Environment variables

DataCoolie does not define a fixed framework-wide environment-variable surface such as DATACOOLIE_* settings. In the current codebase, direct environment variable reads only happen in secret-resolution paths; most other configuration flows through metadata, constructor arguments, or the host runtime.

Direct environment lookups in the library

Pattern	Read by	When it is used
{source}{key}	EnvResolver	When secrets_ref uses an env:<source> resolver key.
{source}{key}	LocalPlatform	When LocalPlatform is the active native secret provider and secrets_ref uses an unprefixed source key.

In both cases, key is the current value already stored in the relevant configure field.

EnvResolver (env:<prefix>)

EnvResolver reads from os.environ using:

{prefix}{configure[field]}

Example:

{
    "configure": {
        "password": "DB_PASSWORD"
    },
    "secrets_ref": {
        "env:APP_": ["password"]
    }
}

This resolves configure["password"] from os.environ["APP_DB_PASSWORD"].

If you want no prefix, use env::

{
    "configure": {
        "password": "DB_PASSWORD"
    },
    "secrets_ref": {
        "env:": ["password"]
    }
}

This resolves from os.environ["DB_PASSWORD"].

LocalPlatform native provider

LocalPlatform is itself a native secret provider backed by os.environ. Without a resolver prefix, it also reads:

{source}{configure[field]}

Example:

{
    "configure": {
        "password": "DB_PASSWORD"
    },
    "secrets_ref": {
        "APP_": ["password"]
    }
}

With LocalPlatform, this resolves configure["password"] from os.environ["APP_DB_PASSWORD"].

Not a DataCoolie environment-variable contract

The following may still matter in some deployments, but they are not read by DataCoolie itself as named environment-variable settings:

• AWS SDK variables such as AWS_REGION, AWS_DEFAULT_REGION, AWS_PROFILE, AWS_ACCESS_KEY_ID, and related credentials are handled by boto3's own provider chain, not by a DataCoolie-specific lookup.
• Spark variables such as SPARK_HOME and PYSPARK_PYTHON are handled by the Spark runtime you launch or the SparkSession you pass in.
• Fabric and Databricks use notebookutils and dbutils in their notebook runtimes rather than a DataCoolie environment-variable layer.
• TZ is not consulted by the current library code; DataCoolie's internal log and runtime timestamps are built with explicit UTC-aware datetimes.