Canvas Agent

Canvas Agent - 画布编排智能体

概述

Canvas Agent 是工作流画布的编排专家，能够理解用户需求并自动完成画布的设计和搭建。

画布位置

画布存储在 canvas_files/workflows/ 目录下。

执行方式

使用 bash 执行Python脚本。脚本位于 skills/canvas-agent/scripts/

脚本调用示例

1. 列出组件

python skills/canvas-agent/scripts/list_components.py --search API --limit 10

返回: {"total": N, "items": [{"path": "类别/名称", "cat": "类别", "desc": "描述", "in": ["输入端口"], "out": ["输出端口"]}]}

2. 获取组件详情

python skills/canvas-agent/scripts/get_component.py --full-path "网络请求/API分页查询器"

返回: {"path": "", "cat": "", "name": "", "desc": "", "in": [], "out": [], "props": []}

3. 创建画布

python skills/canvas-agent/scripts/create_canvas.py --name "我的工作流"

返回: {"name": "", "path": "", "workflow": ""}

4. 添加节点 (必须保存返回的id!)

python skills/canvas-agent/scripts/add_node.py --canvas-path "canvas_files/workflows/我的工作流_xxx" --component "触发器/触发器" --node-name "开始" --x 100 --y 100

返回: {"id": "0x123...", "name": "开始", "path": "", "pos": [100, 100], "ports": {"in": ["input"], "out": ["output"]}}

5. 连接节点

python skills/canvas-agent/scripts/connect_nodes.py --canvas-path "xxx" --from-node "0x123..." --from-port "output" --to-node "0x456..." --to-port "input"

6. 设置属性

python skills/canvas-agent/scripts/set_property.py --canvas-path "xxx" --node-id "0x123..." --property "mode" --value "auto"

7. 设置输入

python skills/canvas-agent/scripts/set_input.py --canvas-path "xxx" --node-id "0x123..." --input-name "url" --value "https://api.example.com"

8. 获取画布

python skills/canvas-agent/scripts/get_canvas.py --canvas-path "xxx"

9. 列出画布

python skills/canvas-agent/scripts/list_canvases.py

常用内置节点

• 触发器/触发器 - 手动/Webhook/定时触发
• 控制流/条件分支 - 条件分支
• 控制流/循环迭代 - 循环处理
• 控制流/并行执行 - 并行分支
• 数据/变量节点 - 定义变量
• 代码执行/代码编辑 - Python代码

关键约束

1. add_node返回的id必须保存 - 用于后续连接和配置
2. 连接前确认端口名 - 用get_component查看in/out
3. 属性/输入值支持JSON - 如 --value '{"key": "val"}'
4. canvas-path格式 - canvas_files/workflows/画布名_时间戳