app.py

import gradio as gr
from transformers import pipeline
import requests
import os  # Import for accessing environment variables

# Load the sentiment analysis model
classifier = pipeline('sentiment-analysis', model='krishnamishra8848/movie_sentiment_analysis')

# Get API key from secrets
RAPIDAPI_KEY = os.environ.get("RAPIDAPI_KEY")  # Securely access the secret

# Language detection function
def detect_language(text):
    detect_url = "https://google-translator9.p.rapidapi.com/v2/detect"
    detect_payload = {"q": text}
    headers = {
        "x-rapidapi-key": RAPIDAPI_KEY,  # Use the secret here
        "x-rapidapi-host": "google-translator9.p.rapidapi.com",
        "Content-Type": "application/json"
    }

    response = requests.post(detect_url, json=detect_payload, headers=headers)
    if response.status_code == 200:
        detections = response.json().get('data', {}).get('detections', [[]])[0]
        if detections:
            return detections[0].get('language')
    return None

# Translation function
def translate_text(text, source_language, target_language="en"):
    translate_url = "https://google-translator9.p.rapidapi.com/v2"
    translate_payload = {
        "q": text,
        "source": source_language,
        "target": target_language,
        "format": "text"
    }
    headers = {
        "x-rapidapi-key": RAPIDAPI_KEY,  # Use the secret here
        "x-rapidapi-host": "google-translator9.p.rapidapi.com",
        "Content-Type": "application/json"
    }

    response = requests.post(translate_url, json=translate_payload, headers=headers)
    if response.status_code == 200:
        translations = response.json().get('data', {}).get('translations', [{}])
        if translations:
            return translations[0].get('translatedText')
    return None

# Main function for Gradio
def analyze_sentiment_with_steps(text):
    # Step 1: Detecting Language
    yield "Detecting Language..."
    detected_language = detect_language(text)
    if not detected_language:
        yield "Error: Could not detect the language."
        return

    yield f"Language Detected: {detected_language.upper()}"

    # Step 2: Translating if necessary
    if detected_language != "en":
        yield "Translating text to English..."
        text = translate_text(text, detected_language)
        if not text:
            yield "Error: Could not translate the input text."
            return

    # Step 3: Sending to model
    yield "Sending to Model..."

    # Step 4: Sentiment analysis
    result = classifier(text)
    label_mapping = {"LABEL_0": "negative", "LABEL_1": "positive"}
    sentiment = label_mapping[result[0]['label']]
    confidence = result[0]['score'] * 100  # Convert to percentage
    color = "green" if sentiment == "positive" else "red"
    yield f"<h1 style='color:{color}'>Prediction: {sentiment.capitalize()}</h1><p>Confidence: {confidence:.2f}%</p>"

# Gradio interface
interface = gr.Interface(
    fn=analyze_sentiment_with_steps,
    inputs=gr.Textbox(
        label="Enter Movie Review",
        placeholder="Type your review in any language...",
        lines=3
    ),
    outputs="html",
    live=True,
    title="Multilingual Movie Sentiment Analysis"
)

# Launch Gradio app
interface.launch(share=True)