RHAPSODY

Runtime for Heterogeneous Applications, Service Orchestration and Dynamism

RHAPSODY is a high-performance runtime system designed for orchestrating complex, heterogeneous workflows that combine traditional HPC tasks with modern AI/ML inference services.

Core Objective

RHAPSODY aims to provide a unified, asynchronous and scalable runtime capabilities for managing AI-HPC workloads and workflows across diverse computing infrastructures, from local workstations to large-scale HPC machines.

Key Features

• Heterogeneous Task Support: Seamlessly mix and scale ComputeTask (HPC/Binary) and AITask (Inference) in a single workflow.
• Asynchronous API: Built on Python's asyncio, allowing for high-throughput task submission and non-blocking state monitoring.
• Extensible Backends: Scale to large number of nodes with multiple execution backends including Dragon, Dask, RADICAL-Pilot, and local concurrent workers.
• Integrations: Integrates with with highly scalable workflow systems such radical.asyncflow and agentic frameworks such as flowgentic

Quick Example

import asyncio
from rhapsody import Session, ComputeTask
from rhapsody.backends import DragonExecutionBackendV3

async def main():
    # 1. Initialize session with a backend
    backend = await DragonExecutionBackendV3(num_workers=2048)
    async with Session(backends=[backend]) as session:

        # 2. Define a task
        task = ComputeTask(executable="/bin/echo", arguments=["Hello, Rhapsody!"])

        # 3. Submit and wait
        await session.submit_tasks([task])
        result = await task

        print(f"Task finished with state: {task.state}")
        print(f"Output: {task.return_value}")

if __name__ == "__main__":
    asyncio.run(main())

Getting Started

Ready to dive in? Check out our Installation Guide or jump straight into the Quick Start.