounded}\n\n"
        f"**Actual Yield (Ground Truth):** {actual}\n\n"
        f"**Absolute Error:** {error_rounded}"
    )

    return "✅ Prediction complete", features_md, result_md


row_choices = [make_row_label(i) for i in range(len(df))]

with gr.Blocks(css="""
body {background:#f5f7fa;color:#0b1220;font-family:Inter,sans-serif;}
.gradio-container{max-width:1200px;margin:20px auto;}
.card{background:white;padding:20px;border-radius:10px;
box-shadow:0 4px 16px rgba(0,0,0,0.05);}
""") as demo:
    with gr.Column(elem_classes="card"):
        gr.Markdown("# EcoGrowAI — Yield Prediction")
        gr.Markdown("Select any experimental row and predict **Yield (g per pot)** using the trained XGBoost model.")

        with gr.Row():
            with gr.Column(scale=1):
                row_dropdown = gr.Dropdown(
                    label="Select Row", choices=row_choices, value=row_choices[0], interactive=True
                )
                row_info = gr.Markdown("No row selected yet.")
                predict_button = gr.Button("Predict", variant="primary")

            with gr.Column(scale=1):
                status = gr.Markdown("")
                features_md = gr.Markdown("")
                result_md = gr.Markdown("")

    row_dropdown.change(on_row_select, row_dropdown, row_info)
    predict_button.click(on_predict, row_dropdown, [status, features_md, result_md])

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)