Getting Started¶

What is ReausoConfig?¶

ReausoConfig provides a simple way to load configuration files (YAML, JSON, or TOML) and instantiate Python objects from them. Unlike heavier frameworks, your application code doesn't need to know about ReausoConfig—only the startup/registration code does.

After instantiation, you get pure Python objects with no framework dependency.

Installation¶

Install ReausoConfig from GitHub using pip:

pip install git+https://github.com/reauso/ReausoConfig.git

Development Setup¶

For contributing or running tests, use pixi:

pixi install
pixi run test

Quick Start¶

1. Define your classes¶

from dataclasses import dataclass

@dataclass
class ModelConfig:
    hidden_size: int
    dropout: float = 0.1

2. Create a config file¶

YAML (config.yaml):

_target_: model
hidden_size: 256
dropout: 0.2

JSON (config.json):

{
    "_target_": "model",
    "hidden_size": 256,
    "dropout": 0.2
}

TOML (config.toml):

_target_ = "model"
hidden_size = 256
dropout = 0.2

The _target_ key maps to a registered class name.

3. Register and instantiate¶

import rconfig as rc

# Register your class
rc.register(name="model", target=ModelConfig)

# Instantiate the object (validates automatically)
# String paths work directly - no need for Path()
model = rc.instantiate(path="config.yaml", expected_type=ModelConfig)
print(model.hidden_size)  # 256
print(model.dropout)      # 0.2

# Optional: validate without instantiating (dry-run)
result = rc.validate(path="config.yaml")
if not result.valid:
    for error in result.errors:
        print(error)

Supported File Formats¶

ReausoConfig has built-in support for three configuration formats:

Format	Extensions	Notes
YAML	`.yaml`, `.yml`	Primary format, preserves comments with ruamel.yaml
JSON	`.json`	Standard JSON, good for programmatic generation
TOML	`.toml`	Python 3.11+ (uses stdlib tomllib)

The loader is selected automatically based on file extension:

# All work the same way - string paths work directly
model = rc.instantiate(path="config.yaml")
model = rc.instantiate(path="config.json")
model = rc.instantiate(path="config.toml")

# Path objects also work
from pathlib import Path
model = rc.instantiate(path=Path("config.yaml"))

Cross-format composition: You can mix formats with _ref_:

# trainer.yaml
_target_: trainer
model:
  _ref_: ./models/resnet.json  # Load JSON from YAML
settings:
  _ref_: ./settings.toml       # Load TOML from YAML