Astrux Python SDK

The official Python SDK for interacting with the Astrux API. Make machine learning predictions in just a few lines of code.

Installation

pip install astrux

Quick Start

from astrux import Astrux

# Initialize the client
client = Astrux(api_key="sk_your_api_key")

# Make a prediction
result = client.models.predict(
    model="my-model",
    input={"feature1": 10, "feature2": 20}
)

print(result["score"])  # Prediction score

Configuration

API Key

You can provide your API key in two ways:

1. As a parameter (recommended):

client = Astrux(api_key="sk_your_api_key")

2. Via environment variable:

export ASTRUX_API_KEY=sk_your_api_key

client = Astrux()  # Automatically loads from ASTRUX_API_KEY

Custom Timeout

The default timeout is 30 seconds. You can customize it:

client = Astrux(api_key="sk_your_api_key", timeout=60.0)

Usage

Simple Prediction

from astrux import Astrux

client = Astrux(api_key="sk_your_api_key")

# Regression
result = client.models.predict(
    model="house-prices",
    input={
        "sqft": 120,
        "bedrooms": 3,
        "city": "Paris"
    }
)
print(f"Estimated price: {result['score']}")

Prediction with Specific Version

# Use a specific model version
result = client.models.predict(
    model="sentiment-analysis",
    input={"text": "This product is excellent!"},
    version=2
)

Classification

For classification tasks, the response includes the predicted class and probabilities:

result = client.models.predict(
    model="image-classification",
    input={"image_url": "https://example.com/image.jpg"}
)

print(f"Class: {result['class_']}")  # Predicted class
print(f"Probabilities: {result['proba']}")  # List of probabilities

Model Metadata

The response may include metadata about the model used:

result = client.models.predict(
    model="my-model",
    input={"feature": 42}
)

print(f"Model ID: {result.get('model_id')}")
print(f"Model name: {result.get('model_name')}")
print(f"Version: {result.get('version')}")
print(f"Task type: {result.get('task_type')}")

Error Handling

The SDK provides specific exceptions for different error types:

from astrux import Astrux
from astrux._errors import (
    AuthenticationError,
    NotFoundError,
    ValidationError,
    RateLimitError,
    ServerError,
    AstruxError
)

client = Astrux(api_key="sk_your_api_key")

try:
    result = client.models.predict(
        model="my-model",
        input={"feature": "value"}
    )
except AuthenticationError as e:
    print(f"Authentication error: {e}")
except NotFoundError as e:
    print(f"Model not found: {e}")
except ValidationError as e:
    print(f"Invalid data: {e}")
    print(f"Details: {e.payload}")
except RateLimitError as e:
    print(f"Rate limit exceeded: {e}")
except ServerError as e:
    print(f"Server error: {e}")
except AstruxError as e:
    print(f"Astrux error: {e}")
    print(f"Status code: {e.status}")

Closing the Client

To release resources, close the client after use:

client = Astrux(api_key="sk_your_api_key")

try:
    result = client.models.predict(model="my-model", input={"x": 1})
    print(result)
finally:
    client.close()

Response Structure

Regression

{
    "score": 42.5,
    "model_id": "abc123",
    "model_name": "my-model",
    "version": 1,
    "task_type": "regression"
}

Classification

{
    "score": 0.95,
    "class_": "positive",  # Note: remapped from "class"
    "proba": [0.05, 0.95],
    "model_id": "def456",
    "model_name": "sentiment",
    "version": 2,
    "task_type": "classification"
}

Requirements

• Python >= 3.8
• httpx >= 0.27.0

Links

• Homepage: https://astrux.io
• Documentation: https://astrux.io/docs
• PyPI: https://pypi.org/project/astrux/

License

MIT - see LICENSE file for details.

Author

Thomas Bodénan - thomas.bodenan@gmail.com

Support

For questions or issues, consult the official documentation or contact Astrux support.