File-backed Pydantic configuration with migration support.
pip install fluxconfor with uv:
uv add fluxconfConfigIO is a generic base class for reading and writing YAML-backed Pydantic models. Subclass it, set file_name and config_type, and you get type-safe read/write out of the box.
from pydantic import BaseModel
from fluxconf import ConfigIO
class AppConfig(BaseModel):
name: str = "my-app"
debug: bool = False
class AppConfigIO(ConfigIO[AppConfig]):
file_name = "app.yml"
config_type = AppConfig
io = AppConfigIO("~/.config/my-app")
io.write(AppConfig(name="my-app", debug=True))
config = io.read() # AppConfig(name='my-app', debug=True)Migrations transform stored configuration data when your schema evolves. Instead of breaking existing config files, you define migration steps that update old data to match the new schema.
To use migrations, inherit from VersionedBaseModel instead of Pydantic's BaseModel. This adds a version field that tracks which migrations have been applied.
Migration keys follow the "N_description" format — the integer prefix determines execution order and is stored as the version in the config file. On read(), any pending migrations run automatically and the file is updated on disk.
The simplest approach: declare JSON Patch (RFC 6902) operations directly. No Python functions needed.
from fluxconf import ConfigIO, VersionedBaseModel
class ServerConfig(VersionedBaseModel):
host: str = "localhost"
port: int = 8080
class ServerConfigIO(ConfigIO[ServerConfig]):
file_name = "server.yml"
config_type = ServerConfig
migrations = {
"1_rename_host": [
{"op": "move", "from": "/hostname", "path": "/host"},
],
"2_add_port": [
{"op": "add", "path": "/port", "value": 8080},
],
}Supported operations: add, remove, replace, move, copy, and test.
When you need conditional logic or complex transforms, use a Python function. Each function receives the raw config dict and must return the updated dict.
from fluxconf import ConfigIO, VersionedBaseModel
class UserConfig(VersionedBaseModel):
full_name: str = ""
email: str = ""
def merge_name_fields(data: dict) -> dict:
first = data.pop("first_name", "")
last = data.pop("last_name", "")
if first or last:
data["full_name"] = f"{first} {last}".strip()
return data
class UserConfigIO(ConfigIO[UserConfig]):
file_name = "user.yml"
config_type = UserConfig
migrations = {
"1_merge_name": merge_name_fields,
}Python functions and JSON Patches can be mixed freely in the same migrations dict.
For projects with many migrations, store each one as a separate file in a directory instead of inlining them all in the class definition.
myapp/migrations/
1_rename_host.json
2_merge_name.py
3_add_defaults.py
_helpers.py # skipped (starts with _)
Only files with an integer prefix are loaded. Files starting with _ or without an integer prefix are silently skipped, so helper modules can live alongside migration files.
.json files contain a JSON array of patch operations:
myapp/migrations/1_rename_host.json
[
{"op": "move", "from": "/hostname", "path": "/host"}
].py files with a patch attribute are equivalent to .json files but written in Python:
myapp/migrations/3_add_defaults.py
patch = [
{"op": "add", "path": "/port", "value": 8080},
{"op": "add", "path": "/retries", "value": 3},
].py files with a migrate function offer full flexibility:
myapp/migrations/2_merge_name.py
def migrate(data: dict) -> dict:
first = data.pop("first_name", "")
last = data.pop("last_name", "")
if first or last:
data["full_name"] = f"{first} {last}".strip()
return dataIf a .py file defines both migrate and patch, the migrate function takes precedence.
Point migrations_dir at the directory to load them:
from pathlib import Path
from fluxconf import ConfigIO, VersionedBaseModel
class AppConfigIO(ConfigIO[AppConfig]):
file_name = "app.yml"
config_type = AppConfig
migrations_dir = Path(__file__).parent / "migrations"migrations and migrations_dir can be used together — fluxconf merges them, raising ValueError on key collisions.
MigrationError is raised when a migration function or patch fails. It carries two attributes:
last_successful_migration— the version of the last migration that completed successfully (or the stored version if none succeeded)original_error— the underlying exception
ValueError is raised when the stored version is ahead of the latest known migration. This typically means the config file was written by a newer version of the software than the one currently running.