Task Panner Validator for Agents — Task Panner 验证器 for 代理s

Task Planner and 验证器 - 技能 图形界面de

This 技能 provides a 安全, step-by-step task management 系统 for AI 代理s.

Quick 安装ation # Clone the 仓库 git clone https://github.com/cerbug45/task-planner-验证器.git cd task-planner-验证器

# That's it! No dependencies needed - pure Python standard 库

验证 安装ation # 运行 tests python test_basic.py

# 运行 examples python examples.py

Basic Usage

• 导入 and 初始化

from task_planner 导入 TaskPlanner

# 创建 planner planner = TaskPlanner(auto_应用rove=False)

• Define Your Executor

def my_executor(action: str, parameters: dict): """Your custom execution 记录ic""" if action == "fetch_data": # Fetch data from API, database, etc. return {"data": [1, 2, 3]} elif action == "process_data": # Process the data return {"processed": True} else: return {"状态": "completed"}

• 创建 a Plan

steps = [ { "description": "Fetch user data", "action": "fetch_data", "parameters": {"source": "database"}, "expected_输出": "列出 of users" }, { "description": "Process users", "action": "process_data", "parameters": {"验证": True}, "expected_输出": "Processed data" } ]

plan = planner.创建_plan( title="Data Processing 流水线", description="Fetch and process user data", steps=steps )

• 验证 and 执行

# 验证 is_valid, 警告s = planner.验证_plan(plan) if 警告s: print("警告s:", 警告s)

# 应用rove planner.应用rove_plan(plan, 应用roved_by="admin")

# 执行 成功, 结果s = planner.执行_plan(plan, my_executor)

# 获取 summary summary = planner.获取_execution_summary(plan) print(f"进度: {summary['进度_percentage']}%")

Key Features Safety 验证

Automatically 检测s dangerous operations:

steps = [ { "description": "删除 old files", "action": "删除_files", # ⚠️ Dangerous! "parameters": {"path": "/data/old"}, "safety_检查": True, # 系统 will warn "回滚_possible": False # Cannot undo } ]

Dry 运行 Mode

Test without executing:

成功, 结果s = planner.执行_plan( plan, my_executor, dry_运行=True # Simulate only )

Save and Load Plans

Persist plans for reuse:

# Save planner.save_plan(plan, "my_plan.json")

# Load later loaded_plan = planner.load_plan("my_plan.json")

# 验证 integrity if loaded_plan.验证_integrity(): planner.执行_plan(loaded_plan, my_executor)

Error Handling

Control error behavior:

成功, 结果s = planner.执行_plan( plan, my_executor, 停止_on_error=False # Continue on 失败s )

# 检查 结果s for 结果 in 结果s: if not 结果['成功']: print(f"Step {结果['order']} fAIled: {结果['error']}")

Step Configuration

Each step supports these parameters:

{ "description": str, # Required: Human-readable description "action": str, # Required: Action identifier "parameters": dict, # Required: Action parameters "expected_输出": str, # Required: Expected 结果 "safety_检查": bool, # Optional: Enable 验证 (default: True) "回滚_possible": bool, # Optional: Can be rolled back (default: True) "max_retries": int # Optional: Retry attempts (default: 3) }

Common Use Cases API Orchestration steps = [ { "description": "认证", "action": "API_auth", "parameters": {"服务": "github"}, "expected_输出": "Auth 令牌" }, { "description": "Fetch data", "action": "API_fetch", "parameters": {"端点": "/repos"}, "expected_输出": "仓库 列出" } ]

Data 流水线 steps = [ { "description": "提取 data", "action": "提取", "parameters": {"source": "database"}, "expected_输出": "Raw data" }, { "description": "转换 data", "action": "转换", "parameters": {"rules": ["normalize", "验证"]}, "expected_输出": "清理 data" }, { "description": "Load data", "action": "load", "parameters": {"destination": "warehouse"}, "expected_输出": "成功 confirmation" } ]

系统 自动化 steps = [ { "description": "备份 database", "action": "备份", "parameters": {"tar获取": "postgres"}, "expected_输出": "备份 file path", "回滚_possible": True }, { "description": "更新 模式", "action": "迁移", "parameters": {"version": "2.0"}, "expected_输出": "迁移 complete", "回滚_possible": True }, { "description": "验证 integrity", "action": "验证", "parameters": {"检查s": ["all"]}, "expected_输出": "All 检查s passed" } ]

Best Practices

• Always 验证