# License key
MARIADB_LICENSE_KEY=

# LLM Configuration
LLM_PROVIDER=gemini
LLM_MODEL=gemini-2.5-flash-lite

# OLLAMA Configuration
#LLM_PROVIDER=ollama
#OLLAMA_HOST=http://localhost:11434

# Embedding Configuration
EMBEDDING_PROVIDER=gemini
embedding_model=gemini-embedding-001

# API Keys
GEMINI_API_KEY=
#OPENAI_API_KEY=

# Database Configuration
DB_HOST=
DB_PORT=3306
DB_USER=
DB_PASSWORD=
DB_NAME=rag_db
DB_SSL_ENABLED=true

# Admin User Configuration
ADMIN_EMAIL=
ADMIN_PASSWORD=

# Security Keys
SECRET_KEY=
JWT_SECRET_KEY=
MCP_AUTH_SECRET_KEY=

# Server Ports
APP_HOST=127.0.0.1
APP_PORT=8000
MCP_HOST=127.0.0.1
MCP_PORT=8002
MCP_MARIADB_HOST=127.0.0.1
MCP_MARIADB_PORT=3306

# RAG Configuration
DOCUMENTS_TABLE=documents_DEMO
VDB_TABLE=vdb_tbl_DEMO

# Document Paths (for MCP file ingestion)
# Files in test_documents/ on host are available at /app/test_documents/ in container
MCP_DOCUMENT_BASE_PATH=/app
 

# ============================================================================
# RERANKING CONFIGURATION
# ============================================================================

# Enable reranking globally (default: false)
RERANKING_ENABLED=true

# Default reranking model
RERANKING_MODEL_TYPE=flashrank
RERANKING_MODEL_NAME=ms-marco-MiniLM-L-12-v2
#RERANKING_API_KEY=

# Retrieval strategy: retrieve (top_k * multiplier), then rerank to top_k
RERANKING_TOP_K_MULTIPLIER=2.0
RERANKING_DEFAULT_TOP_K=5

# Model caching
RERANKING_CACHE_MODELS=true
RERANKING_CACHE_MAX_SIZE=3

# Batch processing
RERANKING_BATCH_SIZE=32
RERANKING_ENABLE_BATCHING=true

# ===============================================================================
# FILE STORAGE CONFIGURATION
# ===============================================================================
# Configure where uploaded files are stored. Supports two backends:
#   - 'local': Store files on local filesystem (for on-prem deployments)
#   - 's3': Store files in AWS S3 bucket (for cloud deployments)
#
# Files are organized in per-user folders for security isolation:
#   Local: {LOCAL_STORAGE_PATH}/{user_id}/{filename}
#   S3: s3://{bucket}/{prefix}{user_id}/{filename}
#
# For Docker deployments with local storage, mount a host volume:
#   volumes:
#     - /var/rag-uploads:/app/uploaded_files
# ===============================================================================

# Storage type: 'local' or 's3' (default: local)
FILE_STORAGE_TYPE=local

LOCAL_STORAGE_PATH=./uploaded_files

# ===============================================================================
# S3 STORAGE CONFIGURATION (used when FILE_STORAGE_TYPE=s3)
# ===============================================================================

#MANAGED_S3_BUCKET=
#MANAGED_S3_REGION=us-east-1
#MANAGED_S3_PREFIX=

#AWS_ACCESS_KEY_ID=
#AWS_SECRET_ACCESS_KEY=
#AWS_DEFAULT_REGION=us-east-1
#AWS_CA_BUNDLE=

# ===============================================================================
# DOCUMENT PROCESSING CONFIGURATION (Layout-Aware Parsers)
# ===============================================================================
# Available processors (selected at /orchestrate/ingestion endpoint):
# - base: Basic text extraction (pdfplumber, python-docx) - always available
# - layout_aware_standard: Docling-based Markdown conversion (local processing)
# - layout_aware_advanced: LlamaParse cloud API (requires API key below)

# --- Layout-Aware Standard (Docling) Server Defaults ---
# OCR provider for scanned documents
DOCLING_OCR_PROVIDER=rapidocr

# Ray cluster address for distributed processing (set to specific address or 'auto')
DOCLING_RAY_HEAD_ADDRESS=auto

# Docling Ray HTTP service (local dev)
DOCLING_RAY_SERVICE_URL=http://localhost:8003
DOCLING_RAY_REQUEST_TIMEOUT=600

# --- Layout-Aware Advanced (LlamaParse) API Key ---
# Get your API key from: https://cloud.llamaindex.ai
#LLAMA_CLOUD_API_KEY=

# =============================================================================
# MCP CONFIGURATION  
# =============================================================================

MCP_READ_ONLY=false
MCP_ENABLE_AUTH=true
MCP_ENABLE_VECTOR_TOOLS=true
MCP_ENABLE_DATABASE_TOOLS=true
MCP_ENABLE_RAG_TOOLS=true
MCP_RAG_HEALTHCHECK_ENABLED=true
MCP_STANDALONE_MODE=false

# ============================================================================
# CELERY CONFIG
# ============================================================================
CELERY_ENABLED=true
CELERY_AUTOSCALE=8,2
CELERY_POOL=threads
CELERY_BROKER_URL=redis://localhost:6379/0
CELERY_RESULT_BACKEND=redis://localhost:6379/0

# Redis Configuration
REDIS_URL=redis://localhost:6379/0

# Optional: Celery Worker Settings
CELERY_WORKER_CONCURRENCY=4  # Number of worker processes
CELERY_WORKER_PREFETCH_MULTIPLIER=4
CELERY_TASK_TIME_LIMIT=3600  # 1 hour hard limit
CELERY_TASK_SOFT_TIME_LIMIT=3300  # 55 minutes soft limit

# CORS Configuration
CORS_ALLOWED_ORIGINS=http://localhost:3000,http://127.0.0.1:3000
CORS_ALLOW_CREDENTIALS=true
CORS_ALLOW_METHODS=GET,POST,PUT,DELETE,OPTIONS
CORS_ALLOW_HEADERS=Authorization,Content-Type
# Development Settings
DEBUG_MODE=false
LOG_SQL_QUERIES=false
MCP_LOG_LEVEL=INFO

# =============================================================================
# ENHANCED SECURITY SETTINGS  
# =============================================================================

# Rate Limiting (requests per minute)
RATE_LIMIT_REQUESTS=100
RATE_LIMIT_WINDOW=60

# File Upload Security
MAX_FILE_SIZE=200MB
ALLOWED_FILE_EXTENSIONS=.pdf,.txt,.docx,.md,.html,.csv,.xml
SCAN_UPLOADS_FOR_MALWARE=false
QUARANTINE_SUSPICIOUS_FILES=true

# Security Headers
SECURITY_HEADERS_ENABLED=true

# =============================================================================
# QUOTA SETTINGS  
# =============================================================================

# Max documents per user
QUOTA_MAX_DOCUMENTS=100        
# Max storage in MB
QUOTA_MAX_STORAGE_MB=1000      
# Max concurrent ingestion tasks
QUOTA_MAX_CONCURRENT_TASKS=10  
# Max files per sync operation
QUOTA_MAX_FILES_PER_SYNC=10    
# Requests per minute
QUOTA_RATE_LIMIT_RPM=60