.env.example

# Arctic Text2SQL Agent - Environment Configuration
# Copy this file to .env and fill in your values
# DO NOT commit .env to version control!

# =============================================================================
# HuggingFace Configuration
# =============================================================================
# Get your token from: https://huggingface.co/settings/tokens
HUGGINGFACE_TOKEN=your_huggingface_token_here

# Model configuration
TEXT2SQL_MODEL=Snowflake/Arctic-Text2SQL-R1-7B
MODEL_DEVICE=auto  # Options: auto, cuda, cpu, mps
ENABLE_8BIT_QUANTIZATION=false
ENABLE_4BIT_QUANTIZATION=false

# =============================================================================
# Database Configuration
# =============================================================================
# Primary database URL (supports PostgreSQL, MySQL, SQLite)
# PostgreSQL: postgresql://user:password@localhost:5432/dbname
# MySQL: mysql://user:password@localhost:3306/dbname
# SQLite: sqlite:///./data/text2sql.db
DATABASE_URL=sqlite:///./data/text2sql.db

# Connection pool settings
DB_POOL_SIZE=5
DB_MAX_OVERFLOW=10
DB_POOL_TIMEOUT=30

# =============================================================================
# API Configuration
# =============================================================================
API_HOST=0.0.0.0
API_PORT=8000
API_DEBUG=false

# CORS settings (comma-separated origins)
CORS_ORIGINS=http://localhost:3000,http://localhost:8080

# Rate limiting
RATE_LIMIT_PER_MINUTE=60
RATE_LIMIT_BURST=10

# =============================================================================
# Agent Configuration
# =============================================================================
# Maximum steps for agent reasoning loop
AGENT_MAX_STEPS=5
# Minimum confidence score to return results (0.0 - 1.0)
AGENT_MIN_CONFIDENCE=0.7
# Enable output validation
AGENT_ENABLE_VALIDATION=true
# Agent model backend (local or hf_inference)
AGENT_MODEL_BACKEND=local
# Inference Providers settings (used when AGENT_MODEL_BACKEND=hf_inference)
AGENT_INFERENCE_PROVIDER=hf-inference
AGENT_INFERENCE_TIMEOUT=120
# AGENT_INFERENCE_BASE_URL=https://your-endpoint.example.com
# AGENT_INFERENCE_BILL_TO=your-org
# AGENT_INFERENCE_MAX_TOKENS=512
# AGENT_INFERENCE_TEMPERATURE=0.1
# AGENT_INFERENCE_TOP_P=0.95

# =============================================================================
# Security Configuration (Phase 2.2: Security Implementation)
# =============================================================================
# CRITICAL: Change these values in production!

# Secret key for JWT tokens
# Generate a secure key with: openssl rand -hex 32
# Or with Python: python -c "import secrets; print(secrets.token_hex(32))"
SECRET_KEY=your-secret-key-change-in-production-MUST-BE-CHANGED

# JWT settings
JWT_ALGORITHM=HS256
JWT_ACCESS_TOKEN_EXPIRE_MINUTES=30

# Authentication toggles
AUTH_ENABLED=true
JWT_AUTH_ENABLED=true
API_KEY_AUTH_ENABLED=true

# API Key Authentication
# Format: key:scope1|scope2, key2:scope1|scope2
# For production, use separate API keys per client/service
API_KEYS=your-api-key-change-in-production:read|write

# Default scopes for keys without an explicit scope list
API_KEY_SCOPES=read

# Optional user credentials for issuing JWTs
# Format: username:password:scope1|scope2;username2:password:scope1|scope2
AUTH_USERS=admin:change_me:admin|write|read

# JWT claims
JWT_SCOPES_CLAIM=scopes
JWT_ROLE_CLAIM=role

# Mutation scopes required for write/management endpoints
MUTATION_SCOPES=write,admin

# Rate limiting storage (Redis recommended in production)
# RATE_LIMIT_STORAGE_URL=redis://localhost:6379/1
RATE_LIMIT_HEADERS_ENABLED=true

# =============================================================================
# Logging Configuration
# =============================================================================
# Log level: DEBUG, INFO, WARNING, ERROR, CRITICAL
LOG_LEVEL=INFO
# Log format: json, console
LOG_FORMAT=json
# Enable request logging
LOG_REQUESTS=true

# =============================================================================
# Monitoring Configuration
# =============================================================================
# Enable Prometheus metrics
ENABLE_METRICS=true
METRICS_PORT=9090

# =============================================================================
# Cache Configuration (Issue #8: Performance Optimization)
# =============================================================================
# Redis URL for caching (optional - falls back to in-memory if not set)
# REDIS_URL=redis://localhost:6379/0

# Default cache TTL in seconds
CACHE_TTL=3600

# TTL for different cache types
CACHE_QUERY_TTL=3600        # Query result cache (1 hour)
CACHE_MODEL_TTL=7200        # Model output cache (2 hours)
CACHE_SCHEMA_TTL=86400      # Schema cache (24 hours)

# Enable/disable caching globally
CACHE_ENABLED=true

# Maximum in-memory cache entries (when Redis unavailable)
CACHE_MAX_MEMORY_ENTRIES=1000

# =============================================================================
# Few-Shot Learning (Issue #16)
# =============================================================================
FEWSHOT_ENABLED=true
FEWSHOT_MAX_EXAMPLES=3
FEWSHOT_MIN_SIMILARITY=0.2
FEWSHOT_EMBEDDING_STRATEGY=hash  # Options: hash, model
FEWSHOT_EMBEDDING_DIM=256
FEWSHOT_EMBEDDING_MAX_LENGTH=256
FEWSHOT_INCLUDE_DEFAULT_EXAMPLES=false
FEWSHOT_VERIFIED_ONLY=true
FEWSHOT_USE_ON_FIRST_ATTEMPT=true
FEWSHOT_USE_ON_RETRY=true

# =============================================================================
# Feedback Collection (Issue #16)
# =============================================================================
FEEDBACK_ENABLED=true
FEEDBACK_AUTO_PROMOTE_TO_EXAMPLES=false
FEEDBACK_MIN_RATING_FOR_PROMOTION=4

# =============================================================================
# Fine-Tuning Pipeline (Issue #16)
# =============================================================================
FINETUNE_ENABLED=false
FINETUNE_OUTPUT_DIR=./data/fine_tuning
FINETUNE_DATASET_NAME=text2sql_finetune
FINETUNE_MAX_EXAMPLES=5000
FINETUNE_INCLUDE_FEEDBACK=true
FINETUNE_TRAIN_EPOCHS=3
FINETUNE_BATCH_SIZE=2
FINETUNE_LEARNING_RATE=2e-5
FINETUNE_MAX_SEQ_LENGTH=1024

# =============================================================================
# Model Versioning (Issue #16)
# =============================================================================
MODEL_VERSION_ENABLED=true
MODEL_VERSION_ACTIVE_VERSION_ID=