Reduce frequency of thinking budget optimization logs (#4746)

This commit is contained in:
pedrohsdb
2026-02-13 14:25:24 -08:00
committed by GitHub
parent 2953a35ac8
commit 1263c09aab

View File

@@ -289,7 +289,7 @@ class LLMAPIHandlerFactory:
else:
# Other reasoning-capable models (Deepseek, etc.) - use "low" for all budget values
parameters["reasoning_effort"] = "low"
LOG.info(
LOG.debug(
"Applied thinking budget optimization (reasoning_effort)",
prompt_name=prompt_name,
budget=new_budget,
@@ -320,7 +320,7 @@ class LLMAPIHandlerFactory:
if model_label is None and isinstance(llm_config, LLMRouterConfig):
model_label = getattr(llm_config, "main_model_group", "router")
LOG.info(
LOG.debug(
"Applied thinking budget optimization (reasoning_effort)",
prompt_name=prompt_name,
budget=new_budget,
@@ -338,7 +338,7 @@ class LLMAPIHandlerFactory:
if model_label is None and isinstance(llm_config, LLMRouterConfig):
model_label = getattr(llm_config, "main_model_group", "router")
LOG.info(
LOG.debug(
"Applied thinking budget optimization (thinking)",
prompt_name=prompt_name,
budget=new_budget,
@@ -366,7 +366,7 @@ class LLMAPIHandlerFactory:
thinking_payload["type"] = "enabled"
parameters["thinking"] = thinking_payload
LOG.info(
LOG.debug(
"Applied thinking budget optimization (budget_tokens)",
prompt_name=prompt_name,
budget=new_budget,