Zero Config 🚀

Smart configuration management with layered overrides and type-aware defaults.

🎯 Core Concept

Zero Config provides layered configuration where each layer can override the previous one:

1. Application Defaults → 2. Environment Variables → 3. Environment Files

from zero_config import setup_environment, get_config

# 1. Define application defaults
default_config = {
    'openai_api_key': '',           # Will be overridden by env var
    'llm.temperature': 0.0,         # Will be overridden by .env file
    'database.host': 'localhost',   # Will stay as default
}

# 2. Set environment variable
# export OPENAI_API_KEY="sk-your-key-here"

# 3. Create .env.zero_config file
# llm.temperature=0.7
# database.port=5432

setup_environment(default_config=default_config)
config = get_config()

# Final configuration:
print(config.get('openai_api_key'))    # "sk-your-key-here" (from env var)
print(config.get('llm.temperature'))   # 0.7 (from .env file, converted to float)
print(config.get('database.host'))     # "localhost" (from defaults)
print(config.get('database.port'))     # 5432 (from .env file, new key as int)

🏗️ Why Layered Configuration?

• Defaults in Code: Your app defines the schema and sensible defaults
• Environment Variables: Perfect for deployment-specific overrides (Docker, CI/CD) - only loaded if defaults exist
• Environment Files: Great for local development and secrets management - loads all variables
• Type Safety: Environment strings are automatically converted to match your default types
• Multi-Level Support: Handle complex nested configurations with ease
• Package Conflict Prevention: Automatic protection when multiple packages use zero-config

🔧 Configuration Sources

🔄 Environment Variable Filtering (NEW in v0.1.5+)

Zero Config uses smart environment variable filtering to prevent configuration pollution:

• OS Environment Variables: Only loaded if they have corresponding defaults in your configuration
• Environment Files: Load ALL variables regardless of defaults

# Example: You have these OS environment variables set
# OPENAI_API_KEY=sk-test123        ← Will be loaded (has default)
# DATABASE_HOST=localhost          ← Will be loaded (has default)
# RANDOM_SYSTEM_VAR=some_value     ← Will be IGNORED (no default)
# PATH=/usr/bin:/bin               ← Will be IGNORED (no default)

default_config = {
    'openai_api_key': '',      # OS env OPENAI_API_KEY will override this
    'database.host': 'local',  # OS env DATABASE__HOST will override this
    # No default for RANDOM_SYSTEM_VAR, so it's ignored
}

setup_environment(default_config=default_config)
config = get_config()

print(config.get('openai_api_key'))    # "sk-test123" (from OS env)
print(config.get('database.host'))     # "localhost" (from OS env)
print(config.get('random_system_var')) # None (filtered out)

This prevents your application configuration from being polluted by unrelated system environment variables.

Environment Variables

# Uppercase env vars automatically override config keys
export OPENAI_API_KEY="sk-your-key-here"    # Becomes: openai_api_key
export DEBUG="true"                         # Becomes: debug (bool)
export MODELS='["gpt-4", "claude-3"]'       # JSON arrays for lists
export DATABASE_URL="host1,host2,host3"     # Strings with commas stay safe

# Section headers with double underscore (multi-level support)
export LLM__TEMPERATURE="0.7"               # Becomes: llm.temperature
export DATABASE__HOST="remote.db.com"       # Becomes: database.host

# Multi-level configuration (NEW in v0.1.5+)
export DATABASE__DEVELOPE__DB_URL="postgresql://localhost:5432/dev"  # Becomes: database.develope.db_url
export DATABASE__DEVELOPE__POOL_SIZE="10"   # Becomes: database.develope.pool_size (converted to int)
export LLM__OPENAI__API_KEY="sk-test123"    # Becomes: llm.openai.api_key
export LLM__ANTHROPIC__MODEL="claude-3"     # Becomes: llm.anthropic.model

Important: OS environment variables are only loaded if they have corresponding defaults in your configuration. This prevents random environment variables from polluting your config.

Environment Files

# .env.zero_config (default) or custom files
openai_api_key=sk-your-local-key
llm.temperature=0.7
database.port=5432
models=["gpt-4", "claude-3"]

# Multi-level configuration in env files (NEW in v0.1.5+)
DATABASE__DEVELOPE__DB_URL=postgresql://localhost:5432/dev
DATABASE__DEVELOPE__POOL_SIZE=10
LLM__OPENAI__API_KEY=sk-env-key
LLM__OPENAI__MODEL=gpt-4

Important: Environment files load ALL variables regardless of whether they have defaults. This allows you to add new configuration keys via env files.

Custom Environment Files

setup_environment(
    default_config=default_config,
    env_files="config/production.env"          # Single file
)

setup_environment(
    default_config=default_config,
    env_files=["base.env", "production.env"]   # Multiple files (later wins)
)

📁 Project Root Detection

Critical: Environment files are loaded relative to your project root.

# Auto-detection (looks for .git, pyproject.toml, setup.py, etc.)
setup_environment(default_config=config)

# Override via environment variable
# export PROJECT_ROOT="/path/to/project"
setup_environment(default_config=config)

Why it matters: Zero Config needs to know your project root to:

• Load .env.zero_config from the correct location
• Resolve relative paths in env_files parameter
• Provide accurate dynamic path helpers (config.data_path(), etc.)

🛠️ Advanced Features

Dynamic Path Helpers

config = get_config()

# Any directory name + '_path' works (Ruby on Rails style)
config.data_path('database.db')      # /project/data/database.db
config.logs_path('app.log')          # /project/logs/app.log
config.cache_path('session.json')    # /project/cache/session.json
config.models_path('gpt4.bin')       # /project/models/gpt4.bin

Section Configuration & Multi-Level Support

# Define sections with dot notation OR nested dictionaries (NEW in v0.1.5+)
default_config = {
    # Flat dot notation (traditional)
    'llm.models': ['gpt-4'],
    'llm.temperature': 0.0,
    'database.host': 'localhost',
    'database.port': 5432,

    # Nested dictionary format (NEW)
    'database': {
        'develope': {
            'db_url': 'sqlite:///dev.db',
            'pool_size': 5
        },
        'production': {
            'db_url': 'postgresql://prod.db.com:5432/prod'
        }
    },

    # Mixed formats work together!
    'llm.openai.api_key': 'default-key',  # Flat notation
    'cache': {                            # Nested format
        'enabled': True,
        'ttl': 3600
    }
}

config = get_config()

# Access sections (returns all keys within that section)
llm_config = config.get('llm')      # {'models': [...], 'temperature': 0.0, 'openai': {'api_key': '...'}}
db_config = config.get('database')  # {'host': 'localhost', 'port': 5432, 'develope': {...}, 'production': {...}}

# Access multi-level keys directly with dot notation
db_url = config.get('database.develope.db_url')     # 'sqlite:///dev.db'
api_key = config.get('llm.openai.api_key')          # 'default-key'

# Access subsections
develope_config = config.get('database.develope')   # {'db_url': '...', 'pool_size': 5}
openai_config = config.get('llm.openai')            # {'api_key': 'default-key'}

Type Conversion

Environment variables are automatically converted to match your default types:

• Numbers: "8000" → 8000 (int), "0.7" → 0.7 (float)
• Booleans: "true" → True, "false" → False
• Lists: '["a","b"]' → ['a','b'] (JSON only - comma strings stay safe)
• Strings: Always preserved as-is (safe for URLs, CSVs, etc.)

📦 Installation

pip install zero-config
# Or install specific version
pip install zero-config==0.1.6

🛡️ Package Conflict Prevention

Critical for libraries: Zero Config prevents configuration conflicts when both your main project and its dependencies use zero-config.

The Problem

Without protection, packages can accidentally overwrite your main project's configuration:

# ❌ Without protection (old behavior)
# Main project sets up config
setup_environment(default_config={'app_name': 'news_app', 'llm.api_key': 'main-key'})

# Package dependency overwrites everything!
setup_environment(default_config={'package_name': 'united_llm'})

# Main project's config is lost 😱
config = get_config()
print(config.get('app_name'))  # None - lost!

The Solution

Zero Config now automatically prevents this:

# ✅ With protection (new behavior)
# Main project initializes first
setup_environment(default_config={'app_name': 'news_app', 'llm.api_key': 'main-key'})

# Package dependency tries to initialize (safely ignored)
setup_environment(default_config={'package_name': 'united_llm'})  # ← Ignored!

# Main project's config is preserved 🎉
config = get_config()
print(config.get('app_name'))      # "news_app" (preserved)
print(config.get('package_name'))  # None (package config ignored)

How It Works

1. First Call Wins: The first setup_environment() call initializes the global configuration
2. Automatic Protection: Subsequent calls are automatically ignored with helpful logging
3. Shared Access: Packages can still access the main project's configuration
4. Override Available: Use force_reinit=True only for testing or special cases

Best Practices

For Main Applications:

# Initialize early in your main application
def main():
    setup_environment(default_config=your_app_config)
    # ... rest of your app

For Package Libraries:

# Packages should call setup_environment but expect it might be ignored
def initialize_package():
    # This will be ignored if main app already initialized
    setup_environment(default_config=package_defaults)

    # Always access config this way
    config = get_config()
    return config.get('llm')  # Access main app's LLM config

🔗 API Reference

# Setup
setup_environment(
    default_config={...},           # Your app's defaults (supports nested dicts and flat dot notation)
    env_files="custom.env",         # Optional: custom env file(s)
    force_reinit=False              # Force re-init (use with caution)
)

# Access
config = get_config()
config.get('key', default)         # Safe access with fallback
config['key']                      # Direct access (raises KeyError if missing)
config.get('llm')                  # Get all llm.* keys as dict (section access)
config.get('database.develope')    # Get subsection as dict (NEW: multi-level)
config.get('database.develope.db_url')  # Direct multi-level access (NEW)
config.to_dict()                   # Get all config as nested dict
config.to_flat_dict()              # Get all config as flat dict with dot notation keys

# Initialization status
is_initialized()                   # Check if already initialized
get_initialization_info()          # Get info about who initialized

# Dynamic paths (Ruby on Rails style)
config.data_path('file.db')        # /project/data/file.db
config.logs_path('app.log')        # /project/logs/app.log
config.any_name_path('file')       # /project/any_name/file

# Multi-level configuration examples (NEW in v0.1.5+)
config.get('database')             # Returns entire database section
config.get('database.develope')    # Returns develope subsection
config.get('llm.openai')          # Returns OpenAI LLM subsection

🔍 Debugging & Troubleshooting

Check Initialization Status

from zero_config import is_initialized, get_initialization_info

# Check if zero-config has been initialized
if is_initialized():
    print(f"Initialized by: {get_initialization_info()}")
else:
    print("Not yet initialized")

Common Issues

Issue: RuntimeError: Configuration not initialized

# ❌ Trying to get config before setup
config = get_config()  # Error!

# ✅ Always call setup_environment first
setup_environment()
config = get_config()  # Works!

Issue: Package config not working

# ❌ Package trying to override main config
setup_environment(default_config=package_config)  # Ignored!

# ✅ Package accessing main config
setup_environment(default_config=package_config)  # Ignored (expected)
config = get_config()  # Access main app's config
llm_config = config.get('llm')  # Get LLM section from main app

Issue: Need to reset for testing

# ✅ For testing only
from zero_config.config import _reset_for_testing

def test_something():
    _reset_for_testing()  # Reset global state
    setup_environment(test_config)
    # ... test code

Logging

Zero Config provides helpful logging. Enable it to see what's happening:

import logging
logging.basicConfig(level=logging.INFO)

setup_environment(default_config=config)
# INFO: Auto-detected project root: /path/to/project
# INFO: 🚀 Environment setup complete

📚 Documentation

For comprehensive guides and advanced usage, see the docs/ directory:

• Multi-Level Configuration - Complete guide to nested configurations ⭐ NEW
• Package Conflict Prevention - Prevent configuration conflicts
• Best Practices - Production deployment and architecture patterns
• Examples - Real-world usage examples
• Documentation Index - Complete documentation overview

🚀 Migration Guide

From v0.1.5 to v0.1.6+

No breaking changes! This is a documentation enhancement release. All v0.1.5 features remain unchanged.

• ✅ Enhanced documentation: Comprehensive guides and examples for all features
• ✅ Documentation index: New docs/README.md with complete overview
• ✅ Improved examples: Better multi-level configuration examples in README

From v0.1.4 to v0.1.5+

No breaking changes! Your existing code continues to work. New features:

• ✅ Multi-level configuration support: DATABASE__DEVELOPE__DB_URL → database.develope.db_url
• ✅ Enhanced DotDict implementation: Native dot notation support for nested access
• ✅ Improved environment variable filtering: OS env vars only loaded if defaults exist
• ✅ Mixed configuration formats: Support both nested dicts and flat dot notation in defaults

From v0.1.0 to v0.1.1+

No breaking changes! Your existing code continues to work. New features:

• ✅ Automatic package conflict prevention
• ✅ New is_initialized() and get_initialization_info() functions
• ✅ New force_reinit=True parameter for special cases

Upgrading Your Package

If you maintain a package that uses zero-config:

# Before (still works)
def initialize():
    setup_environment(default_config=defaults)

# After (recommended - more explicit)
def initialize():
    if not is_initialized():
        setup_environment(default_config=defaults)
    return get_config()