a2aproject / a2aproject/a2a-samples

[Bug]: ADK example returns inconsistent state

未关闭
#46 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
主要语言
Jupyter Notebook
星标
1.8k
派生
751
PR 合并指标
30 天内没有已合并 PR

描述

### What happened?

The Google ADK example agent sometimes returns status `completed` when it should return `input-required`. Attaching an example log where this occurred. Agent responds with text stating `Could you please provide the date, amount, and business justification/purpose for the plane ticket?` but status is set to `completed`

### Relevant log output

```shell
========= starting a new task ========

What do you want to send to the agent? (:q or quit to exit): can you reimburse me for my plane ticket
stream event => {"jsonrpc":"2.0","id":"55d1e4697a0c49da8eaf88314e441bc6","result":{"id":"dd6745735a954d13bfa2f0cf8ecd9bfe","status":{"state":"completed","message":{"role":"agent","parts":[{"type":"text","text":"Could you please provide the date, amount, and business justification/purpose for the plane ticket?\n"}]},"timestamp":"2025-04-14T12:02:43.554273"},"final":false}}
stream event => {"jsonrpc":"2.0","id":"55d1e4697a0c49da8eaf88314e441bc6","result":{"id":"dd6745735a954d13bfa2f0cf8ecd9bfe","artifact":{"parts":[{"type":"text","text":"Could you please provide the date, amount, and business justification/purpose for the plane ticket?\n"}],"index":0,"append":false}}}
stream event => {"jsonrpc":"2.0","id":"55d1e4697a0c49da8eaf88314e441bc6","result":{"id":"dd6745735a954d13bfa2f0cf8ecd9bfe","status":{"state":"completed","timestamp":"2025-04-14T12:02:43.554655"},"final":true}}
```

### Code of Conduct

- [x] I agree to follow this project's Code of Conduct

贡献指南

打开贡献指南

调研方向

Look at the Google ADK example agent code, likely in a Jupyter notebook or Python script. Find where the agent's status is set (state: 'completed' vs 'input-required'). Examine the logic that determines the status based on the agent's response. Run the example to reproduce the issue, then trace through the status-setting code to see why it incorrectly returns 'completed' when asking for more input.

由索引模型根据 Issue 内容生成。

评估

技术栈
python
领域
ai-infra-agents
Issue 类型
缺陷
难度
3/5
预计耗时
1-2 天
活跃度
停滞
描述清晰度
描述清楚
新手友好度
50/100

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。