Support manual retry for failed workflows

Desired Behavior

It would be great to support manually retrying a failed workflow execution (or a failed workflow step) without restarting the entire workflow from the beginning.

In many real-world workflows, a workflow may contain multiple long-running or expensive steps. If one step fails near the end, the workflow is marked as failed and there is currently no simple way for users to manually retry it.

Ideally, OpenWorkflow could support:

manually retrying a failed workflow execution
resuming from the failed step instead of replaying all previous successful steps
retrying a specific failed step
preserving workflow history and outputs from completed steps

This would be especially useful for:

document processing pipelines
data import/export jobs
ETL workflows
multi-stage asynchronous tasks
workflows involving external APIs
Example Usage
// retry an entire failed workflow
await workflow.retry(workflowId)

// retry from the failed step
await workflow.retry(workflowId, {
fromFailedStep: true,
})

// retry a specific step
await workflow.retryStep(workflowId, "generateReport")

Example scenario:

uploadFile
→ validateFile ✅
→ parseData ✅
→ generateReport ❌
→ sendNotification

Instead of restarting the entire workflow, the user could manually retry only the generateReport step from the UI.

This would improve the user experience for long-running workflows and help avoid unnecessary recomputation and repeated external API calls.

L’issue non indica file sorgente, test o punti di ingresso; inizia individuando i punti di ingresso dell’esecuzione del workflow e della gestione degli step. Per considerare il lavoro completato sarebbero necessari un design deciso e un’implementazione per ritentare le esecuzioni o gli step non riusciti, preservando il lavoro completato, la cronologia e gli output.