Implementation:Dagster io Dagster Config Class

Knowledge Sources	Dagster Dagster Docs
Domains	Data_Engineering, Configuration
Last Updated	2026-02-10 00:00 GMT

Overview

Concrete configuration class system for runtime parameterization provided by the Dagster core library.

Description

The Config class is Dagster's base class for defining typed runtime configuration. It extends Pydantic's BaseModel (via MaskedPermissiveConfig), allowing users to define configuration schemas using standard Python type annotations. When an asset or op function declares a parameter of a Config subclass type, Dagster automatically validates user-provided configuration against the schema and injects the validated object at runtime.

The Config system integrates with the Dagster UI to generate configuration form fields, and with the run launcher to validate configuration before execution begins.

Usage

Import Config from the dagster package. Define a subclass with typed fields representing runtime parameters. Declare a parameter of the Config subclass type in your asset or op function. The framework handles validation, UI generation, and injection.

Code Reference

Source Location

Repository: dagster
File: python_modules/dagster/dagster/_config/pythonic_config/config.py

Signature

class Config(MaskedPermissiveConfig):
    """Base class for Dagster configuration. Subclass this to define typed config schemas."""
    pass

# Usage pattern:
class MyConfig(dg.Config):
    param1: str
    param2: int = 10
    param3: float = 0.5

Import

from dagster import Config

# or

import dagster as dg

class MyConfig(dg.Config):
    ...

I/O Contract

Inputs

Name	Type	Required	Description
(user-defined fields)	Any Python type	Varies	Fields defined on the Config subclass. Types are validated by Pydantic. Fields without defaults are required; fields with defaults are optional.

Outputs

Name	Type	Description
(injected object)	Config subclass instance	A validated configuration object injected as a parameter into the asset or op function at runtime.

Usage Examples

Basic Configuration

import dagster as dg

class SampleConfig(dg.Config):
    sample_size: int = 1000
    include_headers: bool = True

@dg.asset
def sampled_data(config: SampleConfig) -> dg.MaterializeResult:
    # config.sample_size and config.include_headers are validated and available
    data = load_data(limit=config.sample_size, headers=config.include_headers)
    return dg.MaterializeResult(metadata={"rows": len(data)})

AI Query Configuration

import dagster as dg

class AskAI(dg.Config):
    question: str

@dg.asset(deps=["embedded_docs"], kinds={"openai"})
def answer(config: AskAI, openai: OpenAIResource) -> dg.MaterializeResult:
    client = openai.get_client()
    response = client.chat.completions.create(
        model="gpt-4-turbo-preview",
        messages=[{"role": "user", "content": config.question}],
    )
    return dg.MaterializeResult(
        metadata={"answer": response.choices[0].message.content}
    )

Configuration with Nested Types

import dagster as dg

class ModelConfig(dg.Config):
    model_name: str = "random_forest"
    learning_rate: float = 0.01
    max_depth: int = 10
    feature_columns: list[str] = ["col_a", "col_b"]

@dg.asset(kinds={"python", "sklearn"})
def trained_model(config: ModelConfig) -> dg.MaterializeResult:
    model = train_model(
        name=config.model_name,
        lr=config.learning_rate,
        depth=config.max_depth,
        features=config.feature_columns,
    )
    return dg.MaterializeResult(
        metadata={"model": config.model_name, "accuracy": model.score}
    )

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment