Orion Python Client (Meesho)

0.1.14 · active · verified Thu Apr 16

The `orion-py-client` is a Python client library developed by Meesho, designed to interact with the Orion Feature Store. Its primary function is to facilitate the pushing and producing of model features, as well as retrieving features' metadata. It aims to integrate machine learning workflows with the Orion Feature Store platform.

Common errors

```
ModuleNotFoundError: No module named 'orion_py_client'
```
cause The `orion-py-client` package is not installed in the current Python environment.

fix Run `pip install orion-py-client` to install the package.
```
An error occurred: Authentication failed (or similar connection error)
```
cause The provided `ORION_API_KEY` is incorrect, expired, or the `ORION_API_URL` is inaccessible or wrong. The client could not establish a connection or authenticate with the Orion Feature Store.

fix Verify that `ORION_API_URL` points to the correct endpoint and `ORION_API_KEY` is valid and active. Check network connectivity to the Orion Feature Store. Consult your Orion Feature Store administrator for valid credentials.
```
AttributeError: 'OrionClient' object has no attribute 'push_feature'
```
cause The method name `push_feature` (or similar, like `produce_feature`) used in the quickstart example might not be the actual method name exposed by the `OrionClient` class. This could be due to inferred API names.

fix Refer to the actual source code or internal documentation of the `orion-py-client` to find the correct method names for interacting with features. If no documentation is available, inspect the `OrionClient` object's methods at runtime (e.g., using `dir(client)` or an IDE's autocomplete).

Warnings

gotcha Public documentation for `orion-py-client` is scarce, making precise API usage and potential breaking changes difficult to ascertain without access to the internal Meesho project. Information regarding detailed API methods, data models, and specific integration patterns is largely inferred from the PyPI summary.
Fix: Consult internal Meesho documentation or source code if available. Rely on PyPI summary and typical Python client patterns for initial integration.
breaking As a client library, breaking changes in the underlying Orion Feature Store API could lead to unexpected behavior or failures, even without a new client release. Always verify compatibility if the backend service is updated.
Fix: Monitor Orion Feature Store backend release notes. Implement robust error handling and API versioning if supported by the Feature Store.
gotcha Authentication relies on an API key and URL. Incorrect or expired credentials will result in connection failures or unauthorized access errors, which may not always be clearly distinguished without verbose logging from the service.
Fix: Ensure `ORION_API_URL` and `ORION_API_KEY` are correctly configured and have the necessary permissions. Implement retry logic for transient connection issues.

Install

pip install orion-py-client Install stable version

Imports

OrionClient
```
from orion_py_client import OrionClient
```
Feature
```
from orion_py_client.models import Feature
```
Assumed path based on common package structure for data models.

Quickstart

Initializes the OrionClient and demonstrates placeholder operations for pushing features and retrieving feature metadata. Replace environment variables with your actual Orion Feature Store URL and API Key. The exact API methods and data structures for feature interaction are inferred based on the library's summary, as detailed documentation is not publicly available.

import os
from orion_py_client import OrionClient

# Placeholder for Orion Feature Store URL and API Key
ORION_API_URL = os.environ.get('ORION_API_URL', 'http://localhost:8080')
ORION_API_KEY = os.environ.get('ORION_API_KEY', 'your_api_key_here')

try:
    client = OrionClient(api_url=ORION_API_URL, api_key=ORION_API_KEY)
    print(f"Successfully initialized OrionClient for {ORION_API_URL}")

    # Example: Push a feature (assuming a 'Feature' model exists)
    # This part is illustrative as exact data model and push method are not publicly documented.
    feature_data = {
        "feature_name": "user_login_count",
        "entity_id": "user_123",
        "value": 15,
        "timestamp": "2026-04-16T10:00:00Z"
    }
    # Assuming a method like 'push_feature' or 'produce_feature'
    # response = client.push_feature(feature_data)
    # print(f"Pushed feature: {response}")

    # Example: Get feature metadata
    # metadata = client.get_feature_metadata(feature_name="user_login_count")
    # print(f"Feature metadata: {metadata}")

except Exception as e:
    print(f"An error occurred: {e}")
    print("Please ensure ORION_API_URL and ORION_API_KEY are correctly set.")

view raw JSON →