Fix structured output leaks in tool-calling loops

lib/crewai/src/crewai/agent/core.py

@@ -28,6 +28,7 @@ from pydantic import (
     ConfigDict,
     Field,
     PrivateAttr,
+    ValidationError,
     model_validator,
 )
 from pydantic.functional_serializers import PlainSerializer
@@ -1691,24 +1692,30 @@ class Agent(BaseAgent):
         elif response_format:
             raw_output = str(output) if not isinstance(output, str) else output
             try:
-                model_schema = generate_model_description(response_format)
-                schema = json.dumps(model_schema, indent=2)
-                instructions = I18N_DEFAULT.slice("formatted_task_instructions").format(
-                    output_format=schema
-                )
-
-                converter = Converter(
-                    llm=cast(BaseLLM, self.llm),
-                    text=raw_output,
-                    model=response_format,
-                    instructions=instructions,
-                )
-
-                conversion_result = converter.to_pydantic()
-                if isinstance(conversion_result, BaseModel):
-                    formatted_result = conversion_result
-            except ConverterError:
+                formatted_result = response_format.model_validate_json(raw_output)
+            except ValidationError:
                 pass
+
+            if formatted_result is None:
+                try:
+                    model_schema = generate_model_description(response_format)
+                    schema = json.dumps(model_schema, indent=2)
+                    instructions = I18N_DEFAULT.slice(
+                        "formatted_task_instructions"
+                    ).format(output_format=schema)
+
+                    converter = Converter(
+                        llm=cast(BaseLLM, self.llm),
+                        text=raw_output,
+                        model=response_format,
+                        instructions=instructions,
+                    )
+
+                    conversion_result = converter.to_pydantic()
+                    if isinstance(conversion_result, BaseModel):
+                        formatted_result = conversion_result
+                except ConverterError:
+                    pass
         else:
             raw_output = str(output) if not isinstance(output, str) else output
 

lib/crewai/src/crewai/agents/crew_agent_executor.py

@@ -350,6 +350,10 @@ class CrewAgentExecutor(BaseAgentExecutor):
 
                 enforce_rpm_limit(self.request_within_rpm_limit)
 
+                effective_response_model = (
+                    None if self.original_tools else self.response_model
+                )
+
                 answer = get_llm_response(
                     llm=cast("BaseLLM", self.llm),
                     messages=self.messages,
@@ -357,11 +361,11 @@ class CrewAgentExecutor(BaseAgentExecutor):
                     printer=PRINTER,
                     from_task=self.task,
                     from_agent=self.agent,
-                    response_model=self.response_model,
+                    response_model=effective_response_model,
                     executor_context=self,
                     verbose=self.agent.verbose,
                 )
-                if self.response_model is not None:
+                if effective_response_model is not None:
                     try:
                         if isinstance(answer, BaseModel):
                             output_json = answer.model_dump_json()
@@ -502,7 +506,7 @@ class CrewAgentExecutor(BaseAgentExecutor):
                     available_functions=None,
                     from_task=self.task,
                     from_agent=self.agent,
-                    response_model=self.response_model,
+                    response_model=None,
                     executor_context=self,
                     verbose=self.agent.verbose,
                 )
@@ -1161,6 +1165,10 @@ class CrewAgentExecutor(BaseAgentExecutor):
 
                 enforce_rpm_limit(self.request_within_rpm_limit)
 
+                effective_response_model = (
+                    None if self.original_tools else self.response_model
+                )
+
                 answer = await aget_llm_response(
                     llm=cast("BaseLLM", self.llm),
                     messages=self.messages,
@@ -1168,12 +1176,12 @@ class CrewAgentExecutor(BaseAgentExecutor):
                     printer=PRINTER,
                     from_task=self.task,
                     from_agent=self.agent,
-                    response_model=self.response_model,
+                    response_model=effective_response_model,
                     executor_context=self,
                     verbose=self.agent.verbose,
                 )
 
-                if self.response_model is not None:
+                if effective_response_model is not None:
                     try:
                         if isinstance(answer, BaseModel):
                             output_json = answer.model_dump_json()
@@ -1314,7 +1322,7 @@ class CrewAgentExecutor(BaseAgentExecutor):
                     available_functions=None,
                     from_task=self.task,
                     from_agent=self.agent,
-                    response_model=self.response_model,
+                    response_model=None,
                     executor_context=self,
                     verbose=self.agent.verbose,
                 )

lib/crewai/src/crewai/experimental/agent_executor.py

@@ -1224,6 +1224,10 @@ class AgentExecutor(Flow[AgentExecutorState], BaseAgentExecutor):
         try:
             enforce_rpm_limit(self.request_within_rpm_limit)
 
+            effective_response_model = (
+                None if self.original_tools else self.response_model
+            )
+
             answer = get_llm_response(
                 llm=self.llm,
                 messages=list(self.state.messages),
@@ -1231,7 +1235,7 @@ class AgentExecutor(Flow[AgentExecutorState], BaseAgentExecutor):
                 printer=PRINTER,
                 from_task=self.task,
                 from_agent=self.agent,
-                response_model=self.response_model,
+                response_model=effective_response_model,
                 executor_context=self,
                 verbose=self.agent.verbose,
             )
@@ -1319,7 +1323,7 @@ class AgentExecutor(Flow[AgentExecutorState], BaseAgentExecutor):
                 available_functions=None,
                 from_task=self.task,
                 from_agent=self.agent,
-                response_model=self.response_model,
+                response_model=None,
                 executor_context=self,
                 verbose=self.agent.verbose,
             )

lib/crewai/src/crewai/lite_agent.py

@@ -23,6 +23,7 @@ from pydantic import (
     BaseModel,
     Field,
     PrivateAttr,
+    ValidationError,
     field_validator,
     model_validator,
 )
@@ -639,29 +640,37 @@ class LiteAgent(FlowTrackable, BaseModel):
             formatted_result = agent_finish.output
         elif active_response_format:
             try:
-                model_schema = generate_model_description(active_response_format)
-                schema = json.dumps(model_schema, indent=2)
-                instructions = I18N_DEFAULT.slice("formatted_task_instructions").format(
-                    output_format=schema
+                formatted_result = active_response_format.model_validate_json(
+                    str(agent_finish.output)
                 )
+            except ValidationError:
+                pass
 
-                converter = Converter(
-                    llm=self.llm,
-                    text=agent_finish.output,
-                    model=active_response_format,
-                    instructions=instructions,
-                )
+            if formatted_result is None:
+                try:
+                    model_schema = generate_model_description(active_response_format)
+                    schema = json.dumps(model_schema, indent=2)
+                    instructions = I18N_DEFAULT.slice(
+                        "formatted_task_instructions"
+                    ).format(output_format=schema)
 
-                result = converter.to_pydantic()
-                if isinstance(result, BaseModel):
-                    formatted_result = result
-            except ConverterError as e:
-                if self.verbose:
-                    PRINTER.print(
-                        content=f"Failed to parse output into response format after retries: {e.message}",
-                        color="yellow",
+                    converter = Converter(
+                        llm=self.llm,
+                        text=agent_finish.output,
+                        model=active_response_format,
+                        instructions=instructions,
                     )
 
+                    result = converter.to_pydantic()
+                    if isinstance(result, BaseModel):
+                        formatted_result = result
+                except ConverterError as e:
+                    if self.verbose:
+                        PRINTER.print(
+                            content=f"Failed to parse output into response format after retries: {e.message}",
+                            color="yellow",
+                        )
+
         # Calculate token usage metrics
         if isinstance(self.llm, BaseLLM):
             usage_metrics = self.llm.get_token_usage_summary()

lib/crewai/tests/agents/test_agent_executor.py

@@ -12,6 +12,7 @@ from typing import Any
 from unittest.mock import AsyncMock, Mock, patch
 
 import pytest
+from pydantic import BaseModel
 
 from crewai.agents.tools_handler import ToolsHandler as _ToolsHandler
 from crewai.agents.step_executor import StepExecutor
@@ -108,6 +109,9 @@ class TestAgentExecutorState:
 class TestAgentExecutor:
     """Test AgentExecutor class."""
 
+    class StructuredResult(BaseModel):
+        value: str
+
     @pytest.fixture
     def mock_dependencies(self):
         """Create mock dependencies for executor."""
@@ -215,6 +219,49 @@ class TestAgentExecutor:
 
         assert result == "check_iteration"
 
+    def test_call_llm_and_parse_does_not_pass_response_model_with_tools(
+        self, mock_dependencies
+    ):
+        """Structured output should not be requested during ReAct tool loops."""
+        executor = _build_executor(
+            **mock_dependencies,
+            original_tools=[Mock()],
+            response_model=self.StructuredResult,
+            callbacks=[],
+        )
+        executor.state.messages = [{"role": "user", "content": "Use a tool"}]
+
+        with patch(
+            "crewai.experimental.agent_executor.get_llm_response",
+            return_value="Thought: done\nFinal Answer: complete",
+        ) as get_llm_response_mock:
+            result = executor.call_llm_and_parse()
+
+        assert result == "parsed"
+        assert get_llm_response_mock.call_args.kwargs["response_model"] is None
+
+    def test_call_llm_native_tools_does_not_pass_response_model_with_tools(
+        self, mock_dependencies
+    ):
+        """Structured output should not be requested during native tool calls."""
+        executor = _build_executor(
+            **mock_dependencies,
+            original_tools=[Mock()],
+            response_model=self.StructuredResult,
+            callbacks=[],
+        )
+        executor._openai_tools = [{"type": "function", "function": {"name": "lookup"}}]
+        executor.state.messages = [{"role": "user", "content": "Use a tool"}]
+
+        with patch(
+            "crewai.experimental.agent_executor.get_llm_response",
+            return_value="complete",
+        ) as get_llm_response_mock:
+            result = executor.call_llm_native_tools()
+
+        assert result == "native_finished"
+        assert get_llm_response_mock.call_args.kwargs["response_model"] is None
+
     def test_finalize_success(self, mock_dependencies):
         """Test finalize with valid AgentFinish."""
         with patch.object(AgentExecutor, "_show_logs") as mock_show_logs: