> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mindlab.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Flow Engine

> The stateful execution layer that ensures every workflow is executed with precision, predictability, and resilience.

Underpinning the Orchestrator and the Context Spine is the Flow Engine, the stateful execution layer that ensures every workflow is executed with precision, predictability, and resilience.

The Flow Engine is designed to provide enterprise-grade guarantees about the state of your AI-driven work. It is not a "best-effort" system; it is a deterministic engine that provides full control over side effects and enables powerful debugging and auditing capabilities.

### Core Architectural Features

<CardGroup cols={2}>
  <Card title="Deterministic DAGs" icon="diagram-project">
    Every workflow is executed as a deterministic Directed Acyclic Graph (DAG) or state machine. This ensures that for a given input, the outcome will be the same every time.
  </Card>

  <Card title="Guarded Side Effects" icon="shield-halved">
    Any interaction with external systems is treated as a guarded side effect. The Flow Engine includes robust mechanisms for timeouts, retries, and error handling, ensuring that the system is resilient to external failures.
  </Card>

  <Card title="Time-Travel Debugging" icon="clock-rotate-left">
    The stateful nature of the Flow Engine allows for replays and diffs of workflow executions. This "time-travel debugging" is a powerful capability for understanding and resolving issues.
  </Card>

  <Card title="Determinism" icon="anchor">
    We provide the ability to use seeded decoding and policy snapshots to ensure reproducible runs, which is critical for debugging, auditing, and compliance.
  </Card>
</CardGroup>