BSO - Boson Model Registry SDK

Python SDK for the Boson Model Registry - manage ML models, releases, and deployments.

Installation

pip install bso

Quick Start

from bso import Registry

# Initialize client
registry = Registry(
    base_url="https://your-registry.example.com/api",
    api_key="your-api-key"
)

# Or use environment variables
# REGISTRY_URL=https://your-registry.example.com/api
# REGISTRY_API_KEY=your-api-key
registry = Registry.from_env()

# List models
models = registry.list_models()

# Create a model
model = registry.create_model(
    name="my-model",
    storage_path="s3://bucket/models/my-model",
    description="My ML model"
)

# Create a release
release = registry.create_release(
    model_name="my-model",
    version="1.0.0",
    tag="v1.0.0",
    digest="sha256:abc123...",
    metadata={"accuracy": 0.95}
)

# Deploy a release
deployment = registry.deploy(
    release_id=release.id,
    environment="production"
)

# Get latest release for an environment
latest = registry.get_latest_release(
    model_name="my-model",
    environment="production"
)

Features

• Model management (create, list, get, delete)
• Release versioning with metadata
• Deployment tracking across environments
• Docker build triggering and status
• Artifact management
• Audit logging

Requirements

• Python 3.8+
• httpx
• pydantic

License

MIT