Skip to content

README.md

Latest commit

 

History

History
196 lines (137 loc) · 4.6 KB

README.md

File metadata and controls

196 lines (137 loc) · 4.6 KB

Sift Client Library

This library provides a high-level Python client for interacting with Sift APIs. It offers both synchronous and asynchronous interfaces, strong type checking, and a Pythonic API design.

Installation

pip install sift-stack-py

Getting Started

Initializing the Client

You can initialize the Sift client with your API key and service URLs:

from sift_client import SiftClient
from datetime import datetime

# Initialize with individual parameters
client = SiftClient(
    api_key="your-api-key",
    grpc_url="your-sift-grpc-url",
    rest_url="your-sift-rest-url"
)

# Or use a connection configuration
from sift_client.transport import SiftConnectionConfig

config = SiftConnectionConfig(
    api_key="your-api-key",
    grpc_url="your-sift-grpc-url",
    rest_url="your-sift-rest-url"
)
client = SiftClient(connection_config=config)

The SiftConnectionConfig provides access to additional configuration options such as use_ssl and cert_via_openssl.

Using Synchronous and Asynchronous APIs

The Sift client provides both synchronous and asynchronous versions of all APIs. You can choose the one that best fits your application's needs.

Synchronous API

The synchronous API is perfect for scripts, notebooks, and applications that don't need asynchronous operation:

# Get an asset by ID
asset = client.assets.get(asset_id="asset123")

# List assets with filtering
assets = client.assets.list_(
    name_contains="example",
    created_after=datetime(2023, 1, 1),
    include_archived=False
)

# Find a single asset matching criteria
asset = client.assets.find(name="my-asset")

Asynchronous API

The asynchronous API is ideal for high-performance applications and services that need to make concurrent API calls:

import asyncio


async def get_asset_async():
    # Get an asset by ID asynchronously
    asset = await client.assets_async.get(asset_id="asset123")

    # Running Sync within async also works
    some_other_asset = client.assets.get(asset_id="asset456")

    return asset


# Run in an async context
asset = asyncio.run(get_asset_async())

Working with Sift Types

Sift types (like Asset, Run, etc.) are immutable Pydantic models that provide a convenient interface for working with Sift resources.

Accessing Properties

# Get an asset
asset = client.assets.get(asset_id="asset123")

# Access properties
print(f"Asset name: {asset.name}")
print(f"Created on: {asset.created_date}")
print(f"Tags: {', '.join(asset.tags)}")
print(f"Is archived: {asset.is_archived}")

Using Methods on Sift Types

Sift types have convenient methods for common operations. These methods use the synchronous API internally. Using these methods will update the instance in-place.

# Get an asset
asset = client.assets.get(asset_id="asset123")

# Archive the asset
asset.archive(archive_runs=True)

# Update the asset
asset.update({
    "tags": ["updated", "example"]
})

Note: Type methods only work with the synchronous API. If you need to use the asynchronous API, you should use the resource APIs directly.

Creating Update Models

For more complex updates, you can create update models (instead of a key-value dictionary):

from sift_client.sift_types.asset import AssetUpdate

# Create an update model
update = AssetUpdate(tags=["new", "tags"])

# Apply the update
asset = client.assets.update(asset="asset123", update=update)

# Or using the asset method
asset = client.assets.get(asset_id="asset123").update(update)

Advanced Usage

Working with Tags

Tags are a powerful way to organize and filter your assets:

# Add tags when updating an asset
asset.update({
    "tags": ["production", "model-v1", "trained"]
})

# Filter assets by tags
production_assets = client.assets.list_(
    tags=["production"]
)

Filtering Assets

The client provides various ways to filter different Sift types:

# Filter by name (exact match)
assets = client.assets.list_(name="my-model")

# Filter by name (contains)
assets = client.assets.list_(name_contains="model")

# Filter by name (regex)
assets = client.assets.list_(name_regex="model-v[0-9]+")

# Filter by creation date
assets = client.assets.list_(
    created_after=datetime(2023, 1, 1),
    created_before=datetime(2023, 12, 31)
)

# Filter by modification date
assets = client.assets.list_(
    modified_after=datetime(2023, 6, 1)
)

# Include archived assets
all_assets = client.assets.list_(include_archived=True)

# Limit the number of results
recent_assets = client.assets.list_(
    limit=10,
    order_by="modified_date desc"
)