import gradio as gr
import numpy as np
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing.sequence import pad_sequences
import pickle
import re

# Load model and tokenizer
model = load_model("best_GRU_tuning_model.h5")
with open("my_tokenizer.pkl","rb") as f:
    tokenizer = pickle.load(f)


def preprocess_text(text):
    text = text.lower()
    text = re.sub(r'[^a-zA-Z\s]', '', text).strip()
    return text


def predict_sentiment(raw_text):
    cleaned = preprocess_text(raw_text)
    seq = tokenizer.texts_to_sequences([cleaned])
    padded_seq = pad_sequences(seq, maxlen=200)
    probs = model.predict(padded_seq)
    predicted_class = np.argmax(probs, axis=1)[0]
    rating = predicted_class + 1
    return f"Predicted rating: {rating} (probabilities={probs[0]})"


demo = gr.Interface(fn=predict_sentiment,
                    inputs="text",
                    outputs="label")
demo.launch()