Module workflow

Workflow Execution Modes in Floxide

Floxide workflows can be executed in several modes, each suited to different use cases:

1. Monolithic (Single-Process) Execution

• Method: [run]
• Description: Runs the entire workflow from start to finish in a single process, with no checkpointing or distributed coordination.
• Use Case: Simple, local workflows where failure recovery and distributed scaling are not needed.

2. Checkpointed (Resumable) Execution

• Method: [run_with_checkpoint]
• Description: Runs the workflow, checkpointing state after each step. If interrupted, can be resumed from the last checkpoint.
• Use Case: Long-running workflows, or those that need to recover from process crashes or restarts.

3. Resuming from Checkpoint

• Method: [resume]
• Description: Resumes a workflow run from its last checkpoint, restoring context and queue from the store.
• Use Case: Recovery after failure, or continuing a workflow after a pause.

4. Distributed Execution Primitives

• a. Orchestration/Seeding
  • Method: [start_distributed]
  • Description: Seeds the distributed workflow by checkpointing the initial state and enqueuing the first work item(s). Does not execute any steps.
  • Use Case: Used by an orchestrator to start a new distributed workflow run, preparing it for workers to process.
• b. Worker Step
  • Method: [step_distributed]
  • Description: A distributed worker dequeues a work item, loads the latest checkpoint, processes the node, enqueues successors, and persists state. Returns output if a terminal node is reached.
  • Use Case: Used by distributed workers to process workflow steps in parallel, with coordination via distributed queue and checkpoint store.

These distributed methods are the core primitives for building scalable, fault-tolerant workflow systems in Floxide.

Traits

WorkItem

Trait for a workflow work item.

Workflow

Trait for a workflow.