Skip to content

Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks.

fix: improve structured output handling across providers and agents

Greyson LaLondea731efac8dfe
diff --git a/lib/crewai/src/crewai/agents/crew_agent_executor.py b/lib/crewai/src/crewai/agents/crew_agent_executor.py
index d6e9c9bc3..1218ceae8 100644
--- a/lib/crewai/src/crewai/agents/crew_agent_executor.py
+++ b/lib/crewai/src/crewai/agents/crew_agent_executor.py
@@ -348,18 +348,36 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
# breakpoint()
if self.response_model is not None:
try:
- self.response_model.model_validate_json(answer)
- formatted_answer = AgentFinish(
- thought="",
- output=answer,
- text=answer,
- )
+ if isinstance(answer, BaseModel):
+ output_json = answer.model_dump_json()
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer,
+ text=output_json,
+ )
+ else:
+ self.response_model.model_validate_json(answer)
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer,
+ text=answer,
+ )
except ValidationError:
+ # If validation fails, convert BaseModel to JSON string for parsing
+ answer_str = (
+ answer.model_dump_json()
+ if isinstance(answer, BaseModel)
+ else str(answer)
+ )
formatted_answer = process_llm_response(
- answer, self.use_stop_words
+ answer_str, self.use_stop_words
) # type: ignore[assignment]
else:
- formatted_answer = process_llm_response(answer, self.use_stop_words) # type: ignore[assignment]
+ # When no response_model, answer should be a string
+ answer_str = str(answer) if not isinstance(answer, str) else answer
+ formatted_answer = process_llm_response(
+ answer_str, self.use_stop_words
+ ) # type: ignore[assignment]
if isinstance(formatted_answer, AgentAction):
# Extract agent fingerprint if available
@@ -520,6 +538,18 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
self._show_logs(formatted_answer)
return formatted_answer
+ if isinstance(answer, BaseModel):
+ output_json = answer.model_dump_json()
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer,
+ text=output_json,
+ )
+ self._invoke_step_callback(formatted_answer)
+ self._append_message(output_json)
+ self._show_logs(formatted_answer)
+ return formatted_answer
+
# Unexpected response type, treat as final answer
formatted_answer = AgentFinish(
thought="",
@@ -570,11 +600,20 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
verbose=self.agent.verbose,
)
- formatted_answer = AgentFinish(
- thought="",
- output=str(answer),
- text=str(answer),
- )
+ if isinstance(answer, BaseModel):
+ output_json = answer.model_dump_json()
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer,
+ text=output_json,
+ )
+ else:
+ answer_str = answer if isinstance(answer, str) else str(answer)
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer_str,
+ text=answer_str,
+ )
self._show_logs(formatted_answer)
return formatted_answer
@@ -1031,18 +1070,36 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
if self.response_model is not None:
try:
- self.response_model.model_validate_json(answer)
- formatted_answer = AgentFinish(
- thought="",
- output=answer,
- text=answer,
- )
+ if isinstance(answer, BaseModel):
+ output_json = answer.model_dump_json()
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer,
+ text=output_json,
+ )
+ else:
+ self.response_model.model_validate_json(answer)
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer,
+ text=answer,
+ )
except ValidationError:
+ # If validation fails, convert BaseModel to JSON string for parsing
+ answer_str = (
+ answer.model_dump_json()
+ if isinstance(answer, BaseModel)
+ else str(answer)
+ )
formatted_answer = process_llm_response(
- answer, self.use_stop_words
+ answer_str, self.use_stop_words
) # type: ignore[assignment]
else:
- formatted_answer = process_llm_response(answer, self.use_stop_words) # type: ignore[assignment]
+ # When no response_model, answer should be a string
+ answer_str = str(answer) if not isinstance(answer, str) else answer
+ formatted_answer = process_llm_response(
+ answer_str, self.use_stop_words
+ ) # type: ignore[assignment]
if isinstance(formatted_answer, AgentAction):
fingerprint_context = {}
@@ -1194,6 +1251,18 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
self._show_logs(formatted_answer)
return formatted_answer
+ if isinstance(answer, BaseModel):
+ output_json = answer.model_dump_json()
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer,
+ text=output_json,
+ )
+ self._invoke_step_callback(formatted_answer)
+ self._append_message(output_json)
+ self._show_logs(formatted_answer)
+ return formatted_answer
+
# Unexpected response type, treat as final answer
formatted_answer = AgentFinish(
thought="",
@@ -1244,11 +1313,20 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
verbose=self.agent.verbose,
)
- formatted_answer = AgentFinish(
- thought="",
- output=str(answer),
- text=str(answer),
- )
+ if isinstance(answer, BaseModel):
+ output_json = answer.model_dump_json()
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer,
+ text=output_json,
+ )
+ else:
+ answer_str = answer if isinstance(answer, str) else str(answer)
+ formatted_answer = AgentFinish(
+ thought="",
+ output=answer_str,
+ text=answer_str,
+ )
self._show_logs(formatted_answer)
return formatted_answer
@@ -1421,7 +1499,12 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
Returns:
Final answer after feedback.
"""
- human_feedback = self._ask_human_input(formatted_answer.output)
+ output_str = (
+ formatted_answer.output
+ if isinstance(formatted_answer.output, str)
+ else formatted_answer.output.model_dump_json()
+ )
+ human_feedback = self._ask_human_input(output_str)
if self._is_training_mode():
return self._handle_training_feedback(formatted_answer, human_feedback)
@@ -1480,7 +1563,12 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
self.ask_for_human_input = False
else:
answer = self._process_feedback_iteration(feedback)
- feedback = self._ask_human_input(answer.output)
+ output_str = (
+ answer.output
+ if isinstance(answer.output, str)
+ else answer.output.model_dump_json()
+ )
+ feedback = self._ask_human_input(output_str)
return answer
diff --git a/lib/crewai/src/crewai/agents/parser.py b/lib/crewai/src/crewai/agents/parser.py
index c338e8360..365443b45 100644
--- a/lib/crewai/src/crewai/agents/parser.py
+++ b/lib/crewai/src/crewai/agents/parser.py
@@ -8,6 +8,7 @@ AgentAction or AgentFinish objects.
from dataclasses import dataclass
from json_repair import repair_json # type: ignore[import-untyped]
+from pydantic import BaseModel
from crewai.agents.constants import (
ACTION_INPUT_ONLY_REGEX,
@@ -40,7 +41,7 @@ class AgentFinish:
"""Represents the final answer from an agent."""
thought: str
- output: str
+ output: str | BaseModel
text: str
diff --git a/lib/crewai/src/crewai/llms/providers/anthropic/completion.py b/lib/crewai/src/crewai/llms/providers/anthropic/completion.py
index e98249567..289a42226 100644
--- a/lib/crewai/src/crewai/llms/providers/anthropic/completion.py
+++ b/lib/crewai/src/crewai/llms/providers/anthropic/completion.py
@@ -23,7 +23,7 @@ if TYPE_CHECKING:
try:
from anthropic import Anthropic, AsyncAnthropic, transform_schema
from anthropic.types import Message, TextBlock, ThinkingBlock, ToolUseBlock
- from anthropic.types.beta import BetaMessage
+ from anthropic.types.beta import BetaMessage, BetaTextBlock
import httpx
except ImportError:
raise ImportError(
@@ -337,6 +337,7 @@ class AnthropicCompletion(BaseLLM):
available_functions: Available functions for tool calling
from_task: Task that initiated the call
from_agent: Agent that initiated the call
+ response_model: Optional response model.
Returns:
Chat completion response or tool call result
@@ -677,31 +678,31 @@ class AnthropicCompletion(BaseLLM):
if _is_pydantic_model_class(response_model) and response.content:
if use_native_structured_output:
for block in response.content:
- if isinstance(block, TextBlock):
- structured_json = block.text
+ if isinstance(block, (TextBlock, BetaTextBlock)):
+ structured_data = response_model.model_validate_json(block.text)
self._emit_call_completed_event(
- response=structured_json,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return structured_data
else:
for block in response.content:
if (
isinstance(block, ToolUseBlock)
and block.name == "structured_output"
):
- structured_json = json.dumps(block.input)
+ structured_data = response_model.model_validate(block.input)
self._emit_call_completed_event(
- response=structured_json,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return structured_data
# Check if Claude wants to use tools
if response.content:
@@ -897,28 +898,29 @@ class AnthropicCompletion(BaseLLM):
if _is_pydantic_model_class(response_model):
if use_native_structured_output:
+ structured_data = response_model.model_validate_json(full_response)
self._emit_call_completed_event(
- response=full_response,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return full_response
+ return structured_data
for block in final_message.content:
if (
isinstance(block, ToolUseBlock)
and block.name == "structured_output"
):
- structured_json = json.dumps(block.input)
+ structured_data = response_model.model_validate(block.input)
self._emit_call_completed_event(
- response=structured_json,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return structured_data
if final_message.content:
tool_uses = [
@@ -1166,31 +1168,31 @@ class AnthropicCompletion(BaseLLM):
if _is_pydantic_model_class(response_model) and response.content:
if use_native_structured_output:
for block in response.content:
- if isinstance(block, TextBlock):
- structured_json = block.text
+ if isinstance(block, (TextBlock, BetaTextBlock)):
+ structured_data = response_model.model_validate_json(block.text)
self._emit_call_completed_event(
- response=structured_json,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return structured_data
else:
for block in response.content:
if (
isinstance(block, ToolUseBlock)
and block.name == "structured_output"
):
- structured_json = json.dumps(block.input)
+ structured_data = response_model.model_validate(block.input)
self._emit_call_completed_event(
- response=structured_json,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return structured_data
if response.content:
tool_uses = [
@@ -1362,28 +1364,29 @@ class AnthropicCompletion(BaseLLM):
if _is_pydantic_model_class(response_model):
if use_native_structured_output:
+ structured_data = response_model.model_validate_json(full_response)
self._emit_call_completed_event(
- response=full_response,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return full_response
+ return structured_data
for block in final_message.content:
if (
isinstance(block, ToolUseBlock)
and block.name == "structured_output"
):
- structured_json = json.dumps(block.input)
+ structured_data = response_model.model_validate(block.input)
self._emit_call_completed_event(
- response=structured_json,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return structured_data
if final_message.content:
tool_uses = [
diff --git a/lib/crewai/src/crewai/llms/providers/azure/completion.py b/lib/crewai/src/crewai/llms/providers/azure/completion.py
index 1de18d984..de9a79174 100644
--- a/lib/crewai/src/crewai/llms/providers/azure/completion.py
+++ b/lib/crewai/src/crewai/llms/providers/azure/completion.py
@@ -557,7 +557,7 @@ class AzureCompletion(BaseLLM):
params: AzureCompletionParams,
from_task: Any | None = None,
from_agent: Any | None = None,
- ) -> str:
+ ) -> BaseModel:
"""Validate content against response model and emit completion event.
Args:
@@ -568,24 +568,23 @@ class AzureCompletion(BaseLLM):
from_agent: Agent that initiated the call
Returns:
- Validated and serialized JSON string
+ Validated Pydantic model instance
Raises:
ValueError: If validation fails
"""
try:
structured_data = response_model.model_validate_json(content)
- structured_json = structured_data.model_dump_json()
self._emit_call_completed_event(
- response=structured_json,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return structured_data
except Exception as e:
error_msg = f"Failed to validate structured output with model {response_model.__name__}: {e}"
logging.error(error_msg)
diff --git a/lib/crewai/src/crewai/llms/providers/gemini/completion.py b/lib/crewai/src/crewai/llms/providers/gemini/completion.py
index 950cdab57..d32cc1ea0 100644
--- a/lib/crewai/src/crewai/llms/providers/gemini/completion.py
+++ b/lib/crewai/src/crewai/llms/providers/gemini/completion.py
@@ -132,6 +132,9 @@ class GeminiCompletion(BaseLLM):
self.supports_tools = bool(
version_match and float(version_match.group(1)) >= 1.5
)
+ self.is_gemini_2_0 = bool(
+ version_match and float(version_match.group(1)) >= 2.0
+ )
@property
def stop(self) -> list[str]:
@@ -439,6 +442,11 @@ class GeminiCompletion(BaseLLM):
Returns:
GenerateContentConfig object for Gemini API
+
+ Note:
+ Structured output support varies by model version:
+ - Gemini 1.5 and earlier: Uses response_schema (Pydantic model)
+ - Gemini 2.0+: Uses response_json_schema (JSON Schema) with propertyOrdering
"""
self.tools = tools
config_params: dict[str, Any] = {}
@@ -466,9 +474,13 @@ class GeminiCompletion(BaseLLM):
if response_model:
config_params["response_mime_type"] = "application/json"
schema_output = generate_model_description(response_model)
- config_params["response_schema"] = schema_output.get("json_schema", {}).get(
- "schema", {}
- )
+ schema = schema_output.get("json_schema", {}).get("schema", {})
+
+ if self.is_gemini_2_0:
+ schema = self._add_property_ordering(schema)
+ config_params["response_json_schema"] = schema
+ else:
+ config_params["response_schema"] = response_model
# Handle tools for supported models
if tools and self.supports_tools:
@@ -632,7 +644,7 @@ class GeminiCompletion(BaseLLM):
messages_for_event: list[LLMMessage],
from_task: Any | None = None,
from_agent: Any | None = None,
- ) -> str:
+ ) -> BaseModel:
"""Validate content against response model and emit completion event.
Args:
@@ -643,24 +655,23 @@ class GeminiCompletion(BaseLLM):
from_agent: Agent that initiated the call
Returns:
- Validated and serialized JSON string
+ Validated Pydantic model instance
Raises:
ValueError: If validation fails
"""
try:
structured_data = response_model.model_validate_json(content)
- structured_json = structured_data.model_dump_json()
self._emit_call_completed_event(
- response=structured_json,
+ response=structured_data.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=messages_for_event,
)
- return structured_json
+ return structured_data
except Exception as e:
error_msg = f"Failed to validate structured output with model {response_model.__name__}: {e}"
logging.error(error_msg)
@@ -673,7 +684,7 @@ class GeminiCompletion(BaseLLM):
response_model: type[BaseModel] | None = None,
from_task: Any | None = None,
from_agent: Any | None = None,
- ) -> str:
+ ) -> str | BaseModel:
"""Finalize completion response with validation and event emission.
Args:
@@ -684,7 +695,7 @@ class GeminiCompletion(BaseLLM):
from_agent: Agent that initiated the call
Returns:
- Final response content after processing
+ Final response content after processing (str or Pydantic model if response_model provided)
"""
messages_for_event = self._convert_contents_to_dict(contents)
@@ -870,7 +881,7 @@ class GeminiCompletion(BaseLLM):
from_task: Any | None = None,
from_agent: Any | None = None,
response_model: type[BaseModel] | None = None,
- ) -> str | list[dict[str, Any]]:
+ ) -> str | BaseModel | list[dict[str, Any]]:
"""Finalize streaming response with usage tracking, function execution, and events.
Args:
@@ -990,7 +1001,7 @@ class GeminiCompletion(BaseLLM):
from_task: Any | None = None,
from_agent: Any | None = None,
response_model: type[BaseModel] | None = None,
- ) -> str | Any:
+ ) -> str | BaseModel | list[dict[str, Any]] | Any:
"""Handle streaming content generation."""
full_response = ""
function_calls: dict[int, dict[str, Any]] = {}
@@ -1190,6 +1201,36 @@ class GeminiCompletion(BaseLLM):
return "".join(text_parts)
+ @staticmethod
+ def _add_property_ordering(schema: dict[str, Any]) -> dict[str, Any]:
+ """Add propertyOrdering to JSON schema for Gemini 2.0 compatibility.
+
+ Gemini 2.0 models require an explicit propertyOrdering list to define
+ the preferred structure of JSON objects. This recursively adds
+ propertyOrdering to all objects in the schema.
+
+ Args:
+ schema: JSON schema dictionary.
+
+ Returns:
+ Modified schema with propertyOrdering added to all objects.
+ """
+ if isinstance(schema, dict):
+ if schema.get("type") == "object" and "properties" in schema:
+ properties = schema["properties"]
+ if properties and "propertyOrdering" not in schema:
+ schema["propertyOrdering"] = list(properties.keys())
+
+ for value in schema.values():
+ if isinstance(value, dict):
+ GeminiCompletion._add_property_ordering(value)
+ elif isinstance(value, list):
+ for item in value:
+ if isinstance(item, dict):
+ GeminiCompletion._add_property_ordering(item)
+
+ return schema
+
@staticmethod
def _convert_contents_to_dict(
contents: list[types.Content],
diff --git a/lib/crewai/src/crewai/llms/providers/openai/completion.py b/lib/crewai/src/crewai/llms/providers/openai/completion.py
index be7991970..5fc0a57b2 100644
--- a/lib/crewai/src/crewai/llms/providers/openai/completion.py
+++ b/lib/crewai/src/crewai/llms/providers/openai/completion.py
@@ -1570,15 +1570,14 @@ class OpenAICompletion(BaseLLM):
parsed_object = parsed_response.choices[0].message.parsed
if parsed_object:
- structured_json = parsed_object.model_dump_json()
self._emit_call_completed_event(
- response=structured_json,
+ response=parsed_object.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return parsed_object
response: ChatCompletion = self.client.chat.completions.create(**params)
@@ -1692,7 +1691,7 @@ class OpenAICompletion(BaseLLM):
from_task: Any | None = None,
from_agent: Any | None = None,
response_model: type[BaseModel] | None = None,
- ) -> str:
+ ) -> str | BaseModel:
"""Handle streaming chat completion."""
full_response = ""
tool_calls: dict[int, dict[str, Any]] = {}
@@ -1728,15 +1727,14 @@ class OpenAICompletion(BaseLLM):
if final_completion.choices:
parsed_result = final_completion.choices[0].message.parsed
if parsed_result:
- structured_json = parsed_result.model_dump_json()
self._emit_call_completed_event(
- response=structured_json,
+ response=parsed_result.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return parsed_result
logging.error("Failed to get parsed result from stream")
return ""
@@ -1887,15 +1885,14 @@ class OpenAICompletion(BaseLLM):
parsed_object = parsed_response.choices[0].message.parsed
if parsed_object:
- structured_json = parsed_object.model_dump_json()
self._emit_call_completed_event(
- response=structured_json,
+ response=parsed_object.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return parsed_object
response: ChatCompletion = await self.async_client.chat.completions.create(
**params
@@ -2006,7 +2003,7 @@ class OpenAICompletion(BaseLLM):
from_task: Any | None = None,
from_agent: Any | None = None,
response_model: type[BaseModel] | None = None,
- ) -> str:
+ ) -> str | BaseModel:
"""Handle async streaming chat completion."""
full_response = ""
tool_calls: dict[int, dict[str, Any]] = {}
@@ -2044,17 +2041,16 @@ class OpenAICompletion(BaseLLM):
try:
parsed_object = response_model.model_validate_json(accumulated_content)
- structured_json = parsed_object.model_dump_json()
self._emit_call_completed_event(
- response=structured_json,
+ response=parsed_object.model_dump_json(),
call_type=LLMCallType.LLM_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
- return structured_json
+ return parsed_object
except Exception as e:
logging.error(f"Failed to parse structured output from stream: {e}")
self._emit_call_completed_event(
diff --git a/lib/crewai/src/crewai/utilities/agent_utils.py b/lib/crewai/src/crewai/utilities/agent_utils.py
index d759368d9..34a368f34 100644
--- a/lib/crewai/src/crewai/utilities/agent_utils.py
+++ b/lib/crewai/src/crewai/utilities/agent_utils.py
@@ -327,7 +327,7 @@ def get_llm_response(
response_model: type[BaseModel] | None = None,
executor_context: CrewAgentExecutor | AgentExecutor | LiteAgent | None = None,
verbose: bool = True,
-) -> str | Any:
+) -> str | BaseModel | Any:
"""Call the LLM and return the response, handling any invalid responses.
Args:
@@ -341,10 +341,11 @@ def get_llm_response(
from_agent: Optional agent context for the LLM call.
response_model: Optional Pydantic model for structured outputs.
executor_context: Optional executor context for hook invocation.
+ verbose: Whether to print output.
Returns:
- The response from the LLM as a string, or tool call results if
- native function calling is used.
+ The response from the LLM as a string, Pydantic model (when response_model is provided),
+ or tool call results if native function calling is used.
Raises:
Exception: If an error occurs.
@@ -393,7 +394,7 @@ async def aget_llm_response(
response_model: type[BaseModel] | None = None,
executor_context: CrewAgentExecutor | AgentExecutor | None = None,
verbose: bool = True,
-) -> str | Any:
+) -> str | BaseModel | Any:
"""Call the LLM asynchronously and return the response.
Args:
@@ -409,8 +410,8 @@ async def aget_llm_response(
executor_context: Optional executor context for hook invocation.
Returns:
- The response from the LLM as a string, or tool call results if
- native function calling is used.
+ The response from the LLM as a string, Pydantic model (when response_model is provided),
+ or tool call results if native function calling is used.
Raises:
Exception: If an error occurs.
@@ -986,32 +987,41 @@ def _setup_before_llm_call_hooks(
def _setup_after_llm_call_hooks(
executor_context: CrewAgentExecutor | AgentExecutor | LiteAgent | None,
- answer: str,
+ answer: str | BaseModel,
printer: Printer,
verbose: bool = True,
-) -> str:
+) -> str | BaseModel:
"""Setup and invoke after_llm_call hooks for the executor context.
Args:
executor_context: The executor context to setup the hooks for.
- answer: The LLM response string.
+ answer: The LLM response (string or Pydantic model).
printer: Printer instance for error logging.
verbose: Whether to print output.
Returns:
- The potentially modified response string.
+ The potentially modified response (string or Pydantic model).
"""
if executor_context and executor_context.after_llm_call_hooks:
from crewai.hooks.llm_hooks import LLMCallHookContext
original_messages = executor_context.messages
- hook_context = LLMCallHookContext(executor_context, response=answer)
+ # For Pydantic models, serialize to JSON for hooks
+ if isinstance(answer, BaseModel):
+ pydantic_answer = answer
+ hook_response: str = pydantic_answer.model_dump_json()
+ original_json: str = hook_response
+ else:
+ pydantic_answer = None
+ hook_response = str(answer)
+
+ hook_context = LLMCallHookContext(executor_context, response=hook_response)
try:
for hook in executor_context.after_llm_call_hooks:
modified_response = hook(hook_context)
if modified_response is not None and isinstance(modified_response, str):
- answer = modified_response
+ hook_response = modified_response
except Exception as e:
if verbose:
@@ -1035,4 +1045,21 @@ def _setup_after_llm_call_hooks(
else:
executor_context.messages = []
+ # If hooks modified the response, update answer accordingly
+ if pydantic_answer is not None:
+ # For Pydantic models, reparse the JSON if it was modified
+ if hook_response != original_json:
+ try:
+ model_class: type[BaseModel] = type(pydantic_answer)
+ answer = model_class.model_validate_json(hook_response)
+ except Exception as e:
+ if verbose:
+ printer.print(
+ content=f"Warning: Hook modified response but failed to reparse as {type(pydantic_answer).__name__}: {e}. Using original model.",
+ color="yellow",
+ )
+ else:
+ # For string responses, use the hook-modified response
+ answer = hook_response
+
return answer
diff --git a/lib/crewai/src/crewai/utilities/converter.py b/lib/crewai/src/crewai/utilities/converter.py
index 09230ee12..67f542d53 100644
--- a/lib/crewai/src/crewai/utilities/converter.py
+++ b/lib/crewai/src/crewai/utilities/converter.py
@@ -62,7 +62,10 @@ class Converter(OutputConverter):
],
response_model=self.model,
)
- result = self.model.model_validate_json(response)
+ if isinstance(response, BaseModel):
+ result = response
+ else:
+ result = self.model.model_validate_json(response)
else:
response = self.llm.call(
[
diff --git a/lib/crewai/tests/llms/anthropic/test_anthropic_async.py b/lib/crewai/tests/llms/anthropic/test_anthropic_async.py
index a93397317..e09e02d4b 100644
--- a/lib/crewai/tests/llms/anthropic/test_anthropic_async.py
+++ b/lib/crewai/tests/llms/anthropic/test_anthropic_async.py
@@ -157,10 +157,10 @@ async def test_anthropic_async_with_response_model():
"Say hello in French",
response_model=GreetingResponse
)
- model = GreetingResponse.model_validate_json(result)
- assert isinstance(model, GreetingResponse)
- assert isinstance(model.greeting, str)
- assert isinstance(model.language, str)
+ # When response_model is provided, the result is already a parsed Pydantic model instance
+ assert isinstance(result, GreetingResponse)
+ assert isinstance(result.greeting, str)
+ assert isinstance(result.language, str)
@pytest.mark.vcr()
diff --git a/lib/crewai/tests/llms/google/test_google.py b/lib/crewai/tests/llms/google/test_google.py
index 33d53da2d..925108072 100644
--- a/lib/crewai/tests/llms/google/test_google.py
+++ b/lib/crewai/tests/llms/google/test_google.py
@@ -799,3 +799,131 @@ def test_google_express_mode_works() -> None:
assert result.token_usage.prompt_tokens > 0
assert result.token_usage.completion_tokens > 0
assert result.token_usage.successful_requests >= 1
+
+
+def test_gemini_2_0_model_detection():
+ """Test that Gemini 2.0 models are properly detected."""
+ # Test Gemini 2.0 models
+ llm_2_0 = LLM(model="google/gemini-2.0-flash-001")
+ from crewai.llms.providers.gemini.completion import GeminiCompletion
+ assert isinstance(llm_2_0, GeminiCompletion)
+ assert llm_2_0.is_gemini_2_0 is True
+
+ llm_2_5 = LLM(model="google/gemini-2.5-flash")
+ assert isinstance(llm_2_5, GeminiCompletion)
+ assert llm_2_5.is_gemini_2_0 is True
+
+ # Test non-2.0 models
+ llm_1_5 = LLM(model="google/gemini-1.5-pro")
+ assert isinstance(llm_1_5, GeminiCompletion)
+ assert llm_1_5.is_gemini_2_0 is False
+
+
+def test_add_property_ordering_to_schema():
+ """Test that _add_property_ordering correctly adds propertyOrdering to schemas."""
+ from crewai.llms.providers.gemini.completion import GeminiCompletion
+
+ # Test simple object schema
+ simple_schema = {
+ "type": "object",
+ "properties": {
+ "name": {"type": "string"},
+ "age": {"type": "integer"},
+ "email": {"type": "string"}
+ }
+ }
+
+ result = GeminiCompletion._add_property_ordering(simple_schema)
+
+ assert "propertyOrdering" in result
+ assert result["propertyOrdering"] == ["name", "age", "email"]
+
+ # Test nested object schema
+ nested_schema = {
+ "type": "object",
+ "properties": {
+ "user": {
+ "type": "object",
+ "properties": {
+ "name": {"type": "string"},
+ "contact": {
+ "type": "object",
+ "properties": {
+ "email": {"type": "string"},
+ "phone": {"type": "string"}
+ }
+ }
+ }
+ },
+ "id": {"type": "integer"}
+ }
+ }
+
+ result = GeminiCompletion._add_property_ordering(nested_schema)
+
+ assert "propertyOrdering" in result
+ assert result["propertyOrdering"] == ["user", "id"]
+ assert "propertyOrdering" in result["properties"]["user"]
+ assert result["properties"]["user"]["propertyOrdering"] == ["name", "contact"]
+ assert "propertyOrdering" in result["properties"]["user"]["properties"]["contact"]
+ assert result["properties"]["user"]["properties"]["contact"]["propertyOrdering"] == ["email", "phone"]
+
+
+def test_gemini_2_0_response_model_with_property_ordering():
+ """Test that Gemini 2.0 models include propertyOrdering in response schemas."""
+ from pydantic import BaseModel, Field
+
+ class TestResponse(BaseModel):
+ """Test response model."""
+ name: str = Field(..., description="The name")
+ age: int = Field(..., description="The age")
+ email: str = Field(..., description="The email")
+
+ llm = LLM(model="google/gemini-2.0-flash-001")
+
+ # Prepare generation config with response model
+ config = llm._prepare_generation_config(response_model=TestResponse)
+
+ # Verify that the config has response_json_schema
+ assert hasattr(config, 'response_json_schema') or 'response_json_schema' in config.__dict__
+
+ # Get the schema
+ if hasattr(config, 'response_json_schema'):
+ schema = config.response_json_schema
+ else:
+ schema = config.__dict__.get('response_json_schema', {})
+
+ # Verify propertyOrdering is present for Gemini 2.0
+ assert "propertyOrdering" in schema
+ assert "name" in schema["propertyOrdering"]
+ assert "age" in schema["propertyOrdering"]
+ assert "email" in schema["propertyOrdering"]
+
+
+def test_gemini_1_5_response_model_uses_response_schema():
+ """Test that Gemini 1.5 models use response_schema parameter (not response_json_schema)."""
+ from pydantic import BaseModel, Field
+
+ class TestResponse(BaseModel):
+ """Test response model."""
+ name: str = Field(..., description="The name")
+ age: int = Field(..., description="The age")
+
+ llm = LLM(model="google/gemini-1.5-pro")
+
+ # Prepare generation config with response model
+ config = llm._prepare_generation_config(response_model=TestResponse)
+
+ # Verify that the config uses response_schema (not response_json_schema)
+ assert hasattr(config, 'response_schema') or 'response_schema' in config.__dict__
+ assert not (hasattr(config, 'response_json_schema') and config.response_json_schema is not None)
+
+ # Get the schema
+ if hasattr(config, 'response_schema'):
+ schema = config.response_schema
+ else:
+ schema = config.__dict__.get('response_schema')
+
+ # For Gemini 1.5, response_schema should be the Pydantic model itself
+ # The SDK handles conversion internally
+ assert schema is TestResponse or isinstance(schema, type)
diff --git a/lib/crewai/tests/llms/openai/test_openai.py b/lib/crewai/tests/llms/openai/test_openai.py
index 8c8f49da6..f4c7f9c8c 100644
--- a/lib/crewai/tests/llms/openai/test_openai.py
+++ b/lib/crewai/tests/llms/openai/test_openai.py
@@ -540,7 +540,9 @@ def test_openai_streaming_with_response_model():
result = llm.call("Test question", response_model=TestResponse)
assert result is not None
- assert isinstance(result, str)
+ assert isinstance(result, TestResponse)
+ assert result.answer == "test"
+ assert result.confidence == 0.95
assert mock_stream.called
call_kwargs = mock_stream.call_args[1]