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

Tune reasoning effort, temp, top_p for gpt-oss models

Browse Source
This commit is contained in:
Debanjum
2025-08-19 23:29:42 -07:00
parent 83d725d2d8
commit dbc3330610
1 changed files with 10 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/khoj/processor/conversation/openai/utils.py
View File

@@ -145,6 +145,11 @@ def completion_with_backoff(
# See https://qwenlm.github.io/blog/qwen3/#advanced-usages
if not deepthought:
add_qwen_no_think_tag(formatted_messages)
elif "gpt-oss" in model_name.lower():
model_kwargs["temperature"] = 1
reasoning_effort = "medium" if deepthought else "low"
model_kwargs["reasoning_effort"] = reasoning_effort
model_kwargs["top_p"] = 1.0
read_timeout = 300 if is_local_api(api_base_url) else 60
if os.getenv("KHOJ_LLM_SEED"):
@@ -346,6 +351,11 @@ async def chat_completion_with_backoff(
# See https://qwenlm.github.io/blog/qwen3/#advanced-usages
if not deepthought:
add_qwen_no_think_tag(formatted_messages)
elif "gpt-oss" in model_name.lower():
temperature = 1
reasoning_effort = "medium" if deepthought else "low"
model_kwargs["reasoning_effort"] = reasoning_effort
model_kwargs["top_p"] = 1.0
read_timeout = 300 if is_local_api(api_base_url) else 60
if os.getenv("KHOJ_LLM_SEED"):