Delete config.py

config.py

@@ -1,317 +0,0 @@
-"""
-Configuration file for the Medical RAG System.
-Centralized settings for easy customization without modifying core files.
-"""
-
-import os
-from pathlib import Path
-
-# ===========================
-# PATHS & DIRECTORIES
-# ===========================
-
-# Project root directory
-PROJECT_ROOT = Path(__file__).parent
-
-# Vector database location
-VECTOR_DB_PATH = os.getenv("VECTOR_DB_PATH", "./MedQuAD_db")
-
-# Model cache directory (HuggingFace cache)
-HF_CACHE_DIR = os.getenv("HF_HOME", "./models")
-
-# ===========================
-# MODEL CONFIGURATION
-# ===========================
-
-# Embedding Model (for query and document encoding)
-EMBEDDING_MODEL_ID = "sentence-transformers/all-MiniLM-L6-v2"
-EMBEDDING_MODEL_DEVICE = "cuda"  # or "cpu"
-
-# Query Rewriter Model
-REWRITER_MODEL_ID = "google/flan-t5-small"
-REWRITER_MAX_LENGTH = 64
-REWRITER_TEMPERATURE = 0.3
-REWRITER_REPETITION_PENALTY = 1.3
-
-# Re-ranker Model (MonoT5)
-RERANKER_MODEL_ID = "castorini/monot5-base-msmarco"
-RERANKER_DEVICE = "cuda"  # or "cpu"
-
-# Fine-tuned Generator Model
-FINETUNED_MODEL_ID = os.getenv(
-    "FINETUNED_MODEL_ID",
-    "amiraghhh/fine-tuned-flan-t5-small"
-)
-
-# Baseline FLAN-T5 (for prompt building)
-BASELINE_MODEL_ID = "google/flan-t5-small"
-
-# ===========================
-# RETRIEVAL CONFIGURATION
-# ===========================
-
-# Default number of context chunks to retrieve
-DEFAULT_TOP_K = 3
-
-# Maximum number of chunks to retrieve before ranking
-MAX_RETRIEVE = 10
-
-# Vector store collection name
-VECTOR_STORE_COLLECTION = "medical_rag"
-
-# Similarity threshold for filtering (0-1, lower is more strict)
-SIMILARITY_THRESHOLD = 0.1
-
-# ===========================
-# GENERATION CONFIGURATION
-# ===========================
-
-# Maximum tokens in generated answer
-GENERATION_MAX_TOKENS = 70
-
-# Number of beams for beam search
-GENERATION_NUM_BEAMS = 3
-
-# Repetition penalty (> 1.0 reduces repetition)
-GENERATION_REPETITION_PENALTY = 1.4
-
-# Do sampling (True) or greedy (False)
-GENERATION_DO_SAMPLE = False
-
-# Temperature for sampling (ignored if do_sample=False)
-GENERATION_TEMPERATURE = 0.7
-
-# ===========================
-# PROMPT CONFIGURATION
-# ===========================
-
-# Maximum tokens for the full prompt
-PROMPT_MAX_TOKENS = 512
-
-# Prompt template - can be customized
-PROMPT_INSTRUCTION = "Medical Context:\n"
-PROMPT_QUERY_FOOTER = "\nQ: {query}\nA:"
-
-# Emergency keywords that should trigger a warning
-EMERGENCY_KEYWORDS = [
-    "emergency", "severe pain", "bleeding", "blind",
-    "lose consciousness", "pass out", "call 911", "911",
-    "critical", "life-threatening"
-]
-
-EMERGENCY_RESPONSE = """I am an AI and cannot provide medical advice for emergencies.
-PLEASE CONTACT EMERGENCY SERVICES OR A MEDICAL PROFESSIONAL IMMEDIATELY."""
-
-# ===========================
-# CONFIDENCE SCORING
-# ===========================
-
-# Confidence thresholds
-CONFIDENCE_LOW_THRESHOLD = 40  # %
-CONFIDENCE_HIGH_THRESHOLD = 70  # %
-
-# How to calculate confidence (based on retrieval distances)
-# distance_range: 0 (identical) to 2 (very different)
-CONFIDENCE_FORMULA = "max(0, min(100, (1 - avg_distance) * 100))"
-
-# ===========================
-# WEB INTERFACE CONFIGURATION
-# ===========================
-
-# Gradio server settings
-GRADIO_SERVER_NAME = "0.0.0.0"
-GRADIO_SERVER_PORT = 7860
-GRADIO_SHARE = False
-GRADIO_DEBUG = False
-GRADIO_SHOW_ERROR = True
-GRADIO_SHOW_TIPS = True
-
-# Gradio theme
-GRADIO_THEME = "soft"
-
-# Page title
-PAGE_TITLE = "Medical Q&A System"
-
-# Example questions to display
-EXAMPLE_QUESTIONS = [
-    "What are the symptoms of type 2 diabetes?",
-    "How is hypertension treated?",
-    "What causes migraines?",
-    "What are the risk factors for heart disease?",
-    "How do I manage chronic pain?",
-    "What is asthma?",
-    "When should I see a doctor for fever?",
-    "What are the causes of back pain?"
-]
-
-# ===========================
-# PERFORMANCE & OPTIMIZATION
-# ===========================
-
-# Batch size for embedding
-EMBEDDING_BATCH_SIZE = 64
-
-# Whether to normalize embeddings
-EMBEDDING_NORMALIZE = True
-
-# Convert embeddings to numpy (True) or keep as tensors (False)
-EMBEDDING_CONVERT_TO_NUMPY = True
-
-# Cache frequently used embeddings
-ENABLE_CACHE = True
-CACHE_SIZE = 1000  # number of queries to cache
-
-# ===========================
-# LOGGING CONFIGURATION
-# ===========================
-
-LOG_LEVEL = "INFO"  # DEBUG, INFO, WARNING, ERROR, CRITICAL
-LOG_FILE = "rag_system.log"
-LOG_FORMAT = "%(asctime)s - %(name)s - %(levelname)s - %(message)s"
-
-# ===========================
-# RANDOM SEEDS (for reproducibility)
-# ===========================
-
-RANDOM_SEED = 1
-NUMPY_SEED = 1
-TORCH_SEED = 1
-CUDA_SEED = 1
-
-# ===========================
-# DATA PREPROCESSING
-# ===========================
-
-# Text cleaning options
-CLEAN_LOWERCASE = True
-CLEAN_REMOVE_URLS = True
-CLEAN_REMOVE_EMAILS = True
-CLEAN_REMOVE_PHONES = True
-CLEAN_REMOVE_ADDRESSES = True
-
-# Chunking options
-CHUNK_SIZE = 350  # tokens
-CHUNK_OVERLAP = 50  # tokens
-
-# ===========================
-# RATE LIMITING & SECURITY
-# ===========================
-
-# Enable rate limiting
-ENABLE_RATE_LIMIT = False
-MAX_REQUESTS_PER_MINUTE = 30
-MAX_REQUEST_LENGTH = 1000  # max question length in characters
-
-# ===========================
-# DEPLOYMENT SETTINGS
-# ===========================
-
-# Environment type
-ENVIRONMENT = os.getenv("ENVIRONMENT", "development")  # development, staging, production
-
-# Enable analytics
-ENABLE_ANALYTICS = False
-
-# API key (for authentication if needed)
-API_KEY = os.getenv("API_KEY", None)
-
-# ===========================
-# VERSION & METADATA
-# ===========================
-
-APP_VERSION = "1.0.0"
-APP_NAME = "Medical RAG System"
-APP_DESCRIPTION = "Retrieval-Augmented Generation for medical Q&A"
-APP_AUTHOR = "Your Name"
-APP_LICENSE = "MIT"
-
-# ===========================
-# HELPER FUNCTIONS
-# ===========================
-
-def get_model_config(model_type):
-    """Get configuration for a specific model type.
-    
-    Args:
-        model_type (str): Type of model ('embedding', 'rewriter', 'reranker', 'generator')
-        
-    Returns:
-        dict: Configuration dictionary
-    """
-    configs = {
-        'embedding': {
-            'model_id': EMBEDDING_MODEL_ID,
-            'device': EMBEDDING_MODEL_DEVICE,
-            'batch_size': EMBEDDING_BATCH_SIZE,
-            'normalize': EMBEDDING_NORMALIZE,
-        },
-        'rewriter': {
-            'model_id': REWRITER_MODEL_ID,
-            'max_length': REWRITER_MAX_LENGTH,
-            'temperature': REWRITER_TEMPERATURE,
-            'repetition_penalty': REWRITER_REPETITION_PENALTY,
-        },
-        'reranker': {
-            'model_id': RERANKER_MODEL_ID,
-            'device': RERANKER_DEVICE,
-        },
-        'generator': {
-            'model_id': FINETUNED_MODEL_ID,
-            'max_tokens': GENERATION_MAX_TOKENS,
-            'num_beams': GENERATION_NUM_BEAMS,
-            'do_sample': GENERATION_DO_SAMPLE,
-            'temperature': GENERATION_TEMPERATURE,
-            'repetition_penalty': GENERATION_REPETITION_PENALTY,
-        }
-    }
-    
-    return configs.get(model_type, {})
-
-
-def is_production():
-    """Check if running in production environment."""
-    return ENVIRONMENT == "production"
-
-
-def is_emergency_query(query):
-    """Check if query contains emergency keywords."""
-    query_lower = query.lower()
-    return any(keyword in query_lower for keyword in EMERGENCY_KEYWORDS)
-
-
-# ===========================
-# VALIDATE CONFIGURATION
-# ===========================
-
-def validate_config():
-    """Validate critical configuration settings."""
-    errors = []
-    
-    # Check paths
-    if not os.path.exists(VECTOR_DB_PATH):
-        errors.append(f"Vector database path not found: {VECTOR_DB_PATH}")
-    
-    # Check model IDs
-    if not FINETUNED_MODEL_ID:
-        errors.append("FINETUNED_MODEL_ID not set")
-    
-    # Check thresholds
-    if not (0 <= CONFIDENCE_LOW_THRESHOLD <= 100):
-        errors.append("CONFIDENCE_LOW_THRESHOLD must be between 0 and 100")
-    
-    if errors:
-        print("Configuration validation errors:")
-        for error in errors:
-            print(f"  - {error}")
-        return False
-    
-    return True
-
-
-if __name__ == "__main__":
-    # Test configuration
-    print("Configuration loaded successfully!")
-    print(f"Environment: {ENVIRONMENT}")
-    print(f"Vector DB: {VECTOR_DB_PATH}")
-    print(f"Fine-tuned Model: {FINETUNED_MODEL_ID}")
-    print(f"Validation: {'PASSED' if validate_config() else 'FAILED'}")