from flask import Flask, request, render_template
import pickle
from preprocessing import preprocess_text  # pastikan file preprocessing.py ada di folder yang sama

app = Flask(__name__)

with open('model_SVM.pkl', 'rb') as f:
    model_SVM = pickle.load(f)
with open('model_NB.pkl', 'rb') as f:
    model_NB = pickle.load(f)

def classify_text(processed_text):
    result_svm = model_SVM.predict([processed_text])[0]
    result_nb = model_NB.predict([processed_text])[0]
    return result_svm, result_nb

@app.route('/', methods=['GET', 'POST'])
def index():
    input_text = ''
    processed_steps = {}
    result_svm = ''
    result_nb = ''
    tfidf_dict = {}

    if request.method == 'POST':
        input_text = request.form['text']
        processed_steps = preprocess_text(input_text)
        # Gunakan hasil akhir (stemming) untuk prediksi dan TF-IDF
        final_text = processed_steps.get("stemming", "")
        result_svm, result_nb = classify_text(final_text)
        
        # Dapatkan nilai TF-IDF dari vectorizer di pipeline model_SVM
        vectorizer = model_SVM.named_steps['vectorizer']
        tfidf_vector = vectorizer.transform([final_text])
        dense = tfidf_vector.todense().tolist()[0]
        features = vectorizer.get_feature_names_out()
        tfidf_dict = {feature: round(value, 3) for feature, value in zip(features, dense) if value > 0}

    return render_template('index.html', 
                           input_text=input_text, 
                           processed_steps=processed_steps,
                           result_svm=result_svm, 
                           result_nb=result_nb,
                           tfidf_dict=tfidf_dict)

if __name__ == '__main__':
    app.run(debug=True)