app.py

"""
SentiMeter — IndoBERT Sentiment Analysis Backend
Uses mdhugol/indonesia-bert-sentiment-classification
"""

import os
import sys
import traceback
from flask import Flask, request, jsonify, send_from_directory
from flask_cors import CORS

app = Flask(__name__, static_folder='.', static_url_path='')
CORS(app)

# ─── Model Loading ────────────────────────────────────
MODEL_NAME = "mdhugol/indonesia-bert-sentiment-classification"
LABEL_MAP = {
    'LABEL_0': 'Positif',
    'LABEL_1': 'Netral',
    'LABEL_2': 'Negatif',
}

sentiment_pipeline = None
model_error = None


def load_model():
    """Load the IndoBERT sentiment model. Returns True on success."""
    global sentiment_pipeline, model_error
    try:
        print(f"[IndoBERT] Loading model: {MODEL_NAME} ...")
        from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline

        tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
        model = AutoModelForSequenceClassification.from_pretrained(MODEL_NAME)
        sentiment_pipeline = pipeline(
            "sentiment-analysis",
            model=model,
            tokenizer=tokenizer,
            truncation=True,
            max_length=512,
        )
        model_error = None
        print("[IndoBERT] Model loaded successfully!")
        return True
    except Exception as e:
        model_error = str(e)
        sentiment_pipeline = None
        print(f"[IndoBERT] ERROR loading model: {e}", file=sys.stderr)
        traceback.print_exc()
        return False


# ─── API Routes ───────────────────────────────────────

@app.route('/api/health', methods=['GET'])
def health():
    """Check if the model is loaded and ready."""
    if sentiment_pipeline is not None:
        return jsonify({"status": "ok", "model": MODEL_NAME})
    else:
        return jsonify({
            "status": "error",
            "model": MODEL_NAME,
            "error": model_error or "Model not loaded"
        }), 503


@app.route('/api/sentiment', methods=['POST'])
def analyze_sentiment():
    """
    Analyze sentiment for a batch of texts.
    Expects JSON: { "texts": ["text1", "text2", ...] }
    Returns JSON:  [ {"label": "Positif", "score": 0.95}, ... ]
    """
    if sentiment_pipeline is None:
        return jsonify({
            "error": True,
            "message": f"Model IndoBERT gagal dimuat: {model_error or 'Unknown error'}. "
                       "Pastikan koneksi internet aktif dan dependensi sudah terinstal.",
        }), 503

    data = request.get_json(silent=True)
    if not data or 'texts' not in data:
        return jsonify({"error": True, "message": "Request harus berisi field 'texts'."}), 400

    texts = data['texts']
    if not isinstance(texts, list) or len(texts) == 0:
        return jsonify({"error": True, "message": "'texts' harus berupa array string yang tidak kosong."}), 400

    # Limit batch size to prevent OOM
    MAX_BATCH = 64
    results = []

    try:
        for i in range(0, len(texts), MAX_BATCH):
            batch = texts[i:i + MAX_BATCH]
            # Ensure all texts are non-empty strings
            batch = [str(t).strip() or "." for t in batch]
            preds = sentiment_pipeline(batch)
            for pred in preds:
                label = LABEL_MAP.get(pred['label'], pred['label'])
                results.append({
                    "label": label,
                    "score": round(pred['score'], 4),
                })
    except Exception as e:
        return jsonify({
            "error": True,
            "message": f"Error saat proses sentimen: {str(e)}"
        }), 500

    return jsonify(results)


# ─── Static File Serving ─────────────────────────────

@app.route('/')
def serve_index():
    return send_from_directory(app.static_folder, 'index.html')


@app.errorhandler(404)
def handle_404(e):
    path = request.path.lstrip('/')
    if not path:
        return send_from_directory(app.static_folder, 'index.html')
    
    # Try serving the file as is (for static assets)
    if os.path.isfile(os.path.join(app.static_folder, path)):
        return send_from_directory(app.static_folder, path)
    
    # Try adding .html (for clean URLs)
    html_path = path + ".html"
    if os.path.isfile(os.path.join(app.static_folder, html_path)):
        return send_from_directory(app.static_folder, html_path)
    
    # Fallback to index.html
    return send_from_directory(app.static_folder, 'index.html')


# ─── Main ────────────────────────────────────────────

if __name__ == '__main__':
    load_model()
    print("\n" + "=" * 50)
    if sentiment_pipeline:
        print("  SentiMeter Server — IndoBERT Ready")
    else:
        print("  SentiMeter Server — WARNING: Model failed to load!")
        print(f"  Error: {model_error}")
    print(f"  Open: http://localhost:7860")
    print("=" * 50 + "\n")
    app.run(host='0.0.0.0', port=7860, debug=False)