strictenv 0.1.0

Strictly typed environment variable parsing and validation with msgspec.

strictenv

strictenv is a fast, strictly typed environment variable loader built on top of msgspec. It gives you explicit schemas, predictable coercion, and runtime validation with a small API.

Install

uv add strictenv

Quickstart

from __future__ import annotations

from typing import Annotated

from msgspec import Struct

from strictenv import BaseSettings, Field, TransformStruct, transform


class Database(TransformStruct):
    host: str
    port: int

    @transform("host", mode="before")
    def normalize_host(value: str) -> str:
        return value.strip().lower()


class AppSettings(BaseSettings):
    debug: bool
    database: Database
    tenant_id: Annotated[str, Field(alias="TENANT")]

    model_config = {
        "env_prefix": "APP_",
        "case_sensitive": False,
        "env_nested_delimiter": "__",
        "env_file": ".env",
        "strict_env_file": True,
    }


settings = AppSettings.load()
AppSettings.write_env_example(".env.example")

Examples:

• APP_DEBUG=true -> debug: bool
• APP_DATABASE={"host":"localhost","port":5432} -> database: Database
• APP_DATABASE__HOST=localhost + APP_DATABASE__PORT=5432 -> nested parsing
• APP_TENANT=acme -> tenant_id via alias

model_config

Key	Type	Default	Description
env_prefix	str	""	Prefix applied to all environment keys.
case_sensitive	bool	False	When False, key lookup is case-insensitive.
env_nested_delimiter	str | None	None	Enables nested mapping like DB__HOST.
env_file	str | None	None	Path to a .env file to load first.
strict_env_file	bool	True	When True, invalid/missing .env files raise explicit errors.
max_nested_struct_depth	int | None	None	Maximum allowed depth for nested Struct traversal.

Field(...)

Field works both in Annotated[...] and as a default value:

from typing import Annotated
from strictenv import BaseSettings, Field

class AppSettings(BaseSettings):
    # Annotated metadata style
    retries: Annotated[int, Field(gt=0, lt=10)]

    # Default value style (alias + default + description)
    tenant_id: str = Field("acme", alias="TENANT", description="Tenant identifier")

    # Required when using `...`
    token: str = Field(...)

Supported quick validations:

• gt, ge, lt, le
• min_length, max_length

Description source priority for metadata/examples:

• Field(description=...) (highest priority)
• attribute docstring right below the field

@transform(...) And TransformStruct

Use @transform(field_name, mode="before" | "after") on classes that inherit from TransformStruct (including BaseSettings).

• before receives raw string input and may return:
  • another str (then normal coercion runs), or
  • a value already in target type.
• after receives already parsed value and must keep a compatible runtime type.

from strictenv import BaseSettings, TransformStruct, transform

class DatabaseConfig(TransformStruct):
    host: str
    port: int

    @transform("host", mode="before")
    def normalize_host(value: str) -> str:
        return value.strip().lower()

    @transform("port", mode="after")
    def keep_int(value: int) -> int:
        return value + 1

class AppSettings(BaseSettings):
    database: DatabaseConfig

Rules:

• field_name must be top-level in that class (no dotted paths).
• Multiple transforms run in definition order.
• Nested transforms apply only when nested type inherits TransformStruct.
• Nested settings can still use plain msgspec.Struct; use TransformStruct only when you need @transform.

Generate .env.example

BaseSettings.write_env_example(path) writes an empty env template for the schema. Field descriptions are emitted as comments:

class AppSettings(BaseSettings):
    debug: bool = Field(..., description="Enable debug logs")
    tenant_id: str = Field(..., alias="TENANT", description="Tenant identifier")

AppSettings.write_env_example(".env.example")

Generated file:

# Enable debug logs
DEBUG=

# Tenant identifier
TENANT=

Value precedence

1. overrides argument in load(...)
2. env argument (or os.environ when env=None)
3. .env file configured with model_config["env_file"]
4. Field defaults in the settings struct

If no source provides a required field, MissingSettingError is raised. If env_file is configured but missing, EnvFileNotFoundError is raised. If env_file cannot be read, EnvFileReadError is raised. If a non-comment line in env_file is not valid KEY=VALUE, EnvFileFormatError is raised. If keys collide in case-insensitive mode, EnvKeyConflictError is raised. If nested struct depth exceeds max_nested_struct_depth, NestedStructDepthError is raised. With strict_env_file=False, .env file errors are tolerated and invalid lines are skipped.

Coercion rules

strictenv performs strict coercion for:

• bool, int, float, str
• Enum (by member name or value)
• datetime, date, time
• timedelta (ISO8601, HH:MM[:SS], or numeric seconds)
• msgspec.Struct (from JSON string)
• list, dict, tuple, set, Mapping (from JSON string)
• Union / Optional (tries non-None members in order)

Invalid values raise ParseSettingError. There is no silent fallback to raw strings. Transform registration/execution failures raise TransformSettingError.

.env parser features:

• Optional export prefix (export KEY=value)
• Inline comments for unquoted values (KEY=value # comment)
• Quoted values with escapes and multiline support
• Variable expansion via ${VAR} (including references to earlier/later keys)

Differences vs pydantic-settings

• API is intentionally smaller and focused on msgspec.Struct.
• Compatibility is partial (supports familiar model_config, aliases, and nested env parsing).
• Automatic field description injection into msgspec.Meta is supported.

Development

uv sync --dev
uv run ruff check .
uv run mypy src
uv run pytest
uv build

Publish

uv publish --dry-run
uv publish