ADK ChatModelAgent 支持基于 tool 执行结果动态终止 ReAct 循环
Open
- Dominant language
- Go
- Stars
- 13k
- Forks
- 1.1k
- Avg merge
- 4h 6m
- Merged PRs (30d)
- 41
Description
在使用 ADK 的 NewChatModelAgent 构建 agent 时,希望能根据 tool 的执行结果动态决定是否终止 ReAct 循环并直接返回该 tool 的结果,而不是在构建 agent 时静态配置 ToolsConfig.ReturnDirectly 。
这是一个常见的需求场景,例如:
- tool 执行后发现结果已经是最终答案(如查询类 tool 命中缓存、检索类 tool 召回质量足够好)
- tool 执行后根据业务规则判断无需再调用 LLM 做后续推理
- 同一个 tool 在不同参数下,有时需要继续 ReAct、有时应该直接返回
Contributor guide
Research direction
Start at the ADK NewChatModelAgent entry point and the ToolsConfig.ReturnDirectly configuration, then trace how tool execution controls the ReAct loop. Done means a tool’s result can dynamically choose between continuing the loop and returning directly; the issue does not name implementation files or tests.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- go
- Domain
- ai
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 55/100