syft-hub SDK

Install

Prerequisites: Install SyftBox from syftbox.net

Install the syft-hub SDK using pip:

pip install syft-hub

Initialize

Create a client that connects to your SyftBox:

from syft_hub import Client

# Auto-detects SyftBox at ~/SyftBox
client = Client()
print(client)

Discover Available Data & Models

Scan the network and retrieve a list of available data sources and models. This might take a second, as it is verifying the availability of the services:

client.show_services()

Run Your First Query

Create a pipeline and add at least one preferred data source that you would like to fetch information from and an LLM to synthesize the response. Then, execute a federated query:

# Load a data source
data_source = client.load_service("irina@openmined.org/openmined-blog")

# Load an AI model
llm = client.load_service("aggregator@openmined.org/claude-sonnet-3.5")

# Create federated pipeline
pipeline = client.pipeline(
    data_sources=[data_source],
    synthesizer=llm
)

# Execute query
response = pipeline.run(messages=[
    {"role": "user", "content": "What is federated learning?"}
])