klbr/khoj
1
0
Fork 0
mirror of https://github.com/khoaliber/khoj.git synced 2026-03-02 21:19:12 +00:00
Code Issues Packages Projects Releases Wiki Activity

Only enable reasoning by qwen3 models in deepthought mode

Browse Source
This commit is contained in:
Debanjum
2025-05-02 10:49:42 -06:00
parent 6eaf54eb7a
commit d10dcc83d4
1 changed files with 8 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/khoj/processor/conversation/openai/utils.py
View File

@@ -81,6 +81,10 @@ def completion_with_backoff(
model_kwargs["reasoning_effort"] = reasoning_effort
elif is_qwen_reasoning_model(model_name, api_base_url):
stream_processor = partial(in_stream_thought_processor, thought_tag="think")
# Reasoning is enabled by default. Disable when deepthought is False.
# See https://qwenlm.github.io/blog/qwen3/#advanced-usages
if not deepthought and len(formatted_messages) > 0:
formatted_messages[-1]["content"] = formatted_messages[-1]["content"] + " /no_think"
model_kwargs["stream_options"] = {"include_usage": True}
if os.getenv("KHOJ_LLM_SEED"):
@@ -193,6 +197,10 @@ async def chat_completion_with_backoff(
formatted_messages = updated_messages
elif is_qwen_reasoning_model(model_name, api_base_url):
stream_processor = partial(ain_stream_thought_processor, thought_tag="think")
# Reasoning is enabled by default. Disable when deepthought is False.
# See https://qwenlm.github.io/blog/qwen3/#advanced-usages
if not deepthought and len(formatted_messages) > 0:
formatted_messages[-1]["content"] = formatted_messages[-1]["content"] + " /no_think"
stream = True
model_kwargs["stream_options"] = {"include_usage": True}