chalkpy

What it is and what it does

Chalk is a feature engineering framework that lets you define machine learning features as Python functions and automatically orchestrates them into pipelines. You write resolvers—decorated functions that compute features from SQL, REST APIs, data warehouses, or other sources—and Chalk handles the orchestration, caching, scheduling, and deployment. It supports both online serving (low-latency feature queries for inference) and offline training (temporally-consistent historical feature sets for model development).

The package integrates with standard Python libraries (pydantic for schemas, requests for HTTP, SQL connectors for databases) and uses a Rust-based execution engine underneath for performance. It provides declarative caching policies, scheduled batch ingestion, reverse ETL to bring slow data sources online, and a REST API for querying features in production. The core idea is that the same source code serves both training and inference, ensuring feature consistency and reducing development overhead.

Use it for

• Define and deploy feature pipelines that compute user credit scores, fraud signals, or recommendation features from multiple databases and APIs.
• Build training datasets with temporal consistency—retrieve historical feature values as they existed at specific past timestamps for model training.
• Serve low-latency features to production models via REST API with declarative caching and staleness policies to balance freshness and cost.
• Ingest batch data from data warehouses on a schedule and automatically make it available for both offline analysis and online serving.
• Integrate vendor APIs and expensive external data sources with caching to reduce latency and per-call costs in online inference.
• Track model provenance and dataset lineage by storing and versioning feature computations alongside training data.

Worth the install?

AI-flagged interpretation of the facts on this page. Verify before relying on it.

Install

chalkpy on PyPI

Before you install

Low install friction with a pure-Python wheel. Active maintenance as of 2026-08-13. Depends on 16 runtime packages including grpcio, pydantic, and a Rust-backed chalkpy-rs component, which are all standard ML/data stack libraries.

Requires Python >=3.10,<3.15. Depends on chalkpy-rs, a compiled Rust component.

License in practice

License treatment is unclear—no SPDX identifier or raw license text is available. Verify the license terms before using in proprietary or commercial projects.

Quickstart

pip install chalkpy

from chalkpy import features, online

@features
class User:
    id: int
    name: str

@online
def get_user(uid: User.id) -> User:
    return User(id=uid, name="example")

from chalkpy import ChalkClient
result = ChalkClient().query(input={User.id: 1}, output=[User.name])

Package facts

Evidence: chalkpy-2.154.36-py3-none-any.whl

Tags

More Python Modules packages

See also azureml-featurestore · tecton · azureml-pipeline · feast · azure-ai-ml · azureml-pipeline-steps · matrice · azureml-pipeline-core · sagemaker-serve · sagemaker-mlops