Arbitrary keys in Pydantic models

I was working on a project that used TOML as the file format for configuration and it looked like this:

[project_a]
vault = "abc"
repo = "def"
 
[project_b]
vault = "ghi"
repo = "jkl"

and I needed to validate the data based on this schema.

When this data is read in with a library like toml, we get a dict:

>>> import toml
>>> with open('config.toml', 'r') as config_file:
...     config_data = toml.load(config_file)
>>> config_data
# {'project_a': {'vault': 'abc', 'repo': 'def'}, 'project_b': {'vault': 'ghi', 'repo': 'jkl'}}

and I needed to let the project names to be anything and wanted to validate the configurations themselves.

Here’s how I did that with pydantic using RootModel:

from typing import Dict
 
from pydantic import RootModel, BaseModel
import toml
 
 
class Configuration(RootModel):
    root: Dict[str, SiteConfiguration]
 
 
class SiteConfiguration(BaseModel):
    vault: str
    repo: str
 
 
def read_configuration(filename: str) -> Configuration:
	with open(filename, 'r') as config_file:
        config_data = toml.load(config_file)
        return Configuration.model_validate(config_data)

Now, our config is:

>>> print(read_configuration('config.toml'))
root={'project_a': SiteConfiguration(vault='abc', repo='def'), 'project_b': SiteConfiguration(vault='ghi', repo='jkl')}

and we can work on it knowing that our configuration is well formed.