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: ensure openai tool call stream is finalized

Greyson LaLonde3a22e80764ca
diff --git a/lib/crewai/src/crewai/llms/providers/openai/completion.py b/lib/crewai/src/crewai/llms/providers/openai/completion.py
index 37b686e3d..22b9cda3b 100644
--- a/lib/crewai/src/crewai/llms/providers/openai/completion.py
+++ b/lib/crewai/src/crewai/llms/providers/openai/completion.py
@@ -1696,6 +1696,99 @@ class OpenAICompletion(BaseLLM):
return content
+ def _finalize_streaming_response(
+ self,
+ full_response: str,
+ tool_calls: dict[int, dict[str, Any]],
+ usage_data: dict[str, int],
+ params: dict[str, Any],
+ available_functions: dict[str, Any] | None = None,
+ from_task: Any | None = None,
+ from_agent: Any | None = None,
+ ) -> str | list[dict[str, Any]]:
+ """Finalize a streaming response with usage tracking, tool call handling, and events.
+
+ Args:
+ full_response: The accumulated text response from the stream.
+ tool_calls: Accumulated tool calls from the stream, keyed by index.
+ usage_data: Token usage data from the stream.
+ params: The completion parameters containing messages.
+ available_functions: Available functions for tool calling.
+ from_task: Task that initiated the call.
+ from_agent: Agent that initiated the call.
+
+ Returns:
+ Tool calls list when tools were invoked without available_functions,
+ tool execution result when available_functions is provided,
+ or the text response string.
+ """
+ self._track_token_usage_internal(usage_data)
+
+ if tool_calls and not available_functions:
+ tool_calls_list = [
+ {
+ "id": call_data["id"],
+ "type": "function",
+ "function": {
+ "name": call_data["name"],
+ "arguments": call_data["arguments"],
+ },
+ "index": call_data["index"],
+ }
+ for call_data in tool_calls.values()
+ ]
+ self._emit_call_completed_event(
+ response=tool_calls_list,
+ call_type=LLMCallType.TOOL_CALL,
+ from_task=from_task,
+ from_agent=from_agent,
+ messages=params["messages"],
+ )
+ return tool_calls_list
+
+ if tool_calls and available_functions:
+ for call_data in tool_calls.values():
+ function_name = call_data["name"]
+ arguments = call_data["arguments"]
+
+ if not function_name or not arguments:
+ continue
+
+ if function_name not in available_functions:
+ logging.warning(
+ f"Function '{function_name}' not found in available functions"
+ )
+ continue
+
+ try:
+ function_args = json.loads(arguments)
+ except json.JSONDecodeError as e:
+ logging.error(f"Failed to parse streamed tool arguments: {e}")
+ continue
+
+ result = self._handle_tool_execution(
+ function_name=function_name,
+ function_args=function_args,
+ available_functions=available_functions,
+ from_task=from_task,
+ from_agent=from_agent,
+ )
+
+ if result is not None:
+ return result
+
+ full_response = self._apply_stop_words(full_response)
+
+ self._emit_call_completed_event(
+ response=full_response,
+ call_type=LLMCallType.LLM_CALL,
+ from_task=from_task,
+ from_agent=from_agent,
+ messages=params["messages"],
+ )
+
+ return full_response
+
def _handle_streaming_completion(
self,
params: dict[str, Any],
@@ -1703,7 +1796,7 @@ class OpenAICompletion(BaseLLM):
from_task: Any | None = None,
from_agent: Any | None = None,
response_model: type[BaseModel] | None = None,
- ) -> str | BaseModel:
+ ) -> str | list[dict[str, Any]] | BaseModel:
"""Handle streaming chat completion."""
full_response = ""
tool_calls: dict[int, dict[str, Any]] = {}
@@ -1820,54 +1913,20 @@ class OpenAICompletion(BaseLLM):
response_id=response_id_stream,
)
- self._track_token_usage_internal(usage_data)
-
- if tool_calls and available_functions:
- for call_data in tool_calls.values():
- function_name = call_data["name"]
- arguments = call_data["arguments"]
-
- # Skip if function name is empty or arguments are empty
- if not function_name or not arguments:
- continue
-
- # Check if function exists in available functions
- if function_name not in available_functions:
- logging.warning(
- f"Function '{function_name}' not found in available functions"
- )
- continue
-
- try:
- function_args = json.loads(arguments)
- except json.JSONDecodeError as e:
- logging.error(f"Failed to parse streamed tool arguments: {e}")
- continue
-
- result = self._handle_tool_execution(
- function_name=function_name,
- function_args=function_args,
- available_functions=available_functions,
- from_task=from_task,
- from_agent=from_agent,
- )
-
- if result is not None:
- return result
-
- full_response = self._apply_stop_words(full_response)
-
- self._emit_call_completed_event(
- response=full_response,
- call_type=LLMCallType.LLM_CALL,
+ result = self._finalize_streaming_response(
+ full_response=full_response,
+ tool_calls=tool_calls,
+ usage_data=usage_data,
+ params=params,
+ available_functions=available_functions,
from_task=from_task,
from_agent=from_agent,
- messages=params["messages"],
- )
-
- return self._invoke_after_llm_call_hooks(
- params["messages"], full_response, from_agent
)
+ if isinstance(result, str):
+ return self._invoke_after_llm_call_hooks(
+ params["messages"], result, from_agent
+ )
+ return result
async def _ahandle_completion(
self,
@@ -2016,7 +2075,7 @@ class OpenAICompletion(BaseLLM):
from_task: Any | None = None,
from_agent: Any | None = None,
response_model: type[BaseModel] | None = None,
- ) -> str | BaseModel:
+ ) -> str | list[dict[str, Any]] | BaseModel:
"""Handle async streaming chat completion."""
full_response = ""
tool_calls: dict[int, dict[str, Any]] = {}
@@ -2142,51 +2201,16 @@ class OpenAICompletion(BaseLLM):
response_id=response_id_stream,
)
- self._track_token_usage_internal(usage_data)
-
- if tool_calls and available_functions:
- for call_data in tool_calls.values():
- function_name = call_data["name"]
- arguments = call_data["arguments"]
-
- if not function_name or not arguments:
- continue
-
- if function_name not in available_functions:
- logging.warning(
- f"Function '{function_name}' not found in available functions"
- )
- continue
-
- try:
- function_args = json.loads(arguments)
- except json.JSONDecodeError as e:
- logging.error(f"Failed to parse streamed tool arguments: {e}")
- continue
-
- result = self._handle_tool_execution(
- function_name=function_name,
- function_args=function_args,
- available_functions=available_functions,
- from_task=from_task,
- from_agent=from_agent,
- )
-
- if result is not None:
- return result
-
- full_response = self._apply_stop_words(full_response)
-
- self._emit_call_completed_event(
- response=full_response,
- call_type=LLMCallType.LLM_CALL,
+ return self._finalize_streaming_response(
+ full_response=full_response,
+ tool_calls=tool_calls,
+ usage_data=usage_data,
+ params=params,
+ available_functions=available_functions,
from_task=from_task,
from_agent=from_agent,
- messages=params["messages"],
)
- return full_response
-
def supports_function_calling(self) -> bool:
"""Check if the model supports function calling."""
return not self.is_o1_model
diff --git a/lib/crewai/tests/llms/openai/test_openai.py b/lib/crewai/tests/llms/openai/test_openai.py
index f88d8639c..a75a37681 100644
--- a/lib/crewai/tests/llms/openai/test_openai.py
+++ b/lib/crewai/tests/llms/openai/test_openai.py
@@ -1,6 +1,7 @@
import os
import sys
import types
+from typing import Any
from unittest.mock import patch, MagicMock
import openai
import pytest
@@ -1578,3 +1579,167 @@ def test_openai_structured_output_preserves_json_with_stop_word_patterns():
assert "Action:" in result.action_taken
assert "Observation:" in result.observation_result
assert "Final Answer:" in result.final_answer
+
+
+def test_openai_streaming_returns_tool_calls_without_available_functions():
+ """Test that streaming returns tool calls list when available_functions is None.
+
+ This mirrors the non-streaming path where tool_calls are returned for
+ the executor to handle. Reproduces the bug where streaming with tool
+ calls would return empty text instead of tool_calls when
+ available_functions was not provided (as the crew executor does).
+ """
+ llm = LLM(model="openai/gpt-4o-mini", stream=True)
+
+ mock_chunk_1 = MagicMock()
+ mock_chunk_1.choices = [MagicMock()]
+ mock_chunk_1.choices[0].delta = MagicMock()
+ mock_chunk_1.choices[0].delta.content = None
+ mock_chunk_1.choices[0].delta.tool_calls = [MagicMock()]
+ mock_chunk_1.choices[0].delta.tool_calls[0].index = 0
+ mock_chunk_1.choices[0].delta.tool_calls[0].id = "call_abc123"
+ mock_chunk_1.choices[0].delta.tool_calls[0].function = MagicMock()
+ mock_chunk_1.choices[0].delta.tool_calls[0].function.name = "calculator"
+ mock_chunk_1.choices[0].delta.tool_calls[0].function.arguments = '{"expr'
+ mock_chunk_1.choices[0].finish_reason = None
+ mock_chunk_1.usage = None
+ mock_chunk_1.id = "chatcmpl-1"
+
+ mock_chunk_2 = MagicMock()
+ mock_chunk_2.choices = [MagicMock()]
+ mock_chunk_2.choices[0].delta = MagicMock()
+ mock_chunk_2.choices[0].delta.content = None
+ mock_chunk_2.choices[0].delta.tool_calls = [MagicMock()]
+ mock_chunk_2.choices[0].delta.tool_calls[0].index = 0
+ mock_chunk_2.choices[0].delta.tool_calls[0].id = None
+ mock_chunk_2.choices[0].delta.tool_calls[0].function = MagicMock()
+ mock_chunk_2.choices[0].delta.tool_calls[0].function.name = None
+ mock_chunk_2.choices[0].delta.tool_calls[0].function.arguments = 'ession": "1+1"}'
+ mock_chunk_2.choices[0].finish_reason = None
+ mock_chunk_2.usage = None
+ mock_chunk_2.id = "chatcmpl-1"
+
+ mock_chunk_3 = MagicMock()
+ mock_chunk_3.choices = [MagicMock()]
+ mock_chunk_3.choices[0].delta = MagicMock()
+ mock_chunk_3.choices[0].delta.content = None
+ mock_chunk_3.choices[0].delta.tool_calls = None
+ mock_chunk_3.choices[0].finish_reason = "tool_calls"
+ mock_chunk_3.usage = MagicMock()
+ mock_chunk_3.usage.prompt_tokens = 10
+ mock_chunk_3.usage.completion_tokens = 5
+ mock_chunk_3.id = "chatcmpl-1"
+
+ with patch.object(
+ llm.client.chat.completions, "create", return_value=iter([mock_chunk_1, mock_chunk_2, mock_chunk_3])
+ ):
+ result = llm.call(
+ messages=[{"role": "user", "content": "Calculate 1+1"}],
+ tools=[{
+ "type": "function",
+ "function": {
+ "name": "calculator",
+ "description": "Calculate expression",
+ "parameters": {"type": "object", "properties": {"expression": {"type": "string"}}},
+ },
+ }],
+ available_functions=None,
+ )
+
+ assert isinstance(result, list), f"Expected list of tool calls, got {type(result)}: {result}"
+ assert len(result) == 1
+ assert result[0]["function"]["name"] == "calculator"
+ assert result[0]["function"]["arguments"] == '{"expression": "1+1"}'
+ assert result[0]["id"] == "call_abc123"
+ assert result[0]["type"] == "function"
+
+
+async def test_openai_async_streaming_returns_tool_calls_without_available_functions():
+ """Test that async streaming returns tool calls list when available_functions is None.
+
+ Same as the sync test but for the async path (_ahandle_streaming_completion).
+ """
+ llm = LLM(model="openai/gpt-4o-mini", stream=True)
+
+ mock_chunk_1 = MagicMock()
+ mock_chunk_1.choices = [MagicMock()]
+ mock_chunk_1.choices[0].delta = MagicMock()
+ mock_chunk_1.choices[0].delta.content = None
+ mock_chunk_1.choices[0].delta.tool_calls = [MagicMock()]
+ mock_chunk_1.choices[0].delta.tool_calls[0].index = 0
+ mock_chunk_1.choices[0].delta.tool_calls[0].id = "call_abc123"
+ mock_chunk_1.choices[0].delta.tool_calls[0].function = MagicMock()
+ mock_chunk_1.choices[0].delta.tool_calls[0].function.name = "calculator"
+ mock_chunk_1.choices[0].delta.tool_calls[0].function.arguments = '{"expr'
+ mock_chunk_1.choices[0].finish_reason = None
+ mock_chunk_1.usage = None
+ mock_chunk_1.id = "chatcmpl-1"
+
+ mock_chunk_2 = MagicMock()
+ mock_chunk_2.choices = [MagicMock()]
+ mock_chunk_2.choices[0].delta = MagicMock()
+ mock_chunk_2.choices[0].delta.content = None
+ mock_chunk_2.choices[0].delta.tool_calls = [MagicMock()]
+ mock_chunk_2.choices[0].delta.tool_calls[0].index = 0
+ mock_chunk_2.choices[0].delta.tool_calls[0].id = None
+ mock_chunk_2.choices[0].delta.tool_calls[0].function = MagicMock()
+ mock_chunk_2.choices[0].delta.tool_calls[0].function.name = None
+ mock_chunk_2.choices[0].delta.tool_calls[0].function.arguments = 'ession": "1+1"}'
+ mock_chunk_2.choices[0].finish_reason = None
+ mock_chunk_2.usage = None
+ mock_chunk_2.id = "chatcmpl-1"
+
+ mock_chunk_3 = MagicMock()
+ mock_chunk_3.choices = [MagicMock()]
+ mock_chunk_3.choices[0].delta = MagicMock()
+ mock_chunk_3.choices[0].delta.content = None
+ mock_chunk_3.choices[0].delta.tool_calls = None
+ mock_chunk_3.choices[0].finish_reason = "tool_calls"
+ mock_chunk_3.usage = MagicMock()
+ mock_chunk_3.usage.prompt_tokens = 10
+ mock_chunk_3.usage.completion_tokens = 5
+ mock_chunk_3.id = "chatcmpl-1"
+
+ class MockAsyncStream:
+ """Async iterator that mimics OpenAI's async streaming response."""
+
+ def __init__(self, chunks: list[Any]) -> None:
+ self._chunks = chunks
+ self._index = 0
+
+ def __aiter__(self) -> "MockAsyncStream":
+ return self
+
+ async def __anext__(self) -> Any:
+ if self._index >= len(self._chunks):
+ raise StopAsyncIteration
+ chunk = self._chunks[self._index]
+ self._index += 1
+ return chunk
+
+ async def mock_create(**kwargs: Any) -> MockAsyncStream:
+ return MockAsyncStream([mock_chunk_1, mock_chunk_2, mock_chunk_3])
+
+ with patch.object(
+ llm.async_client.chat.completions, "create", side_effect=mock_create
+ ):
+ result = await llm.acall(
+ messages=[{"role": "user", "content": "Calculate 1+1"}],
+ tools=[{
+ "type": "function",
+ "function": {
+ "name": "calculator",
+ "description": "Calculate expression",
+ "parameters": {"type": "object", "properties": {"expression": {"type": "string"}}},
+ },
+ }],
+ available_functions=None,
+ )
+
+ assert isinstance(result, list), f"Expected list of tool calls, got {type(result)}: {result}"
+ assert len(result) == 1
+ assert result[0]["function"]["name"] == "calculator"
+ assert result[0]["function"]["arguments"] == '{"expression": "1+1"}'
+ assert result[0]["id"] == "call_abc123"
+ assert result[0]["type"] == "function"