qdhenry Workflow
Instructions
The Workflow component adds durable execution to Convex functions, allowing you to build reliable long-running processes that survive server restarts and handle failures gracefully. It combines queries, mutations, and actions into deterministic workflows with configurable retry policies, parallel execution, and reactive status monitoring. Workflows can run for months with custom delays, be canceled mid-execution, and provide real-time progress updates to multiple subscribers.
Installation
npm install @convex-dev/workflow
Use cases
• User onboarding flows that span multiple days with email verification, content generation, and follow-up messages that must complete reliably • Data processing pipelines that orchestrate multiple API calls, transformations, and database updates with automatic retries on transient failures • Payment and fulfillment workflows that coordinate between payment processors, inventory systems, and shipping APIs with precise error handling • Content moderation systems that combine AI analysis, human review queues, and notification steps with configurable delays between stages • Batch job orchestration that processes large datasets by coordinating multiple parallel workers with backpressure and failure recovery
How it works
You define workflows using workflow.define() with deterministic handler functions that orchestrate Convex queries, mutations, and actions through a step context. The component restricts access to non-deterministic APIs like fetch and patches others like Math.random() to ensure reproducible execution across server restarts.
Workflows are started from mutations or actions using workflow.start(), which returns a WorkflowId for tracking progress. Steps execute via ctx.runQuery(), ctx.runMutation(), and ctx.runAction() methods, with results passed between steps. Parallel execution happens through Promise.all() calls, while retry behavior is configured per-step or globally through workpool options.
The component provides reactive status monitoring through workflow.status(), cancellation with workflow.cancel(), and completion handling via onComplete callbacks. Failed workflows can be restarted from specific steps using workflow.restart(), and the system automatically manages step parallelism and resource cleanup.
When NOT to use
- When a simpler built-in solution exists for your specific use case
- If you are not using Convex as your backend
- When the functionality provided by Workflow is not needed