Skip to content

Add reasoning field to LLMResponse for reasoning models #140

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
ArthurConmy wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Add reasoning field to LLMResponse for reasoning models #140

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions safetytooling/apis/inference/openrouter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -269,6 +269,7 @@ async def __call__(
generated_content = []
duration = None
api_duration = None
reasoning = None

error_list = []
for i in range(max_attempts):
Expand All @@ -295,6 +296,13 @@ async def __call__(

api_duration = time.time() - api_start

# Extract reasoning trace if present (for reasoning models like DeepSeek-R1, o1)
reasoning = None
if response_data.choices and response_data.choices[0].message:
msg = response_data.choices[0].message
if hasattr(msg, "reasoning") and msg.reasoning:
reasoning = msg.reasoning

if (
response_data.choices is None
or len(response_data.choices) == 0
Expand All @@ -314,6 +322,7 @@ async def __call__(
LLMResponse(
model_id=model_id,
completion="",
reasoning=reasoning,
generated_content=[],
stop_reason="stop_sequence",
api_duration=api_duration,
Expand Down Expand Up @@ -356,10 +365,15 @@ async def __call__(
assert not tools, "Multiple choices not supported with tools"
responses = []
for choice in response_data.choices:
# Extract per-choice reasoning if available
choice_reasoning = None
if hasattr(choice.message, "reasoning") and choice.message.reasoning:
choice_reasoning = choice.message.reasoning
responses.append(
LLMResponse(
model_id=model_id,
completion=choice.message.content or "",
reasoning=choice_reasoning,
generated_content=[self._convert_message_to_chat_message(choice.message)],
stop_reason=choice.finish_reason,
api_duration=api_duration,
Expand All @@ -378,6 +392,7 @@ async def __call__(
LLMResponse(
model_id=model_id,
completion=completion,
reasoning=reasoning,
generated_content=generated_content,
stop_reason=response_data.choices[0].finish_reason,
api_duration=api_duration,
Expand Down
1 change: 1 addition & 0 deletions safetytooling/data_models/inference.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,6 +71,7 @@ class LLMResponse(pydantic.BaseModel):
completion: str
stop_reason: StopReason | GeminiStopReason | GeminiBlockReason
cost: float = 0
reasoning: str | None = None # Reasoning trace from reasoning models (e.g., DeepSeek-R1, o1)
generated_content: List[ChatMessage] | None = None
audio_out: str | Path | None = None
duration: float | None = None
Expand Down