Skip to main content
The Runtime service executes workflow definitions durably. It tracks active runs, schedules tasks, routes context between steps, handles failures, and logs execution traces.

​
Key responsibilities

  • Durable Execution: Saves step progress to ensure that workflows can resume from the exact last-run state after system restarts or network drops.
  • Node Scheduling: Determines which steps to execute and coordinates parallel branches.
  • Failures & Retries: Handles step failures using configurable back-off and retry strategies.
  • Agent Invocation: Manages calls to LLMs and foundational models for AI agent steps.
  • Context Routing: Formats and passes input and output payloads between executing steps.
  • Auditing & Tracing: Logs execution traces, console output, and performance metrics for every run.

​
Workflow vs runtime

To distinguish the two services:
  • Workflow defines the logic schema (the blueprint).
  • Runtime manages the actual execution (the constructor and operations team).