fix: preserve multi-yield tool results

Author: SunmiJJW

astrbot/core/agent/runners/tool_loop_agent_runner.py

@@ -1096,11 +1096,19 @@ async def _handle_function_tools(
         logger.info(f"Agent 使用工具: {llm_response.tools_call_name}")
 
         def _append_tool_call_result(tool_call_id: str, content: str) -> None:
+            content = self._merge_follow_up_notice(content)
+            if (
+                tool_call_result_blocks
+                and tool_call_result_blocks[-1].tool_call_id == tool_call_id
+            ):
+                previous = tool_call_result_blocks[-1]
+                previous.content = f"{previous.content}\n\n{content}"
+                return
             tool_call_result_blocks.append(
                 ToolCallMessageSegment(
                     role="tool",
                     tool_call_id=tool_call_id,
-                    content=self._merge_follow_up_notice(content),
+                    content=content,
                 ),
             )
 

tests/test_tool_loop_agent_runner.py

@@ -99,6 +99,24 @@ async def generator():
         return generator()
 
 
+class MultiYieldToolExecutor:
+    """Simulate a local tool that yields multiple results for one call."""
+
+    @classmethod
+    def execute(cls, tool, run_context, **tool_args):
+        async def generator():
+            from mcp.types import CallToolResult, TextContent
+
+            yield CallToolResult(
+                content=[TextContent(type="text", text="first streamed result")]
+            )
+            yield CallToolResult(
+                content=[TextContent(type="text", text="second streamed result")]
+            )
+
+        return generator()
+
+
 class LargeTextToolExecutor:
     """模拟返回超长文本的工具执行器"""
 
@@ -621,6 +639,34 @@ async def snapshot_context_manager(messages, trusted_token_usage=0):
     assert "工具执行结果" in tool_messages[0].content
 
 
+@pytest.mark.asyncio
+async def test_tool_loop_merges_multiple_results_from_one_tool_call(
+    runner, provider_request, mock_hooks
+):
+    """Multiple executor yields must remain one complete provider tool result."""
+    provider = CapturingToolLoopProvider("test_tool")
+
+    await runner.reset(
+        provider=provider,
+        request=provider_request,
+        run_context=ContextWrapper(context=None),
+        tool_executor=MultiYieldToolExecutor,
+        agent_hooks=mock_hooks,
+        streaming=False,
+    )
+
+    async for _ in runner.step_until_done(3):
+        pass
+
+    assert provider.call_count == 2
+    second_contexts = provider.received_contexts[1]
+    tool_messages = [msg for msg in second_contexts if msg.role == "tool"]
+    assert len(tool_messages) == 1
+    assert tool_messages[0].tool_call_id == "call_context_refresh"
+    assert "first streamed result" in tool_messages[0].content
+    assert "second streamed result" in tool_messages[0].content
+
+
 @pytest.mark.asyncio
 async def test_normal_completion_without_max_step(
     runner, mock_provider, provider_request, mock_tool_executor, mock_hooks