andre/crewAI
1
0
Fork 0
mirror of https://github.com/crewAIInc/crewAI.git synced 2026-01-22 06:18:14 +00:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

fix llm_call_completed event serialization issue

Browse Source
This commit is contained in:
lorenzejay
2026-01-21 12:59:28 -08:00
parent d6e04ba24d
commit 7d5a64af0d
7 changed files with 109 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
lib/crewai/src/crewai/llms/base_llm.py
View File

@@ -292,14 +292,16 @@ class BaseLLM(ABC):
from_agent: Agent | None = None,
) -> None:
"""Emit LLM call started event."""
from crewai.utilities.serialization import to_serializable
if not hasattr(crewai_event_bus, "emit"):
raise ValueError("crewai_event_bus does not have an emit method") from None
crewai_event_bus.emit(
self,
event=LLMCallStartedEvent(
messages=messages,
tools=tools,
messages=to_serializable(messages),
tools=to_serializable(tools),
callbacks=callbacks,
available_functions=available_functions,
from_task=from_task,
@@ -317,11 +319,13 @@ class BaseLLM(ABC):
messages: str | list[LLMMessage] | None = None,
) -> None:
"""Emit LLM call completed event."""
from crewai.utilities.serialization import to_serializable
crewai_event_bus.emit(
self,
event=LLMCallCompletedEvent(
messages=messages,
response=response,
messages=to_serializable(messages),
response=to_serializable(response),
call_type=call_type,
from_task=from_task,
from_agent=from_agent,

14
lib/crewai/src/crewai/llms/providers/anthropic/completion.py
View File

@@ -589,6 +589,13 @@ class AnthropicCompletion(BaseLLM):
# This allows the executor to manage tool execution with proper
# message history and post-tool reasoning prompts
if not available_functions:
self._emit_call_completed_event(
response=list(tool_uses),
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
return list(tool_uses)
# Handle tool use conversation flow internally
@@ -1004,6 +1011,13 @@ class AnthropicCompletion(BaseLLM):
if tool_uses:
# If no available_functions, return tool calls for executor to handle
if not available_functions:
self._emit_call_completed_event(
response=list(tool_uses),
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
return list(tool_uses)
return await self._ahandle_tool_use_conversation(

17
lib/crewai/src/crewai/llms/providers/azure/completion.py
View File

@@ -628,6 +628,13 @@ class AzureCompletion(BaseLLM):
# If there are tool_calls but no available_functions, return the tool_calls
# This allows the caller (e.g., executor) to handle tool execution
if message.tool_calls and not available_functions:
self._emit_call_completed_event(
response=list(message.tool_calls),
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
return list(message.tool_calls)
# Handle tool calls
@@ -804,7 +811,7 @@ class AzureCompletion(BaseLLM):
# If there are tool_calls but no available_functions, return them
# in OpenAI-compatible format for executor to handle
if tool_calls and not available_functions:
return [
formatted_tool_calls = [
{
"id": call_data.get("id", f"call_{idx}"),
"type": "function",
@@ -815,6 +822,14 @@ class AzureCompletion(BaseLLM):
}
for idx, call_data in tool_calls.items()
]
self._emit_call_completed_event(
response=formatted_tool_calls,
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
return formatted_tool_calls
# Handle completed tool calls
if tool_calls and available_functions:

24
lib/crewai/src/crewai/llms/providers/bedrock/completion.py
View File

@@ -546,6 +546,18 @@ class BedrockCompletion(BaseLLM):
"I apologize, but I received an empty response. Please try again."
)
# If there are tool uses but no available_functions, return them for the executor to handle
tool_uses = [block["toolUse"] for block in content if "toolUse" in block]
if tool_uses and not available_functions:
self._emit_call_completed_event(
response=tool_uses,
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=messages,
)
return tool_uses
# Process content blocks and handle tool use correctly
text_content = ""
@@ -935,6 +947,18 @@ class BedrockCompletion(BaseLLM):
"I apologize, but I received an empty response. Please try again."
)
# If there are tool uses but no available_functions, return them for the executor to handle
tool_uses = [block["toolUse"] for block in content if "toolUse" in block]
if tool_uses and not available_functions:
self._emit_call_completed_event(
response=tool_uses,
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=messages,
)
return tool_uses
text_content = ""
for content_block in content:

17
lib/crewai/src/crewai/llms/providers/gemini/completion.py
View File

@@ -661,6 +661,13 @@ class GeminiCompletion(BaseLLM):
# If there are function calls but no available_functions,
# return them for the executor to handle (like OpenAI/Anthropic)
if function_call_parts and not available_functions:
self._emit_call_completed_event(
response=function_call_parts,
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=self._convert_contents_to_dict(contents),
)
return function_call_parts
# Otherwise execute the tools internally
@@ -799,7 +806,7 @@ class GeminiCompletion(BaseLLM):
# If there are function calls but no available_functions,
# return them for the executor to handle
if function_calls and not available_functions:
return [
formatted_function_calls = [
{
"id": call_data["id"],
"function": {
@@ -810,6 +817,14 @@ class GeminiCompletion(BaseLLM):
}
for call_data in function_calls.values()
]
self._emit_call_completed_event(
response=formatted_function_calls,
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=self._convert_contents_to_dict(contents),
)
return formatted_function_calls
# Handle completed function calls
if function_calls and available_functions:

17
lib/crewai/src/crewai/llms/providers/openai/completion.py
View File

@@ -431,6 +431,13 @@ class OpenAICompletion(BaseLLM):
# If there are tool_calls but no available_functions, return the tool_calls
# This allows the caller (e.g., executor) to handle tool execution
if message.tool_calls and not available_functions:
self._emit_call_completed_event(
response=list(message.tool_calls),
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
return list(message.tool_calls)
# If there are tool_calls and available_functions, execute the tools
@@ -734,9 +741,13 @@ class OpenAICompletion(BaseLLM):
# If there are tool_calls but no available_functions, return the tool_calls
# This allows the caller (e.g., executor) to handle tool execution
if message.tool_calls and not available_functions:
print("--------------------------------")
print("lorenze tool_calls", list(message.tool_calls))
print("--------------------------------")
self._emit_call_completed_event(
response=list(message.tool_calls),
call_type=LLMCallType.TOOL_CALL,
from_task=from_task,
from_agent=from_agent,
messages=params["messages"],
)
return list(message.tool_calls)
# If there are tool_calls and available_functions, execute the tools

22
lib/crewai/src/crewai/utilities/serialization.py
View File

@@ -66,11 +66,23 @@ def to_serializable(
if key not in exclude
}
if isinstance(obj, BaseModel):
return to_serializable(
obj=obj.model_dump(exclude=exclude),
max_depth=max_depth,
_current_depth=_current_depth + 1,
)
try:
return to_serializable(
obj=obj.model_dump(exclude=exclude),
max_depth=max_depth,
_current_depth=_current_depth + 1,
)
except Exception:
try:
return {
_to_serializable_key(k): to_serializable(
v, max_depth=max_depth, _current_depth=_current_depth + 1
)
for k, v in obj.__dict__.items()
if k not in (exclude or set())
}
except Exception:
return repr(obj)
return repr(obj)