Skip to content
This repository was archived by the owner on Mar 11, 2026. It is now read-only.

Commit 8364941

Browse files
author
astrbot-docs-agent[bot]
committed
docs: update for AstrBotDevs/AstrBot#5484
1 parent dd95160 commit 8364941

4 files changed

Lines changed: 44 additions & 2 deletions

File tree

en/config/agent-runners/astrbot-agent-runner.md

Lines changed: 11 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -6,3 +6,14 @@ By default, AstrBot uses the built-in Agent Runner as the default executor. You
66

77
With the built-in Agent Runner, you can use AstrBot's [MCP Server](/use/mcp), [Knowledge Base](/use/knowledge-base), [Web Search](/use/websearch), and persona features.
88

9+
## Follow-up Message Handling
10+
11+
When the Agent is executing tool calls (such as web search, code execution, etc.), new messages sent by the user are automatically captured and queued. These follow-up messages are injected into the context when the next tool result is returned, allowing the LLM to prioritize them.
12+
13+
This feature allows users to continue the conversation while the Agent is performing long-running tasks, without waiting for the current task to complete. The Agent will respond to your follow-up messages immediately after processing the current tool call.
14+
15+
**Example Use Cases:**
16+
17+
- While the Agent is performing a web search, you can send "also search for XXX"
18+
- While the Agent is calling the code executor, you can add new instructions or modify requirements
19+
- Multiple follow-up messages are processed in the order they were sent

en/use/function-calling.md

Lines changed: 11 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -46,6 +46,16 @@ Below are some common tool calling demos:
4646
![image](https://files.astrbot.app/docs/source/images/function-calling/image-1.png)
4747

4848

49+
## Follow-up Messages During Tool Execution
50+
51+
When the Agent is executing tool calls, new messages you send are automatically captured and queued. These follow-up messages are injected into the context when the next tool result is returned, allowing the LLM to prioritize them.
52+
53+
For example, while the Agent is performing a web search, you can continue to send "also search for XXX", and the Agent will respond to your follow-up request immediately after processing the current search.
54+
55+
:::tip
56+
This feature is only available when using the AstrBot built-in Agent Runner.
57+
:::
58+
4959
## MCP
5060

51-
Please refer to this documentation: [AstrBot - MCP](/use/mcp).
61+
Please refer to this documentation: [AstrBot - MCP](/use/mcp).

zh/config/agent-runners/astrbot-agent-runner.md

Lines changed: 11 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -6,3 +6,14 @@
66

77
在内置 Agent 执行器下,您可以使用 AstrBot 的 [MCP 服务器](/use/mcp)[知识库](/use/knowledge-base)[网页搜索](/use/websearch)、人格功能。
88

9+
## 后续消息处理
10+
11+
当 Agent 正在执行工具调用(如网页搜索、代码执行等)时,用户发送的新消息会被自动捕获并排队。这些后续消息会在下一次工具结果返回时被注入到上下文中,让 LLM 能够优先处理这些消息。
12+
13+
这一特性使得用户可以在 Agent 执行长时间任务时继续对话,而无需等待当前任务完成。Agent 会在处理完当前工具调用后,立即响应您的后续消息。
14+
15+
**使用场景示例:**
16+
17+
- Agent 正在执行网页搜索时,您可以发送"顺便也搜索一下 XXX"
18+
- Agent 正在调用代码执行器时,您可以补充新的指令或修改需求
19+
- 多条后续消息会按发送顺序依次处理

zh/use/function-calling.md

Lines changed: 11 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,7 @@ outline: deep
88

99
函数调用旨在提供大模型**调用外部工具的能力**,以此实现 Agentic 的一些功能。
1010

11-
比如,问大模型:帮我搜索一下关于“猫”的信息,大模型会调用用于搜索的外部工具,比如搜索引擎,然后返回搜索结果。
11+
比如,问大模型:帮我搜索一下关于"猫"的信息,大模型会调用用于搜索的外部工具,比如搜索引擎,然后返回搜索结果。
1212

1313
目前,支持的模型包括但不限于
1414

@@ -46,6 +46,16 @@ outline: deep
4646
![image](https://files.astrbot.app/docs/source/images/function-calling/image-1.png)
4747

4848

49+
## 工具执行期间的后续消息
50+
51+
当 Agent 正在执行工具调用时,您发送的新消息会被自动捕获并排队。这些后续消息会在下一次工具结果返回时被注入到上下文中,让 LLM 能够优先处理。
52+
53+
例如,当 Agent 正在执行网页搜索时,您可以继续发送"顺便也搜索一下 XXX",Agent 会在处理完当前搜索后立即响应您的后续请求。
54+
55+
:::tip
56+
此功能仅在使用 AstrBot 内置 Agent 执行器时可用。
57+
:::
58+
4959
## MCP
5060

5161
请前往此文档 [AstrBot - MCP](/use/mcp) 查看。

0 commit comments

Comments
 (0)