[Bug]: 疑似planner输出的某个步骤没有携带“next” 跳转导致报错

[JanuaryHuang]

What happened?

输入指令为
run-l -t agent_workflow -u demo_user_002 -m "为电商搜索推荐搭建一个可执行的多智能体工作流。输入是自然语言查询（示例：'运动鞋，夏天，预算100加币'）。工作流目标：1) 解析查询抽取意图与约束（类目/季节/预算）；2) 从本地示例商品库或在线搜索获取候选；3) 生成个性化推荐并给出理由；4) 产出一个用于前端渲染的UI布局JSON（含卡片顺序与优先级）；5) 将完整结果写入 ./outputs/report.json；保存完成后立即结束工作流。"

log：

┏━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Role ┃ Content                                                                                                                                                        ┃
┡━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩
│ User │ 为电商搜索推荐搭建一个可执行的多智能体工作流。输入是自然语言查询（示例：'运动鞋，夏天，预算100加币'）。工作流目标：1)                                          │
│      │ 解析查询抽取意图与约束（类目/季节/预算）；2) 从本地示例商品库或在线搜索获取候选；3) 生成个性化推荐并给出理由；4)                                               │
│      │ 产出一个用于前端渲染的UI布局JSON（含卡片顺序与优先级）；5) 将完整结果写入 ./outputs/report.json；保存完成后立即结束工作流。                                    │
└──────┴────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘
╭──────── CoorAgent ─────────╮
│ Workflow execution started │
╰────────────────────────────╯


>>> coordinator starting execution...



<<< coordinator execution finished



>>> planner starting execution...


{
  "thought": "用户希望搭建一个电商搜索推荐的多智能体工作流，能够解析自然语言查询并生成个性化推荐。这个工作流需要多个步骤，包括意图解析、候选商品获取、推荐生成和结果输出
  "title": "电商搜索推荐多智能体工作流",
  "new_agents_needed": [],
  "steps": [
    {
      "agent_name": "EcommerceSearchRecommendationDemo",
      "title": "解析查询并提取意图与约束",
      "description": "使用EcommerceSearchRecommendationDemo智能体解析用户的自然语言查询，提取出类目、季节和预算等信息。",
      "note": "输入示例：'运动鞋，夏天，预算100加币'"
    },
    {
      "agent_name": "EcommerceSearchRecommendationDemo",
      "title": "获取候选商品",
      "description": "从本地示例商品库或在线搜索获取符合提取条件的候选商品。",
      "note": "确保候选商品符合类目、季节和预算的约束。"
    },
    {
      "agent_name": "EcommerceSearchRecommendationDemo",
      "title": "生成个性化推荐",
      "description": "基于候选商品生成个性化推荐，并给出推荐理由。",
      "note": "推荐理由应考虑用户的偏好和预算。"
    },
    {
      "agent_name": "coder",
      "title": "生成UI布局JSON",
      "description": "将推荐结果转换为前端渲染所需的UI布局JSON格式，包括卡片顺序与优先级。",
      "note": "确保JSON格式符合前端需求。"
    },
    {
      "agent_name": "reporter",
      "title": "保存结果到文件",
      "description": "将完整结果写入 ./outputs/report.json，并确保保存完成后结束工作流。",
      "note": "报告应包含所有推荐信息和生成的JSON布局。"
    }
  ]
}


<<< planner execution finished



>>> publisher starting execution...



<<< publisher execution finished



>>> agent_proxy starting execution...


{
  "category": "运动鞋",
  "season": "夏天",
  "budget": "100 加币"
}


<<< agent_proxy execution finished



>>> publisher starting execution...



<<< publisher execution finished



>>> agent_proxy starting execution...


{
  "candidates": [
    {
      "product_id": 1,
      "name": "轻便运动鞋",
      "price": 90,
      "season": "夏天",
      "category": "运动鞋",
      "description": "透气轻便，适合夏季运动。",
      "image_url": "http://example.com/product1.jpg"
    },
    {
      "product_id": 2,
      "name": "跑步鞋",
      "price": 85,
      "season": "夏天",
      "category": "运动鞋",
      "description": "专为跑步设计，舒适透气。",
      "image_url": "http://example.com/product2.jpg"
    },
    {
      "product_id": 3,
      "name": "休闲运动鞋",
      "price": 75,
      "season": "夏天",
      "category": "运动鞋",
      "description": "适合日常穿着，时尚舒适。",
      "image_url": "http://example.com/product3.jpg"
    }
  ]
}


<<< EcommerceSearchRecommendationDemo execution finished



>>> publisher starting execution...

Traceback (most recent call last):
  File "/Users/januaryh/GitHub/cooragent/src/workflow/process.py", line 196, in _process_workflow
    command = await node_func(state)
              ^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/januaryh/GitHub/cooragent/src/workflow/coor_task.py", line 92, in publisher_node
    agent = response["next"]
            ~~~~~~~~^^^^^^^^
KeyError: 'next'
⠸ Starting execution: publisher...2025-08-22 01:40:27 - src.workflow.process - ERROR - Error in Agent workflow: 'next'
╭───────── CoorAgent ──────────╮
│ Workflow execution finished! │
╰──────────────────────────────╯
CoorAgent> run-o -u demo_user_002
╭─────────────────────────────────────────────────────────────────────────────── Result ────────────────────────────────────────────────────────────────────────────────╮
│ No matching workflow found                                                                                                                                            │
╰───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
⠋ Fetched 0 workflow!
Select Command:
1 - Select workflow
2 - Exit
Enter part number: 2```

从日志结果来看，publisher_node 在读取 response["next"] 时抛了 KeyError: 'next'，说明 Planner 输出的某个步骤没有携带 “next” 跳转，Publisher 没法继续路由，于是崩掉。

### Distribution version

macOS ARM 64 (Apple Silicon)

### App Version

_No response_

### What browsers are you seeing the problem on if you're using web version?

_No response_

### Are you self-hosting?

- [ ] Yes

### Self-hosting Version

_No response_

### Relevant log output

```shell

Anything else?

No response