ValidationError when importing library

[mdavis-xyz]

It seems that something is checking for the presence of the API in the environment vars when I import the library, and is throwing an exception when not found.

Steps to reproduce

• Install openelectricity == 0.9.2
• Do not set API key in the env vars
• Run this line of python: from openelectricity import OEClient

Expected Result

The import returns a client constructor.
No exception is thrown.

Then I can subsequently pass the key as an explicit argument

with open(api_key_path, 'r') as f:
    key = f.read().strip()
 
client = OEClient(api_key=key)

The documentation says I can pass a key either as an env var, or as an argument. So I think this behavior does not match the intended behavior.

Actual Result

s> python
Python 3.14.2 | packaged by conda-forge | (main, Dec  6 2025, 11:17:27) [MSC v.1944 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> from openelectricity import OEClient
Traceback (most recent call last):
  File "<python-input-0>", line 1, in <module>
    from openelectricity import OEClient
  File "C:\Users\DAVIS_M\AppData\Local\anaconda3\envs\ausbids\Lib\site-packages\openelectricity\__init__.py", line 7, in <module>
    from openelectricity.client import AsyncOEClient, OEClient
  File "C:\Users\DAVIS_M\AppData\Local\anaconda3\envs\ausbids\Lib\site-packages\openelectricity\client.py", line 13, in <module>
    from openelectricity.logging import get_logger
  File "C:\Users\DAVIS_M\AppData\Local\anaconda3\envs\ausbids\Lib\site-packages\openelectricity\logging.py", line 12, in <module>
    from openelectricity.settings_schema import settings
  File "C:\Users\DAVIS_M\AppData\Local\anaconda3\envs\ausbids\Lib\site-packages\openelectricity\settings_schema.py", line 24, in <module>
    settings = Settings()
  File "C:\Users\DAVIS_M\AppData\Local\anaconda3\envs\ausbids\Lib\site-packages\pydantic_settings\main.py", line 194, in __init__
    super().__init__(
    ~~~~~~~~~~~~~~~~^
        **__pydantic_self__._settings_build_values(
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    ...<27 lines>...
        )
        ^
    )
    ^
  File "C:\Users\DAVIS_M\AppData\Local\anaconda3\envs\ausbids\Lib\site-packages\pydantic\main.py", line 250, in __init__
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
pydantic_core._pydantic_core.ValidationError: 1 validation error for Settings
OPENELECTRICITY_API_KEY
  Field required [type=missing, input_value={}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/missing

Discussion

I prefer to use the explicit approach, because I'm in a new company where the actual environment variables are unclear. (Conda, pip, proxies, antivirus. Argh, it's hard. If I can the key explicitly in python, that's one less thing to go wrong.)

Even if I choose to pass the key as an env var, I don't think it should be checked when the library is imported. What if I do:

import os
from openelectricity import OEClient

# get secret from disk, or AWS secrets manager etc
with open(api_key_path, 'r') as f:
    os.environ["OPENELECTRICITY_API_KEY"] = f.read().strip()

client = OEClient()

Also, what if I change the API key while the script is running? Or what if I want two keys at once for different queries? I don't have those use cases, but I would expect to be able to have two different clients to manage that kind of difference.