config.py

import os
from pathlib import Path

class Config:
    """Configuration class for Smart RAG API"""
    
    # Base directories
    BASE_DIR = Path(__file__).parent
    UPLOAD_DIR = BASE_DIR / "uploads"
    VECTOR_STORE_DIR = BASE_DIR / "vector_store"
    TEMP_DIR = BASE_DIR / "temp"
    
    # File processing
    MAX_FILE_SIZE = int(os.getenv("MAX_FILE_SIZE", 10 * 1024 * 1024))  # 10MB default
    ALLOWED_EXTENSIONS = {
        '.pdf', '.docx', '.txt', '.jpg', '.jpeg', '.png', '.csv', '.db'
    }
    
    # Text chunking
    CHUNK_SIZE = int(os.getenv("CHUNK_SIZE", 500))
    CHUNK_OVERLAP = int(os.getenv("CHUNK_OVERLAP", 50))
    
    # Hugging Face Models (Free)
    EMBEDDING_MODEL = os.getenv("EMBEDDING_MODEL", "sentence-transformers/all-MiniLM-L6-v2")
    
    # LLM Model options (choose based on performance needs)
    LLM_MODEL = os.getenv("LLM_MODEL", "google/flan-t5-base")
    # Alternative models:
    # "microsoft/DialoGPT-medium" - for conversational responses
    # "google/flan-t5-small" - faster, smaller model
    # "facebook/bart-large-cnn" - good for summarization
    
    # Vector search
    VECTOR_SEARCH_K = int(os.getenv("VECTOR_SEARCH_K", 5))
    SIMILARITY_THRESHOLD = float(os.getenv("SIMILARITY_THRESHOLD", 0.1))
    
    # OCR settings
    TESSERACT_CMD = os.getenv("TESSERACT_CMD", "/usr/bin/tesseract")
    OCR_LANGUAGE = os.getenv("OCR_LANGUAGE", "eng")
    
    # API settings
    API_HOST = os.getenv("API_HOST", "0.0.0.0")
    API_PORT = int(os.getenv("API_PORT", 7860))
    
    # Gradio settings
    GRADIO_SHARE = os.getenv("GRADIO_SHARE", "true").lower() == "true"
    GRADIO_DEBUG = os.getenv("GRADIO_DEBUG", "false").lower() == "true"
    
    # Model cache directory (for Hugging Face models)
    HF_CACHE_DIR = os.getenv("HF_HOME", BASE_DIR / "model_cache")
    
    # Performance settings
    TORCH_THREADS = int(os.getenv("TORCH_THREADS", 4))
    USE_GPU = os.getenv("USE_GPU", "false").lower() == "true"
    
    # Logging
    LOG_LEVEL = os.getenv("LOG_LEVEL", "INFO")
    
    @classmethod
    def setup_environment(cls):
        """Setup environment variables and directories"""
        
        # Set Hugging Face cache directory
        os.environ["HF_HOME"] = str(cls.HF_CACHE_DIR)
        os.environ["TRANSFORMERS_CACHE"] = str(cls.HF_CACHE_DIR)
        
        # Set PyTorch settings
        os.environ["OMP_NUM_THREADS"] = str(cls.TORCH_THREADS)
        os.environ["MKL_NUM_THREADS"] = str(cls.TORCH_THREADS)
        
        # Disable tokenizers parallelism warning
        os.environ["TOKENIZERS_PARALLELISM"] = "false"
        
        # Set Tesseract command if available
        if os.path.exists(cls.TESSERACT_CMD):
            import pytesseract
            pytesseract.pytesseract.tesseract_cmd = cls.TESSERACT_CMD

# File type configurations
FILE_TYPE_CONFIG = {
    '.pdf': {
        'icon': '📄',
        'description': 'PDF Document',
        'processor': 'pdf'
    },
    '.docx': {
        'icon': '📝',
        'description': 'Word Document',
        'processor': 'docx'
    },
    '.txt': {
        'icon': '📃',
        'description': 'Text File',
        'processor': 'text'
    },
    '.jpg': {
        'icon': '🖼️',
        'description': 'JPEG Image',
        'processor': 'image'
    },
    '.jpeg': {
        'icon': '🖼️',
        'description': 'JPEG Image',
        'processor': 'image'
    },
    '.png': {
        'icon': '🖼️',
        'description': 'PNG Image',
        'processor': 'image'
    },
    '.csv': {
        'icon': '📊',
        'description': 'CSV Data',
        'processor': 'csv'
    },
    '.db': {
        'icon': '🗄️',
        'description': 'SQLite Database',
        'processor': 'database'
    }
}

# Model configurations for different use cases
MODEL_CONFIGS = {
    'fast': {
        'embedding': 'sentence-transformers/all-MiniLM-L6-v2',
        'llm': 'google/flan-t5-small',
        'description': 'Fast processing, lower accuracy'
    },
    'balanced': {
        'embedding': 'sentence-transformers/all-MiniLM-L6-v2',
        'llm': 'google/flan-t5-base',
        'description': 'Balanced speed and accuracy'
    },
    'accurate': {
        'embedding': 'sentence-transformers/all-mpnet-base-v2',
        'llm': 'google/flan-t5-large',
        'description': 'Higher accuracy, slower processing'
    }
}

# Initialize configuration
Config.setup_environment()