orion-py-client

What it is and what it does

Orion Python Client is a lightweight wrapper for pushing ML model features from offline data sources (Hive/Delta tables, cloud-stored Parquet/Delta folders) into Orion Feature Store. It handles feature metadata retrieval, protobuf serialization of feature values, and Kafka message production to enable asynchronous feature ingestion pipelines.

The package is designed for teams running feature engineering workflows on Spark who need to serialize and stream features to a centralized feature store. It supports scalar types (FP32, FP64, Int32, Int64, UInt32, UInt64, String, Bool) and vector types, with Kafka acting as the transport layer. You initialize a client with metadata source credentials, call methods to fetch feature schemas and generate protobuf-encoded DataFrames, then push those to Kafka topics for downstream consumption.

Use it for:

• Push batch-computed features from Spark DataFrames to Orion Feature Store via Kafka for model serving.
• Retrieve feature metadata and schema information from Orion to validate offline feature computation pipelines.
• Serialize ML feature vectors into protobuf format for efficient Kafka transport.
• Integrate offline feature engineering on Delta/Hive tables with a centralized feature store backend.
• Automate feature ingestion workflows that read from cloud storage and publish to Orion.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

A Python client for pushing ML feature data from offline sources to Orion Feature Store and retrieving feature metadata via protobuf serialization and Kafka integration.

Yes, if you are already running Spark-based feature engineering and have Orion Feature Store deployed as your backend. The low install friction and permissive license make it a straightforward addition. However, the aging maintenance status (343 days since last release) means you should verify that it remains compatible with your Spark, Kafka, and Orion versions before committing to production use.

Install

orion-py-client on PyPI

pip install orion-py-client

uv add orion-py-client

poetry add orion-py-client

Installing orion-py-client

Before you install

Low install friction with only 2 runtime dependencies (cloudpathlib and protobuf). Maintenance status is aging—last release was 343 days ago—so expect slower response to issues or updates.

License in practice

Licensed under Apache-2.0 (permissive), so you can use, modify, and distribute this package freely in commercial or private projects without copyleft obligations.

Quickstart

pip install orion-py-client==0.1.14

from orion_py_client import OrionPyClient

client = OrionPyClient(
    features_metadata_source_url="your_features_metadata_source_url",
    job_id="your_job_id",
    job_token="your_job_token"
)

offline_src_type_columns, offline_col_to_default_values_map, entity_column_names = client.get_features_details()

Requires Python 3.7+; Kafka feature push functionality requires Apache Spark 3.0+ and spark-sql-kafka.

Verify before relying

• Whether the package actively maintains compatibility with recent versions of Apache Spark and Kafka ecosystems.
• Whether cloudpathlib and protobuf version constraints are documented or if any known incompatibilities exist.
• Whether the Orion Feature Store backend this client targets is actively maintained and production-ready.

Package facts

Evidence: orion_py_client-0.1.14-py3-none-any.whl

Tags

More Database packages