AI

These interfaces enable integrations into key AI packages, such as PyTorch . A Common use is for enhanced data loading.

Python Reference

torch.DragonDataset

This is a PyTorch dataset that utilizes the dragon distributed dictionary to store the training data and labels.

collective_group

Tools for creating a ProcessGroup for NCCL-like applications

Inference

Distributed, multi-GPU and multi-node LLM inference with a pull-based load balancing component managed through RDMA-enabled shared Dragon Queues, dynamic batching, optional prompt guardrails, and a vLLM backend. See the Inference for the full API reference.

Agent Framework

A multi-agent orchestration system for executing LLM-powered DAG workflows on HPC clusters. See the Agent Framework for the full API reference.