# app.py (with human-readable labels)
import gradio as gr
import joblib

# --- Load Model and Vectorizer ---
try:
    model = joblib.load('logistic_regression_model.joblib')
    vectorizer = joblib.load('tfidf_vectorizer.joblib')
    print("✅ Model and vectorizer loaded successfully!")
except Exception as e:
    print(f"🛑 Error loading files: {e}")
    model, vectorizer = None, None

# --- Define Prediction Logic with Label Mapping ---
def predict_sentiment(text):
    if not model or not vectorizer:
        return "ERROR: Model is not loaded. Check terminal logs."
    
    try:
        # 1. Transform the input text
        vectorized_text = vectorizer.transform([text])
        
        # 2. Make a numerical prediction
        prediction = model.predict(vectorized_text)
        numeric_prediction = prediction[0]
        
        # --- NEW CODE: Map prediction to labels ---
        # 3. Define the mapping from numbers to labels
        sentiment_map = {
            0: "NEUTRAL",
            1: "HAPPY",
            2: "SAD"
        }
        
        # 4. Get the label from the map.
        #    The .get() method safely returns a default value if the key isn't found.
        sentiment_label = sentiment_map.get(numeric_prediction, "Unknown Prediction")
        
        # 5. Return the final human-readable label
        return sentiment_label
        
    except Exception as e:
        print(f"--- PREDICTION ERROR --- \n{e}\n --- END ---")
        return f"An error occurred during prediction: {e}"

# --- Create and Launch the Gradio Interface ---
iface = gr.Interface(
    fn=predict_sentiment,
    inputs=gr.Textbox(lines=5, label="Enter a Sentence"),
    outputs=gr.Label(label="Predicted Sentiment"),
    title="Sentiment Analysis Model",
    description="Analyzes text and classifies it as Happy, Sad, or Neutral."
)

print("🚀 Launching Gradio app...")
iface.launch()