.env.example

# WronAI Environment Configuration
# Copy this file to .env and fill in your values

# =============================================================================
# Logging & Monitoring
# =============================================================================

# Weights & Biases
WANDB_API_KEY=your_wandb_api_key_here
WANDB_PROJECT=wronai
WANDB_ENTITY=your_wandb_entity

# Logging level (DEBUG, INFO, WARNING, ERROR)
LOG_LEVEL=INFO
LOG_FILE=logs/wronai.log

# =============================================================================
# Model Configuration
# =============================================================================

# Default model settings
DEFAULT_MODEL_NAME=mistralai/Mistral-7B-v0.1
DEFAULT_MAX_LENGTH=2048
DEFAULT_BATCH_SIZE=1

# HuggingFace Hub
HUGGINGFACE_HUB_TOKEN=your_hf_token_here
HF_DATASETS_CACHE=/path/to/hf/cache

# =============================================================================
# Training Configuration
# =============================================================================

# CUDA settings
CUDA_VISIBLE_DEVICES=0
CUDA_LAUNCH_BLOCKING=1

# Training paths
CHECKPOINTS_DIR=./checkpoints
DATA_DIR=./data
LOGS_DIR=./logs

# Training parameters
LEARNING_RATE=2e-4
WARMUP_RATIO=0.03
GRADIENT_ACCUMULATION_STEPS=16
SAVE_STEPS=500
EVAL_STEPS=500

# =============================================================================
# Data Configuration
# =============================================================================

# Polish data sources
POLISH_WIKIPEDIA_PATH=data/processed/polish_wikipedia.jsonl
OSCAR_POLISH_PATH=data/processed/oscar_polish.jsonl
INSTRUCTION_DATA_PATH=data/processed/polish_instructions.json

# Data processing
MAX_DATASET_SIZE=100000
PREPROCESSING_WORKERS=4

# =============================================================================
# Hardware Configuration
# =============================================================================

# Memory settings
MAX_MEMORY_GB=7
ENABLE_GRADIENT_CHECKPOINTING=true
ENABLE_MIXED_PRECISION=true

# Quantization
ENABLE_QUANTIZATION=true
QUANTIZATION_BITS=4
QUANTIZATION_TYPE=nf4

# LoRA settings
ENABLE_LORA=true
LORA_R=16
LORA_ALPHA=32
LORA_DROPOUT=0.1

# =============================================================================
# API Configuration
# =============================================================================

# API server settings
API_HOST=0.0.0.0
API_PORT=8000
API_WORKERS=1

# Rate limiting
RATE_LIMIT_REQUESTS=100
RATE_LIMIT_WINDOW=3600

# =============================================================================
# Database Configuration (if using)
# =============================================================================

# Database URL for conversation storage
DATABASE_URL=sqlite:///wronai_conversations.db

# Redis for caching (optional)
REDIS_URL=redis://localhost:6379/0

# =============================================================================
# Cloud Configuration
# =============================================================================

# AWS S3 for model storage (optional)
AWS_ACCESS_KEY_ID=your_aws_access_key
AWS_SECRET_ACCESS_KEY=your_aws_secret_key
AWS_DEFAULT_REGION=eu-central-1
S3_BUCKET_NAME=wronai-models

# Google Cloud Storage (optional)
GOOGLE_APPLICATION_CREDENTIALS=/path/to/gcp/credentials.json
GCS_BUCKET_NAME=wronai-models

# =============================================================================
# Security Configuration
# =============================================================================

# API security
API_SECRET_KEY=your_super_secret_key_here
JWT_SECRET_KEY=your_jwt_secret_key

# CORS settings
CORS_ORIGINS=["http://localhost:3000", "http://localhost:8080"]

# =============================================================================
# Development Configuration
# =============================================================================

# Development mode
DEBUG=false
DEVELOPMENT_MODE=false

# Testing
PYTEST_WORKERS=auto
TEST_DATA_DIR=tests/fixtures

# Profiling
ENABLE_PROFILING=false
PROFILE_OUTPUT_DIR=profiles/

# =============================================================================
# Docker Configuration
# =============================================================================

# Docker build args
DOCKER_PYTHON_VERSION=3.10
DOCKER_CUDA_VERSION=11.8
DOCKER_UBUNTU_VERSION=22.04

# Container settings
CONTAINER_NAME=wronai
CONTAINER_RESTART_POLICY=unless-stopped

# =============================================================================
# Polish Language Specific
# =============================================================================

# Language model settings
POLISH_VOCAB_SIZE=32000
ENABLE_MORPHOLOGICAL_ANALYSIS=true
POLISH_STOPWORDS_PATH=wronai/data/polish/stopwords.txt

# Content filtering
ENABLE_CONTENT_FILTER=true
TOXIC_THRESHOLD=0.8
ENABLE_BIAS_DETECTION=true

# =============================================================================
# Performance Tuning
# =============================================================================

# Optimization
TORCH_COMPILE=false
FLASH_ATTENTION=false
TORCH_BACKENDS=inductor

# CPU settings
OMP_NUM_THREADS=8
MKL_NUM_THREADS=8

# =============================================================================
# Monitoring & Alerting
# =============================================================================

# Prometheus metrics
ENABLE_PROMETHEUS=false
PROMETHEUS_PORT=9090

# Health checks
HEALTH_CHECK_INTERVAL=30
HEALTH_CHECK_TIMEOUT=10

# Alerts
SLACK_WEBHOOK_URL=your_slack_webhook_url
EMAIL_ALERTS=admin@example.com