Skip to main content

`konfetti` provides a framework-independent way for configuration of applications or libraries written in Python.

Project description


codecov Build Version Python versions License


NOTE: The documentation is in progress

konfetti provides a framework-independent way for configuration of applications or libraries written in Python.

Key features:

  • Lazy evaluation;
  • Built-in environment variables support;
  • Built-in Vault support;
  • Helpers for tests.

The primary motivation for building this library is to unify all configuration in different projects and make it as lazy as possible. In this case, we could get these benefits:

  • No need to full app configuration when running a subset of tests that don't need a full config;
  • Avoid network calls during imports until necessary;

The interface design and features are heavily inspired by Django & decouple.

Supported Python versions: 2.7 & 3.5 - 3.8


To use konfetti you need to define:

  • configuration variables in a module or a class;
  • an access point;

Settings module

# app_name/settings/
from konfetti import env, vault


DEBUG = env("DEBUG", default=False)
DATABASE_URI = vault("path/to/db")

NOTE: The naming convention for variables names is to use upper case, other variables will be ignored.

Access point

# app_name/settings/
from konfetti import Konfig, VaultBackend

config = Konfig(vault_backend=VaultBackend("/secret/team"))

konfetti relies on KONFETTI_SETTINGS environment variable to discover your settings module, in the case above:

export KONFETTI_SETTINGS=app_name.settings.production


The settings module/class with configuration options shouldn't be accessed directly, because the aforementioned features are implemented in the access point level.

from app_name.settings import config

def something():


Table of contents:

Lazy evaluation

Until a config option is accessed it is not evaluated - it is lazy. To avoid side effects on imports accessing configuration should be avoided on a module level.

This concept allows you to choose when you actually evaluate the config. Why?

  • Testing. If you need to test a small piece of code that doesn't require any configuration - you don't have to setup it;
  • Faster application startup; Use only what you need at the moment

It is still possible to evaluate the config eagerly on the app startup - access the needed variables in the entry points. It could be done either with direct accessing needed variables or with config.require(...) / config.asdict() calls.


from konfetti import env

VARIABLE = env("VARIABLE_NAME", default="foo") 

Since environment variables are strings, there is a cast option to convert given variable from a string to the desired type:

from konfetti import env

VARIABLE = env("VARIABLE_NAME", default=42, cast=int)

You can pass any callable as cast. If there is a need to use the environment variable immediately, it could be evaluated via str call (other ways could be added on demand):

from konfetti import env, vault

DATABASE_ROLE = env("DATABASE_ROLE", default="booking")


If cast is specified, then it will be applied before evaluation as well.

.env support

It is possible to specify a path to the .env file and it will be used as a source of data for environment variables.

dotenv_override parameter specifies whether the .env value should be used if both the environment variable and the .env record exists, False by default.

# app_name/settings/
from konfetti import Konfig

config = Konfig(dotenv="path/to/.env", dotenv_override=False)


Backend configuration

To use Vault as a secrets storage you need to configure the access point:

# app_name/settings/
from konfetti import Konfig, VaultBackend

config = Konfig(vault_backend=VaultBackend("your/prefix"))

There are two Vault backends available:

  • konfetti.VaultBackend
  • konfetti.AsyncVaultBackend

The main difference is that the latter requires using await to access the secret value (the call will be handled asynchronously under the hood), otherwise the interfaces and capabilities are the same.

Each backend requires a prefix to be specified, the trailing / leading slashes don't matter, "your/prefix" will work the same as "/your/prefix/".


Every Vault secret needs a path to be used as a lookup (leading and trailing slashes don't matter as well):

# app_name/settings/
from konfetti import vault

WHOLE_SECRET = vault("path/to")

In this case all key/value pairs will be loaded on evaluation:

>>> from app_name.settings import config
>>> config.WHOLE_SECRET
{'key': 'value', 'foo': 'bar'}

You can specify a specific key to be returned for a config option with [] syntax:

# app_name/settings/
from konfetti import vault

KEY = vault("path/to")["key"]
>>> from app_name.settings import config
>>> config.KEY

Using square brackets will not trigger evaluation - you could specify as many levels as you want:

# app_name/settings/
from konfetti import vault

DEEP = vault("path/to")["deeply"]["nested"]["key"]

Casting could be specified as well:

# app_name/settings/
from decimal import Decimal
from konfetti import vault

DECIMAL = vault("path/to", cast=Decimal)["fee_amount"]  # stored as string
>>> from app_name.settings import config
>>> config.DECIMAL

Sometimes you need to access to some secrets dynamically. Konfig provides a way to do it:

>>> from app_name.settings import config
>>> config.get_secret("path/to")["key"]
Secret files

It is possible to get a file-like interface for vault secret.

# app_name/settings/
from konfetti import vault_file

KEY = vault_file("path/to/file")["key"]
>>> from app_name.settings import config
>>> config.KEY.readlines()

It is possible to specify the default value for vault variable. Value could be any type for a key in a secret and a dict for the whole secret.

DEFAULT = vault("path/to", default="default")["DEFAULT"]
DEFAULT_SECRET = vault("path/to", default={"DEFAULT_SECRET": "default_secret"})

>>> from app_name.settings import config
>>> config.DEFAULT
{"DEFAULT_SECRET": "default_secret"}

Defaults could be disabled entirely if VAULT_DISABLE_DEFAULTS is set

Overriding Vault secrets

In some cases, secrets need to be overridden in runtime on the application level. You can define some custom values for tests or you just want to run the app with some different configuration without changing data in Vault.

There is a way to do it using environment variables or .env records To redefine certain config option you need to redefine the whole secret with a JSON encoded string.


# app_name/settings/
from konfetti import vault

KEY = vault("path/to")["key"]
>>> from app_name.settings import config
>>> config.KEY
>>> import os
>>> os.environ["PATH__TO"] = '{"key": "overridden"}'
>>> config.KEY

To check how to override certain option there is a config.vault.get_override_examples() helper:

>>> config.vault.get_override_examples()
        "PATH__TO__NESTED": '{"NESTED_SECRET": {"nested": "example_value"}}'
    "SECRET": {
        "PATH__TO": '{"SECRET": "example_value"}'
        "PATH__TO": "{}"

By default, when the evaluation will happen on a Vault secret, the environment will be checked first. If you don't need this behavior, it could be turned off with try_env_first=False option to the chosen backend:

# app_name/settings/
from konfetti import Konfig, VaultBackend

config = Konfig(vault_backend=VaultBackend("your/prefix", try_env_first=False))
Disabling access to secrets

If you want to forbid any access to Vault (e.g. in your tests) you can set KONFETTI_DISABLE_SECRETS environment variable with 1 / on / true / yes.

>>> import os
>>> from app_name.settings import config
>>> os.environ["KONFETTI_DISABLE_SECRETS"] = "1"
>>> config.get_secret("path/to")["key"]
RuntimeError: Access to secrets is disabled. Unset KONFETTI_DISABLE_SECRETS variable to enable it. 

Vault values could be cached in memory:

config = Konfig(vault_backend=VaultBackend("your/prefix", cache_ttl=60))

By default, caching is disabled.

Lazy options

If there is a need to calculate config options dynamically (e.g., if it depends on values of other options) konfetti provides lazy:

from konfetti import lazy

LAZY_LAMBDA = lazy(lambda config: config.KEY + "/" + config.SECRET + "/" + config.REQUIRED)

def lazy_property(config):
    return config.KEY + "/" + config.SECRET + "/" + config.REQUIRED


It is usually a good idea to use a slightly different configuration for tests (disabled tracing, sentry, etc.).

export KONFETTI_SETTINGS=app_name.settings.tests

It is very useful to override some config options in tests. Konfig.override will override config options defined in the settings module. It works as a context manager or a decorator to provide explicit setup & clean up for overridden options.

from app_name.settings import config

# DEBUG will be `True` for `test_everything`
def test_everything():
    # DEBUG will be `False` again for this block 
    with config.override(DEBUG=False):

Overrides could be nested, and deeper level has precedence over all levels above:

from app_name.settings import config

@config.override(FOO=1, BAR=2)
def test_many_things():
    with config.override(BAR=3):
        assert config.FOO == 1
        assert config.BAR == 3
    # As it was before
    assert config.BAR == 2

Also, override works for classes (including inherited from unittest.TestCase):

class TestOverride:

    def test_override(self):
        assert config.INTEGER == 123

    def test_another_override(self):
        assert config.INTEGER == 456

def test_not_affected():
    assert config.INTEGER == 1

NOTE. setup_class/setUp and teardown_class/tearDown methods will work with override.

konfetti includes a pytest integration that gives you a fixture, that allows you to override given config without using a context manager/decorator approach and automatically rollbacks changes made:

import pytest
from app_name.settings import config
from konfetti.pytest_plugin import make_fixture

# create a fixture. the default name is "settings",
# but could be specified via `name` option

def global_settings(settings):
    settings.INTEGER = 456

def test_something(settings):
    assert settings.INTEGER == 456
    assert config.INTEGER == 456

    # fixture overriding
    settings.INTEGER = 123
    assert settings.INTEGER == 123
    assert config.INTEGER == 123

    # context manager should work as well
    with settings.override(INTEGER=7):
        assert settings.INTEGER == 7
        assert config.INTEGER == 7

    # Context manager changes are rolled back
    assert settings.INTEGER == 123
    assert config.INTEGER == 123

# This test is not affected by the fixture
def test_disable(settings):
    assert config.INTEGER == 1
    assert settings.INTEGER == 1

NOTE. It is forbidden to create two fixtures from the same config instances.


The environment variable name could be customized via config_variable_name option:

config = Konfig(config_variable_name="APP_CONFIG")

Alternatively, it is possible to specify class-based settings:

from konfetti import env, vault

class ProductionSettings:

    DEBUG = env("DEBUG", default=False)
    DATABASE_URI = vault("path/to/db")

It possible to load the whole config and get its content as a dict:

>>> config.asdict()
    "ENV": "env value",
    "KEY": "static value",
    "SECRET": "secret_value",

If you need to validate that certain variables are present in the config, there is require:

>>> config.require("SECRET")
MissingError: Options ['SECRET'] are required

Or to check that they are defined:

>>> "SECRET" in config

Configuration 101

There are a couple of principles that will help you to avoid problems when you specify or use your configuration.

Do not access configuration on the module level

Do this:

from app_name.settings import config

def get_redis_client():
    return StrictRedis.from_url(config.REDIS_URL)

Instead of this:

from redis import StrictRedis
from app_name.settings import config

cache_redis = StrictRedis.from_url(config.REDIS_URL)

However, if you want to have a global Redis instance, consider using python-lazy-object-proxy:

pip install lazy-object-proxy
import lazy_object_proxy


cache_redis = lazy_object_proxy.Proxy(get_redis_client)


Accessing configuration on the module level leads to side-effects on imports, this fact could produce unrelated errors when you run your test suite:

  • Simple unit tests will fail due to lack of configuration options or Vault unavailability;
  • Slow tests due to config initialization and long network calls (they could time out as well);

Having your config access lazy will prevent many for those cases because that code branches won't be executed on imports and will not affect your test suite.

Code formatting

In order to maintain code formatting consistency we use black to format the python files. A pre-commit hook that formats the code is provided but it needs to be installed on your local git repo, so...

In order to install the pre-commit framework run pip install pre-commit or if you prefer homebrew brew install pre-commit

Once you have installed pre-commit just run pre-commit install on your repo folder

If you want to exclude some files from Black (e.g. automatically generated database migrations or test snapshots) please follow instructions for pyproject.toml


To run all tests:

docker run -p 8200:8200 -d --cap-add=IPC_LOCK -e 'VAULT_DEV_ROOT_TOKEN_ID=test_root_token' vault:0.9.6
tox -p all

Note that tox doesn't know when you change the requirements.txt and won't automatically install new dependencies for test runs. Run pip install tox-battery to install a plugin which fixes this silliness.

It also possible to run tests via docker-compose that will start up all required environment:

$ make docker-test

or alternatively:

$ docker-compose -f docker-compose-tests.yml run konfetti



Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for konfetti, version 0.6.0
Filename, size File type Python version Upload date Hashes
Filename, size konfetti-0.6.0-py2.py3-none-any.whl (24.4 kB) File type Wheel Python version py2.py3 Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page