AI SOP Engine

The AI SOP (Standard Operating Procedure) Engine enables defining and executing automated workflows through YAML configuration.

Features

YAML-based Definition: Define workflows declaratively
Multiple Step Types: tool (call HertzBeat tools), llm (AI reasoning)
Unified Output: Consistent result structure with SopResult
I18n Support: Multi-language output (zh/en)
Multiple APIs: Streaming (SSE), Sync (JSON), AI-friendly (Text)

Quick Start

1. Define a Skill

Create skills/my_skill.yml:

yaml

name: my_skill
description: "My custom skill"
version: "1.0"

output:
  type: report      # report/simple/data/action
  format: markdown
  language: zh      # zh/en

steps:
  - id: get_data
    type: tool
    tool: queryMonitors
    args:
      status: 9
      
  - id: analyze
    type: llm
    prompt: |
      Analyze: ${get_data}

2. Execute

bash

# Streaming (SSE)
POST /api/ai/sop/execute/{skillName}

# Sync (JSON)
POST /api/ai/sop/execute/{skillName}/sync

# AI Format (Text)
POST /api/ai/sop/execute/{skillName}/ai

Output Types

Type	Use Case
`report`	Daily inspection, analysis
`simple`	Restart, clear cache
`data`	Query, statistics
`action`	Pending confirmation

Architecture

YAML Definition → SkillRegistry → SopEngine → Executors → SopResult
                                      ↓
                               ToolExecutor / LlmExecutor
                                      ↓
                               ToolRegistry (auto-discover @Tool methods)

Adding New Tools

Just add @Tool annotation to your method - no other code changes needed:

java

@Tool(name = "myNewTool", description = "...")
public String myNewTool(@ToolParam(...) Long param) {
    // implementation
}

ToolRegistry auto-discovers all @Tool methods at runtime.