dynaconf - The dynamic configurator for your Python Project

dynaconf a layered configuration system for Python applications (with strong support for 12-factor applications).

# install it in a Python 3 environment
pip3 install dynaconf
# optionally
pip3 install PyYAML toml redis

How does it work?

# import the unique dynaconf object
from dynaconf import settings

# access your config variables
Connect(user=settings.USERNAME, passwd=settings.PASSWD)

# You can provide defaults in case config is missing
Connect(user=settings('USERNAME', 'admin'), passwd=settings('PASSWD', 1234))

Where the values come from?

• settings.{py|yaml|toml|ini} file
• .env file
• exported Environment Variables
• Remote storage servers
• Multiple customizable sources

12factor recommended example (environment variables):

# put some variable in a .env file
echo "DYNACONF_USERNAME=admin" >> .env

# Or export directly
export DYNACONF_USERNAME=admin
export DYNACONF_PASSWD='@int 1234'  # you can type the values!

Just read it

# import the unique dynaconf object
from dynaconf import settings

# access your config variables
Connect(user=settings.USERNAME, passwd=settings.PASSWD)

NOTE: You can customize the prefix DYNACONF_ to your own namespace like MYAPP_USERNAME.

Features

• Read config variables from unique dynaconf object
• Values can come from different sources:
  • Load values from settings.{py|yaml|toml|ini} file
  • Load values from .env files
  • Load values from System's Exported Environment Variables
  • And even more:
    • Load values from a remote Redis server
    • Load values from a remote SQL database
    • Load values from a remote memcached
    • Load values from a remote Secrets Vault
  • And if you want:
    • Load values from anywhere you want, easy to implement your own loader!
• Flexible typing system
  • Export Typed environment variables using dynaconf data type markers export FOO=@int 42
• Flask Support
  • In your Flask application just do FlaskDynaconf(app) and then read values from app.config object
• Feature Flag System
  • Implement a feature flag system using dynaconf as store and checker
• Value validation
  • Validate the config variables and define rules

Examples

Namespace support

When you are working with multiple projects using the same environment maybe you want to use different namespaces for ENV vars based configs

export DYNACONF_DATABASE="DYNADB"
export PROJ1_DATABASE="PROJ1DB"
export PROJ2_DATABASE="PROJ2DB"

and then access them

>>> from dynaconf import settings

# access default namespace settings
>>> settings.DATABASE
'DYNADB'

# switch namespaces
>>> settings.namespace('PROJ1')
>>> settings.DATABASE
'PROJ1DB'

>>> settings.namespace('PROJ2')
>>> settings.DATABASE
'PROJ2DB'

# return to default, call it without args
>>> settings.namespace()
>>> settings.DATABASE
'DYNADB'

You can also use the context manager:

>>> settings.DATABASE
'DYNADB'

>>> with settings.using_namespace('PROJ1'):
...    settings.DATABASE
    'PROJ1DB'

>>> with settings.using_namespace('PROJ2'):
...    settings.DATABASE
    'PROJ2DB'

>>> settings.DATABASE
'DYNADB'

namespace() and using_namespace() takes optional argument clean defaults to True. If you want to keep the pre-loaded values when switching namespaces set it to False.

Namespaced environment

It is usual to have e.g production and development environments, the way to set this is:

Using settings.py as base file you just need other <environment>_settings.py files.

settings.py
development_settings.py
production_settings.py

Then in your environment.

export DYNACONF_NAMESPACE=DEVELOPMENT|PRODUCTION  # switch enviroment using env vars.

Or using namespace

with settings.using_namespace('development'):
    # code here

settings.namespace('development')

NOTE: settings.py is the base and namespace specific overrides its vars.

using YAML

you need to install PyYAML pip install PyYAML

Just save a settings.yaml in the root dir.

Using YAML is easier because it support multiple namespace in a single file.

Lets say you have DYNACONF_NAMESPACE=DYNACONF (the default)

DYNACONF:  # this is the global namespace
  VARIABLE: 'this variable is available on every namespace'
  HOST: null  # this variable is set to None

DEVELOPMENT:
  HOST: devserver.com  # overrides the global or sets new

production:  # upper or lower case does not matter
  host: prodserver.com

Then it will be applied using env var DYNACONF_NAMESPACE or context manager.

HINT: When using yaml namespace identifier and first level vars are case insensitive, dynaconf will always have them read as upper case.

using TOML

you need to install toml pip install toml

Just save a settings.toml in the root dir.

Using TOML is easier because it support multiple namespace in a single file.

Lets say you have DYNACONF_NAMESPACE=DYNACONF (the default)

[dynaconf]:  # this is the global namespace
variable = 'this variable is available on every namespace'
HOST = false  # this variable is set to False

[DEVELOPMENT]
HOST = 'devserver.com'  # overrides the global or sets new

[production]  # upper or lower case does not matter
host = 'prodserver.com'

Then it will be applied using env var DYNACONF_NAMESPACE or context manager.

HINT: When using toml namespace identifier and first level vars are case insensitive, dynaconf will always have them read as upper case.

casting values from envvars

Sometimes you need to set some values as specific types, boolean, integer, float or lists and dicts.

built in casts

• @int (as_int)
• @bool (as_bool)
• @float (as_float)
• @json (as_json)

@json / as_json will use json to load a Python object from string, it is useful to get lists and dictionaries. The return is always a Python object.

strings does not need converters.

You have 2 ways to use the casts.

Casting on declaration

Just start your ENV settigs with this

export DYNACONF_DEFAULT_THEME='material'
export DYNACONF_DEBUG='@bool True'
export DYNACONF_DEBUG_TOOLBAR_ENABLED='@bool False'
export DYNACONF_PAGINATION_PER_PAGE='@int 20'
export DYNACONF_MONGODB_SETTINGS='@json {"DB": "quokka_db"}'
export DYNACONF_ALLOWED_EXTENSIONS='@json ["jpg", "png"]'

Starting the settings values with @ will make dynaconf.settings to cast it in the time od load.

Casting on access

export DYNACONF_USE_SSH='yes'

from dynaconf import settings

use_ssh = settings.get('USE_SSH', cast='@bool')
# or
use_ssh = settings('USE_SSH', cast='@bool')
# or
use_ssh = settings.as_bool('USE_SSH')

print use_ssh

True

more examples

export DYNACONF_USE_SSH='enabled'

export DYNACONF_ALIST='@json [1,2,3]'
export DYNACONF_ADICT='@json {"name": "Bruno"}'
export DYNACONF_AINT='@int 42'
export DYNACONF_ABOOL='@bool on'
export DYNACONF_AFLOAT='@float 42.5'

from dynaconf import settings

# original value
settings('USE_SSH')
'enabled'

# cast as bool
settings('USE_SSH', cast='@bool')
True

# also cast as bool
settings.as_bool('USE_SSH')
True

# cast defined in declaration '@bool on'
settings.ABOOL
True

# cast defined in declaration '@json {"name": "Bruno"}'
settings.ADICT
{u'name': u'Bruno'}

# cast defined in declaration '@json [1,2,3]'
settings.ALIST
[1, 2, 3]

# cast defined in decalration '@float 42.5'
settings.AFLOAT
42.5

# cast defined in declaration '@int 42'
settings.AINT
42

Defining default namespace

Include in the file defined in DYNACONF_SETTINGS or in the .env file or in the customized Settings class the desired namespace

DYNACONF_NAMESPACE = 'DYNACONF'

Storing settings in databases

Using REDIS

Redis support relies on the following two settings that you can setup in the DYNACONF_SETTINGS file

1 Add the configuration for redis client

REDIS_FOR_DYNACONF = {
    'host': 'localhost',
    'port': 6379,
    'db': 0
}

NOTE: if running on Python 3 include 'decode_responses': True in REDIS_FOR_DYNACONF

Include redis_loader in dynaconf LOADERS_FOR_DYNACONF

the order is the precedence

# Loaders to read namespace based vars from diferent data stores
LOADERS_FOR_DYNACONF = [
    'dynaconf.loaders.env_loader',
    'dynaconf.loaders.redis_loader'
]

You can now write variables direct in to a redis hash named DYNACONF_< NAMESPACE >

By default DYNACONF_DYNACONF

You can also use the redis writer

from dynaconf.utils import redis_writer
from dynaconf import settings

redis_writer.write(settings, name='Bruno', database='localhost', PORT=1234)

The above data will be converted to namespaced values and recorded in redis as a hash:

DYNACONF_DYNACONF:
    NAME='Bruno'
    DATABASE='localhost'
    PORT='@int 1234'

if you want to skip type casting, write as string intead of PORT=1234 use PORT='1234' as redis stores everything as string anyway

Data is read from redis and another loaders only once or when namespace() and using_namespace() are invoked. You can access the fresh value using settings.get_fresh(key)

There is also the fresh context manager

from dynaconf import settings

print settings.FOO  # this data was loaded once on import

with settings.fresh():
    print settings.FOO  # this data is being directly read from loaders

And you can also force some variables to be fresh setting in your setting file

DYNACONF_ALWAYS_FRESH_VARS = ['MYSQL_HOST']

or using env vars

export DYNACONF_ALWAYS_FRESH_VARS='@json ["MYSQL_HOST"]'

Then

from dynaconf import settings

print settings.FOO  # This data was loaded once on import

print settings.MYSQL_HOST # This data is being read from redis imediatelly!

Using programatically

Sometimes you want to override settings for your existing Package or Framework lets say you have a conf module exposing a settings object and used to do:

from myprogram.conf import settings

Now you want to use Dynaconf, open that conf.py or conf/__init__.py and do:

# coding: utf-8
from dynaconf import LazySettings

settings = LazySettings(
    ENVVAR_FOR_DYNACONF="MYPROGRAM_SETTINGS_MODULE",
    DYNACONF_NAMESPACE='MYPROGRAM'
)

Now you can import settings from your own program and dynaconf will do the rest!

Flask Extension

Dynaconf provides an extension to make your app.config in Flask to be a dynaconf instance.

from flask import Flask
from dynaconf import FlaskDynaconf

app = Flask(__name__)
FlaskDynaconf(app)

The FlaskDynaconf takes optional arguments

ENVVAR_FOR_DYNACONF="MYSITE_SETTINGS_MODULE"  # defaults to `DYNACONF_SETTINGS` 
DYNACONF_NAMESPACE='MYSITE'  # defaults to `FLASK_`
SETTINGS_MODULE_FOR_DYNACONF='settings.yml'  # defaults to `settings.py`
YAML='.secrets.yml',  # aditional config
EXTRA_VALUE='You can add aditional config vars here'

DEBUGGING

By default Dynaconf only outputs the ERROR level to debug it change

export DYNACONF_DEBUG_LEVEL='INFO'

The loading precedende order

Dynaconf will perform loads in this order:

1. Load Default configuration
2. Load Environment variables (pre load to read initial DYNACONF_ config)
3. Load Settings file in the order defined in SETTINGS_MODULE_FOR_DYNACONF by default will to load 'settings.py,settings.yaml,settings.toml' in this order overriding previous values
4. Load all loaders defined in LOADERS_FOR_DYNACONF by default only environment variables will be read again and get higher precedence

Customizing the loaders

In a setting file like settings.{py|yaml|toml} define:

LOADERS_FOR_DYNACONF = [
    'dynaconf.loaders.env_loader',
    'dynaconf.loaders.redis_loader',
    'YourCustomLoaderPath'
]

export also works

export LOADERS_FOR_DYNACONF='@json ["loader1", "loader2"]'

Loaders will be executed in that order.

To disable loaders do:

LOADERS_FOR_DYNACONF=0

This will cause environment variables to lose the higher precedence

More examples:

Take a look at example/ for more.

This was inspired by flask.config and django.conf.settings