Skip to content

Settings Files

Optionally you can store settings in files, dynaconf supports multiple file formats, you are recommended to choose one format but you can also use mixed settings formats across your application.


You are not required to use settings files, if not specified dynaconf can load your data only from environment variables

Supported formats

  • .toml - Default and recommended file format.
  • .yaml|.yml - Recommended for Django applications. (see yaml caveats)
  • .json - Useful to reuse existing or exported settings.
  • .ini - Useful to reuse legacy settings.
  • .py - Not Recommended but supported for backwards compatibility.
  • .env - Useful to automate the loading of environment variables.


Create your settings in desired format and specify it on settings_files argument on your dynaconf instance or pass it in -f <format> if using dynaconf init command.


Can't find the file format you need for your settings? You can create your custom loader and read any data source. read more on extending dynaconf


To use the .ini or .properties file format you need to install extra dependency pip install configobj or pip install dynaconf[ini]

Reading settings from files

On files by default dynaconf loads all the existing keys and sections as first level settings.

name = "Bruno"
name: Bruno
{"name": "Bruno"}
name = 'Bruno'
NAME = "Bruno"

⚠️ on .py files dynaconf only read UPPERCASE variables.

Then on your application code: == "Bruno"


The default encoding when loading the settings files is utf-8 and it can be customized via encoding parameter.

Settings file location

Dynaconf will search files specified in settings_file option starting the search tree on the current working dir (the directory where your program is located).


from dynaconf import Dynaconf

settings = Dynaconf(settings_files=["settings.toml", "/etc/program/foo.yaml"])


To use python -m module, where the module uses dynaconf you will need to specify your settings.toml path, for example, like this: settings_file="module/config/settings.toml".


In the above example, dynaconf will try to load settings.toml from the same directory where the program is located, also known as . and then will keep traversing the folders in backwards order until the root is located.

root is either the path where the program was invoked, or the O.S root or the root specified in root_path.


Dynaconf will then recognize this as an absolute path and will try to load it directly from the specified location.

Local Settings files

For each file specified in settings_files dynaconf will also try to load an optional name.local.extension.

For example, settings_files=["settings.toml"] will make dynaconf to search for settings.toml and then also search for settings.local.toml


You can also specify includes so dynaconf can include those settings after the normal loading.

as a parameter

settings = Dynaconf(includes=["path/to/file.toml", "or/a/glob/*.yaml])

as a variable in a file

dynaconf_include = ["path/to/file.toml"]
key = value
anotherkey = value


The paths passed to includes are relative to the root_path of Dynaconf instance or if that is not set, relative to the directory where the first loaded file is located, includes also accepts globs and absolute paths.

Layered environments on files

It is also possible to make dynaconf to read the files separated by layered environments so each section or first level key is loaded as a distinct environment.

settings = Dynaconf(environments=True)
name = ""
name = "developer"
name = "admin"
    name: ''
    name: developer
    name: admin
    "default": {
        "name": ""
    "development": {
        "name": "developer"
    "production": {
        "name": "admin"
name = ""
name = "developer"
name = "admin"

ℹ️ You can define custom environment using the name you want [default] and [global] are the only environments that are special. You can for example name it [testing] or [anything]

Then in your program you can use environment variables to switch environments.

export ENV_FOR_DYNACONF=development == "developer"

export ENV_FOR_DYNACONF=production == "admin"


On Flask and Django extensions the default behaviour is already the layered environments. Also to switch the environment you use export FLASK_ENV=production or export DJANGO_ENV=production respectively.


It is also possible to switch environments programmatically passing env="development" to Dynaconf class on instantiation.

YAML Caveats


Yaml parser used by dynaconf (ruamel.yaml) reads undefined values as None so

key: ~
key: null

All those 3 keys will be parsed as python's None object.

When using a validator to set a default value for those values you might want to use one of:

Validator("key", default="thing", apply_default_on_none=True)

This way dynaconf will consider the default value even if setting is None on yaml.

or on yaml you can set the value to @empty

key: "@empty"

NEW in 3.1.9

Last update: 2022-09-03