Function predict

webui/app.py:405–624

Perform prediction

()

@app.route('/api/predict', methods=['POST'])
def predict():
    """Perform prediction"""
    try:
        data = request.get_json()
        file_path = data.get('file_path')
        lookback = int(data.get('lookback', 400))
        pred_len = int(data.get('pred_len', 120))
        
        # Get prediction quality parameters
        temperature = float(data.get('temperature', 1.0))
        top_p = float(data.get('top_p', 0.9))
        sample_count = int(data.get('sample_count', 1))
        
        if not file_path:
            return jsonify({'error': 'File path cannot be empty'}), 400
        
        # Load data
        df, error = load_data_file(file_path)
        if error:
            return jsonify({'error': error}), 400
        
        if len(df) < lookback:
            return jsonify({'error': f'Insufficient data length, need at least {lookback} rows'}), 400
        
        # Perform prediction
        if MODEL_AVAILABLE and predictor is not None:
            try:
                # Use real Kronos model
                # Only use necessary columns: OHLCV, excluding amount
                required_cols = ['open', 'high', 'low', 'close']
                if 'volume' in df.columns:
                    required_cols.append('volume')
                
                # Process time period selection
                start_date = data.get('start_date')
                
                if start_date:
                    # Custom time period - fix logic: use data within selected window
                    start_dt = pd.to_datetime(start_date)
                    
                    # Find data after start time
                    mask = df['timestamps'] >= start_dt
                    time_range_df = df[mask]
                    
                    # Ensure sufficient data: lookback + pred_len
                    if len(time_range_df) < lookback + pred_len:
                        return jsonify({'error': f'Insufficient data from start time {start_dt.strftime("%Y-%m-%d %H:%M")}, need at least {lookback + pred_len} data points, currently only {len(time_range_df)} available'}), 400
                    
                    # Use first lookback data points within selected window for prediction
                    x_df = time_range_df.iloc[:lookback][required_cols]
                    x_timestamp = time_range_df.iloc[:lookback]['timestamps']
                    
                    # Use last pred_len data points within selected window as actual values
                    y_timestamp = time_range_df.iloc[lookback:lookback+pred_len]['timestamps']
                    
                    # Calculate actual time period length
                    start_timestamp = time_range_df['timestamps'].iloc[0]
                    end_timestamp = time_range_df['timestamps'].iloc[lookback+pred_len-1]