rapha-ai 0.1.0

The official SDK for Rapha Protocol: Compute-to-Data AI Training

Rapha AI SDK

The official Python SDK for the Rapha Protocol—a decentralized "Compute-to-Data" network for AI model training over sensitive health data.

Instead of bringing data to the model, Rapha brings your model to the data using TEEs (Trusted Execution Environments) and ZK-TLS cryptography.

Installation

Install the package directly from PyPI:

pip install rapha-ai

Quickstart

Initialize the RaphaClient, fund the escrow contract (for compute nodes), and execute remote training over HIPAA-compliant medical datasets securely.

import torch
import torch.nn as nn
from rapha import RaphaClient

# 1. Define your PyTorch Model Architecture
class MedicalNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc1 = nn.Linear(3, 10)
        self.fc2 = nn.Linear(10, 1)
        
    def forward(self, x):
        x = torch.relu(self.fc1(x))
        return self.fc2(x)

model = MedicalNet()

# 2. Initialize the Rapha Client
# node_url defaults to production: https://api.rapha.ltd
# For local testing, pass: node_url="http://127.0.0.1:8000"
client = RaphaClient(escrow_contract_address="0xYourContractAddress")

# 3. Fund your job with USDC
job_id = client.fund_job(amount=100.0)
print(f"Funded Job: {job_id}")

# 4. Trigger remote training
# The model weights are packaged, sent to the node, computed against the dataset,
# and returned with a valid Zero-Knowledge proof. 
zk_proof = client.train(model, target_dataset_id="hospital_dataset_1")

# 5. Model state is updated in-place automatically
print("Training Complete. Model weights updated securely.")

# 6. Push proof to on-chain settlement
client.settle(zk_proof)
print("Node provider paid. Transaction complete.")

Learn More

Visit rapha.ltd for protocol documentation and enterprise features.