Frequently Asked Questions¶

Here are some curated questions about Dynaconf usage.

To get more help, please use the GitHub discussions.
If you'd like to suggest some new questions here, use GitHub issues.

Dynamically set `settings_files` path¶

I have multiple environments and I don't want to rely on a particular location of the settings file. How do I set its path dynamically?

Dynaconf-specific config options, such as settings_file or settings_files, have two ways of being defined:

On instance options: Dynaconf(settings_file="path.to.settings.toml")
On environment options: export SETTINGS_FILE_FOR_DYNACONF="path.to.settings"

So, for each environment, just export the right path for it. Note that it should be UPPERCASE and have _FOR_DYNACONF suffix

Environment settings will, by default, override instance settings.

Problem running tests in IDE¶

I'm trying to run a test on my IDE, but there it raises an OSError: Starting path not found exception, although running from CLI works fine. What can it be?

Dynaconf path resolving relies on your CWD (current working directory) being at the project's root. Some IDEs may modify the CWD to run tests, and this will cause this issue.

PyCharm, for example, may change the CWD to ./tests in their integrated test system. If this is your specific case, go to "Modify Run Configuration > Working Dir" on the test tab and assure that it points to the root of your project.

Conditional Validation¶

Is it possible to create a Validator that would require the presence of an option if another option is present?

Yes. Validators provide great flexibility in the ways you can arrange them. If foo must only be present if bar exists, you could do:

Validator(
        "foo",
        must_exist=True,
        when=Validator(
            "bar", must_exist=True,
        ),
    ),

See more about validators

Deprecation mechanism¶

I have a quite busy project and sometimes old config files get in the way. Is there a deprecation mechanism to deal with this?

There is already a feature request about a more specific deprecation mechanism, but right now you can use hooks to achieve that. You might do something like:

# dynaconf_hooks.py
def post(settings: Dynaconf) -> dict:
    DEPRECATED = {
        "FOO": "NEWKEY"
    }
    for key, new in DEPRECATED.items():
        if value := settings.get(key):
            warnings.warn(f"{key} has been replaced by {new}")
            settings.set(new, value)

See more about hooks

Django functions inside settings¶

I use some django utility functions in my django settings.py. Can I still use them if I choose to use yaml to manage my config?

Yes, you may add a custom converter for any functions you like. Refer here for a full example.

Default-override workflow¶

I'd like to use a base-file as default settings (like django settings.py) and a user-file settings where I would add my custom overrides more explicitly. Can this be done?

Yes, this is entirely possible. By default, top-level settings with different keys will both be kept, while the ones with the same keys will be overridden.

# settingsA.yaml
name="John"

# settingsB.yaml
age=23

# resultant
name="John"
age=23

Note that this will override entire nested structures too (like lists or dicts).

# settingsA.yaml
bucket=[1,2,3,4]

# settingsB.yaml
bucket=["a", "b", "c", "d"]

# resultant (default)
bucket=["a", "b", "c", "d"]

To control how you want to merge these structures, you may want to have a look at merging There are several ways you can merge data: settings the global option merge_enabled, using dunder syntax or marking a whole file, a specific env, or just an option for merging. Choose what fits best for your case.

Error with Pyinstaller executable¶

I had some trouble running an executable app that uses Dynaconf. What could it be?

A user reported this situation while using Pyinstaller. For his case, the fix was to package dynaconf and python-dotenv[cli] without compiling by using the --collect-all argument of pyinstaller.

Dynaconf instance from dict¶

I would like a Dynaconf object to be created from a python dict, while still being able to use validators on it. Is it possible?

Yes, it is possible.

Let's say you got a stringfied json directly from an API and you'd like to integrate it to a Dynaconf object using some validation. You could do this:

data = get_data_as_dict()

settings = Dynaconf()

settings.validators.register(Validator("aws", "elasticsearch", must_exist=True))
settings.update(data)
settings.validators.validate()