config.py

import os
import logging
from dotenv import load_dotenv

load_dotenv()

# --- Logging Setup ---
logger = logging.getLogger(__name__)
if not logger.handlers:
    logging.basicConfig(
        level=logging.INFO,
        format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
    )

_MODULE_BASE_DIR = os.path.dirname(os.path.abspath(__file__))

# API Authentication for n8n (Basic Auth)
API_USERNAME = os.getenv("API_USERNAME", "admin")
API_PASSWORD = os.getenv("API_PASSWORD", "password")

# Admin fallback credentials for dashboard (used when users.csv is missing or unavailable)
ADMIN_USERNAME = os.getenv('FLASK_ADMIN_USERNAME', 'admin')
ADMIN_PASSWORD = os.getenv('FLASK_ADMIN_PASSWORD', '1234')

# URL Fetcher configs (Generalized from Rentry)
URL_FETCH_ENABLED = os.getenv("URL_FETCH_ENABLED", "False").lower() == "true"
EXTERNAL_URL = os.getenv("EXTERNAL_URL", os.getenv("RENTRY_URL", ""))
URL_UPDATE_PERIOD_MINUTES = int(os.getenv("URL_UPDATE_PERIOD_MINUTES", os.getenv("RENTRY_UPDATE_PERIOD_MINUTES", "60")))

# CSV Configuration
RAG_CSV_MAX_RESULTS = int(os.getenv("RAG_CSV_MAX_RESULTS", "5"))
RAG_CSV_CONFIDENCE_THRESHOLD = float(os.getenv("RAG_CSV_CONFIDENCE_THRESHOLD", "0.5"))
# Separate CSV retrieval controls. CSV is row-by-row, so it usually needs a wider candidate pool.
RAG_CSV_INITIAL_FETCH_K = int(os.getenv("RAG_CSV_INITIAL_FETCH_K", "50"))
RAG_CSV_FINAL_K = int(os.getenv("RAG_CSV_FINAL_K", str(RAG_CSV_MAX_RESULTS)))
# Number of CSV rows stored in each vector-search chunk. Use 1 for strict row-by-row CSV indexing.
RAG_CSV_ROWS_PER_CHUNK = max(1, int(os.getenv("RAG_CSV_ROWS_PER_CHUNK", "1")))

RAG_FAISS_INDEX_SUBDIR_NAME = "faiss_index"
RAG_STORAGE_PARENT_DIR = os.getenv("RAG_STORAGE_DIR", os.path.join(_MODULE_BASE_DIR, "faiss_storage"))
RAG_SOURCES_DIR = os.getenv("SOURCES_DIR", os.path.join(_MODULE_BASE_DIR, "sources"))
RAG_CHUNKED_SOURCES_FILENAME = "pre_chunked_sources.json"

# App FAQ Configuration
# Keep app-only FAQ content here (default: sources/app_faqs/faqs.csv).
# It has separate FAISS storage, so it does not mix into the main knowledgebase.
FAQ_SOURCES_DIR = os.getenv("FAQ_SOURCES_DIR", os.path.join(RAG_SOURCES_DIR, "app_faqs"))
FAQ_STORAGE_PARENT_DIR = os.getenv("FAQ_STORAGE_DIR", os.path.join(_MODULE_BASE_DIR, "faiss_storage_app_faqs"))
FAQ_CSV_FILENAME = os.getenv("FAQ_CSV_FILENAME", "faqs.csv")
FAQ_INITIAL_FETCH_K = int(os.getenv("FAQ_INITIAL_FETCH_K", "50"))
FAQ_FINAL_K = int(os.getenv("FAQ_FINAL_K", "5"))
FAQ_CONFIDENCE_THRESHOLD = float(os.getenv("FAQ_CONFIDENCE_THRESHOLD", str(RAG_CSV_CONFIDENCE_THRESHOLD)))

os.makedirs(RAG_SOURCES_DIR, exist_ok=True)
os.makedirs(RAG_STORAGE_PARENT_DIR, exist_ok=True)
os.makedirs(FAQ_SOURCES_DIR, exist_ok=True)
os.makedirs(FAQ_STORAGE_PARENT_DIR, exist_ok=True)

# Embedding and model configuration
RAG_EMBEDDING_MODEL_NAME = os.getenv("RAG_EMBEDDING_MODEL", "BAAI/bge-small-en")
RAG_EMBEDDING_USE_GPU = os.getenv("RAG_EMBEDDING_GPU", "False").lower() == "true"
RAG_LOAD_INDEX_ON_STARTUP = os.getenv("RAG_LOAD_INDEX", "True").lower() == "true"

# Retrieval Settings
RAG_INITIAL_FETCH_K = int(os.getenv("RAG_INITIAL_FETCH_K", 20))
RAG_RERANKER_K = int(os.getenv("RAG_RERANKER_K", 5))
RAG_MAX_FILES_FOR_INCREMENTAL = int(os.getenv("RAG_MAX_FILES_FOR_INCREMENTAL", "50"))

# Chunk configuration
RAG_CHUNK_SIZE = int(os.getenv("RAG_CHUNK_SIZE", 2000))
RAG_CHUNK_OVERLAP = int(os.getenv("RAG_CHUNK_OVERLAP", 150))

# Reranker configuration
RAG_RERANKER_MODEL_NAME = os.getenv("RAG_RERANKER_MODEL", "jinaai/jina-reranker-v2-base-multilingual")
RAG_RERANKER_ENABLED = os.getenv("RAG_RERANKER_ENABLED", "True").lower() == "true"

# GDrive configuration for RAG sources
GDRIVE_SOURCES_ENABLED = os.getenv("GDRIVE_SOURCES_ENABLED", "False").lower() == "true"
GDRIVE_FOLDER_ID_OR_URL = os.getenv("GDRIVE_FOLDER_URL")

# GDrive configuration for downloading a pre-built FAISS index
GDRIVE_INDEX_ENABLED = os.getenv("GDRIVE_INDEX_ENABLED", "False").lower() == "true"
GDRIVE_INDEX_ID_OR_URL = os.getenv("GDRIVE_INDEX_URL")

# GDrive configuration for downloading users.csv
GDRIVE_USERS_CSV_ENABLED = os.getenv("GDRIVE_USERS_CSV_ENABLED", "False").lower() == "true"
GDRIVE_USERS_CSV_ID_OR_URL = os.getenv("GDRIVE_USERS_CSV_URL")

RAG_DETAILED_LOGGING = os.getenv("RAG_DETAILED_LOGGING", "True").lower() == "true"

logger.info(f"RAG Config Loaded - Chunk Size: {RAG_CHUNK_SIZE}, Chunk Overlap: {RAG_CHUNK_OVERLAP}")
logger.info(f"Embedding Model: {RAG_EMBEDDING_MODEL_NAME}")
logger.info(f"Reranker Model: {RAG_RERANKER_MODEL_NAME}")
logger.info(f"Retrieval Pipeline: Initial Fetch K={RAG_INITIAL_FETCH_K}, Reranker Final K={RAG_RERANKER_K}")
logger.info(f"CSV Filters: Max Results={RAG_CSV_MAX_RESULTS}, Threshold={RAG_CSV_CONFIDENCE_THRESHOLD}")
logger.info(f"CSV Retrieval: Initial Fetch K={RAG_CSV_INITIAL_FETCH_K}, Final K={RAG_CSV_FINAL_K}")
logger.info(f"App FAQ Source: {os.path.join(FAQ_SOURCES_DIR, FAQ_CSV_FILENAME)}")
logger.info(f"App FAQ Retrieval: Initial Fetch K={FAQ_INITIAL_FETCH_K}, Final K={FAQ_FINAL_K}")
logger.info(f"App FAQ Filter: Threshold={FAQ_CONFIDENCE_THRESHOLD}")
logger.info(f"CSV Chunking: Rows Per Chunk={RAG_CSV_ROWS_PER_CHUNK}")
logger.info(f"URL Fetching: {'ENABLED' if URL_FETCH_ENABLED else 'DISABLED'}")
logger.info(f"Detailed Logging: {'ENABLED' if RAG_DETAILED_LOGGING else 'DISABLED'}")
logger.info(f"GDrive Sources Download: {'ENABLED' if GDRIVE_SOURCES_ENABLED else 'DISABLED'}")
logger.info(f"GDrive Pre-built Index Download: {'ENABLED' if GDRIVE_INDEX_ENABLED else 'DISABLED'}")
logger.info(f"GDrive users.csv Download: {'ENABLED' if GDRIVE_USERS_CSV_ENABLED else 'DISABLED'}")