How does the workflow stop mechanism work?

It uses a Redis key-value pair to signal a stop request. The workflow engine polls this key at a high frequency (100ms) and terminates execution before starting the next node if the signal is present.

Why is there a 60-second TTL on stop signals?

The 60-second Time-To-Live prevents 'key leakage' and memory bloat in Redis if a workflow fails to manually delete its stop signal after termination.

Does this skill stop HTTP nodes mid-request?

While it terminates LLM streaming immediately, external HTTP requests may complete their current cycle before the workflow checks the stop status at the next node boundary.

What happens if Redis fails during a stop request?

The system includes fallback logic that defaults to 'continue' (false) to ensure workflow reliability, while logging the error for system monitoring.

FastGPT Workflow Lifecycle Control

Name: FastGPT Workflow Lifecycle Control
Author: labring

bylabring

•

26,978

•

生产力与工作流

Implements a robust Redis-based mechanism for gracefully stopping and managing AI agent workflows in real-time.

This skill provides a comprehensive design and implementation framework for the FastGPT ecosystem to handle workflow interruptions. It utilizes a Redis-based state management system to synchronize 'stop' signals across distributed environments, ensuring that long-running AI tasks can be terminated gracefully. The skill includes standardized Redis key structures, TTL-based cleanup to prevent memory leaks, and integration patterns for workflow dispatchers to check status before executing subsequent nodes. It is ideal for developers building complex AI orchestration platforms that require precise control over execution cycles and resource consumption.

主要功能

0126,978 GitHub stars

02Sub-100ms polling mechanism for responsive workflow interruption

03Comprehensive error handling and fallback strategies for state management

04Standardized API schema for cross-platform stop command integration

05Graceful termination logic with automatic TTL-based cleanup

06Redis-driven status synchronization for distributed workflows

使用场景

01Interrupting long-running LLM generation to save tokens and costs

02Implementing manual 'Cancel' buttons in AI chat interfaces

03Managing resource cleanup in complex multi-step agentic workflows

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add labring/fastgpt core

For use in Claude.ai and ChatGPT

Download Skill