API Reference

The best place to begin with Dragon is with Python multiprocessing . Users move to the Dragon API directly as they find the need for more control, such as explicitly placing processes or objects on CPU and GPU nodes, or they want to use other features in Dragon like the in-memory distributed dictionary.

This reference is split into two layers. Most readers should begin with the User API, which exposes Dragon’s main capabilities through higher-level interfaces for multiprocessing, data movement, AI, telemetry, workflows, and native process orchestration. Move to the Low-level API when you need direct control over the underlying runtime objects, infrastructure services, or client abstractions that those user-facing APIs build on.

In practice, a simple rule works well: start with the User API unless you are building new Dragon features, debugging runtime internals, or integrating with lower-level services that are not surfaced in the higher-level interfaces.

User API

Choose the User API when you want to write or extend applications with Dragon without managing runtime internals directly. These pages are usually the right starting point for application developers.

• Python multiprocessing
  • DragonContext
  • Dragon multiprocessing Classes
• Data
  • Python Reference
  • C Reference
  • C++ Reference
• Telemetry
  • Python Reference
• AI
  • Python Reference
  • Inference
  • Agent Framework
• Workflows
  • Python Reference
• Native
  • Python Reference
  • C++ Reference
• Resource Placement and Affinity
  • Python Reference

Low-level API

Choose the Low-level API when you need to understand or control the runtime implementation itself, such as infrastructure objects, client interfaces, or core communication primitives.

• Infrastructure
  • Python Reference
  • C Reference
• Global Services Client
  • Python Reference
• Core
  • Python Reference
  • Low-level Python Reference
  • C Reference