The functionality of the Smyth Runtime Environment is partitioned into several discrete Subsystems. Each subsystem is responsible for a specific domain of the SRE's capabilities, such as memory management, I/O, or security.
Each subsystem exposes its functionality through one or more services, which are in turn powered by the SRE's pluggable Connector architecture. This structure ensures a clean separation of concerns and allows for a high degree of modularity.
Here are the primary subsystems within the SRE:
Agent Manager: The heart of agent execution. Responsible for the agent lifecycle, performance monitoring, and component workflow orchestration.
Memory Manager: Provides intelligent state and context management for agents, including multi-tiered caching and conversation history.
LLM Manager: A powerful abstraction layer for various LLM providers, handling API variations, smart inference, and response caching.
IO Subsystem: The gateway to the outside world. It provides unified connector interfaces for all input/output operations like Storage, VectorDBs, and Logging.
Security Subsystem: Manages all security-related primitives, including the Vault service for secret management and the Account service for identity and authentication.