Nightjar

A config dispatch library. Supports dataclasses, Pydantic models, and custom converters.

Install

uv add nightjar

Usage

from dataclasses import dataclass
from nightjar import dispatch, register


@dataclass
class StorageConfig:
    pass


@dataclass
class LocalConfig(StorageConfig):
    kind: str = "local"
    path: str = "."


@dataclass
class MemoryConfig(StorageConfig):
    kind: str = "memory"
    capacity: int = 100


@register(kind="local")
@dataclass
class LocalStorage:
    config: LocalConfig


@register(kind="memory")
@dataclass
class MemoryStorage:
    config: MemoryConfig


storage = dispatch(StorageConfig, {"kind": "memory", "capacity": "20"})
assert isinstance(storage, MemoryStorage)
assert storage.config.capacity == 20

local = dispatch(StorageConfig, {"kind": "local", "path": "./data"})
assert isinstance(local, LocalStorage)

The caller supplies StorageConfig; kind selects the implementation. @register infers each concrete config type from its annotation.

Compatibility

• Python 3.9+: use from __future__ import annotations for A | B annotations on 3.9.
• Pydantic v1.10 and v2 are supported without an exact version pin. Conversion behavior follows the installed version; uv resolves a compatible release.

Preview the documentation with uv run --group docs mkdocs serve.

Documentation · AI agent guide · Issues · MIT License