vodorod/Dorod-Sky
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

refactor gemini reasoning effor (#3292)

Browse Source
This commit is contained in:
LawyZheng
2025-08-25 23:42:42 +08:00
committed by GitHub
parent f0134cbcb9
commit e0043d002c
3 changed files with 57 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
skyvern/config.py
View File

@@ -245,7 +245,8 @@ class Settings(BaseSettings):
# GEMINI
GEMINI_API_KEY: str | None = None
GEMINI_REASONING_EFFORT: str | None = None
GEMINI_INCLUDE_THOUGHT: bool = False
GEMINI_THINKING_BUDGET: int | None = None
# VERTEX_AI
VERTEX_CREDENTIALS: str | None = None

65
skyvern/forge/sdk/api/llm/config_registry.py
View File

@@ -787,7 +787,12 @@ if settings.ENABLE_GEMINI:
supports_vision=True,
add_assistant_prefix=False,
max_completion_tokens=65536,
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
litellm_params=LiteLLMParams(
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
),
)
LLMConfigRegistry.register_config(
@@ -798,7 +803,12 @@ if settings.ENABLE_GEMINI:
supports_vision=True,
add_assistant_prefix=False,
max_completion_tokens=65536,
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
litellm_params=LiteLLMParams(
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
),
)
LLMConfigRegistry.register_config(
@@ -809,7 +819,12 @@ if settings.ENABLE_GEMINI:
supports_vision=True,
add_assistant_prefix=False,
max_completion_tokens=65536,
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
litellm_params=LiteLLMParams(
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
),
)
LLMConfigRegistry.register_config(
@@ -820,7 +835,12 @@ if settings.ENABLE_GEMINI:
supports_vision=True,
add_assistant_prefix=False,
max_completion_tokens=65536,
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
litellm_params=LiteLLMParams(
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
),
)
LLMConfigRegistry.register_config(
@@ -831,7 +851,12 @@ if settings.ENABLE_GEMINI:
supports_vision=True,
add_assistant_prefix=False,
max_completion_tokens=65536,
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
litellm_params=LiteLLMParams(
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
),
)
@@ -1027,8 +1052,11 @@ if settings.ENABLE_VERTEX_AI and settings.VERTEX_CREDENTIALS:
vertex_credentials=settings.VERTEX_CREDENTIALS,
api_base=f"{api_base}/gemini-2.5-pro" if api_base else None,
vertex_location=settings.VERTEX_LOCATION,
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
),
)
LLMConfigRegistry.register_config(
@@ -1043,8 +1071,11 @@ if settings.ENABLE_VERTEX_AI and settings.VERTEX_CREDENTIALS:
vertex_credentials=settings.VERTEX_CREDENTIALS,
api_base=f"{api_base}/gemini-2.5-pro-preview-05-06" if api_base else None,
vertex_location=settings.VERTEX_LOCATION,
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
),
)
LLMConfigRegistry.register_config(
@@ -1059,8 +1090,11 @@ if settings.ENABLE_VERTEX_AI and settings.VERTEX_CREDENTIALS:
vertex_credentials=settings.VERTEX_CREDENTIALS,
api_base=f"{api_base}/gemini-2.5-flash" if api_base else None,
vertex_location=settings.VERTEX_LOCATION,
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
),
)
LLMConfigRegistry.register_config(
@@ -1075,8 +1109,11 @@ if settings.ENABLE_VERTEX_AI and settings.VERTEX_CREDENTIALS:
vertex_credentials=settings.VERTEX_CREDENTIALS,
api_base=f"{api_base}/gemini-2.5-flash-preview-05-20" if api_base else None,
vertex_location=settings.VERTEX_LOCATION,
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
),
)
LLMConfigRegistry.register_config(
@@ -1091,8 +1128,11 @@ if settings.ENABLE_VERTEX_AI and settings.VERTEX_CREDENTIALS:
vertex_credentials=settings.VERTEX_CREDENTIALS,
api_base=f"{api_base}/gemini-2.5-flash-preview-04-17" if api_base else None,
vertex_location=settings.VERTEX_LOCATION,
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
),
)
LLMConfigRegistry.register_config(
@@ -1107,8 +1147,11 @@ if settings.ENABLE_VERTEX_AI and settings.VERTEX_CREDENTIALS:
vertex_credentials=settings.VERTEX_CREDENTIALS,
api_base=f"{api_base}/gemini-2.5-flash-preview-05-20" if api_base else None,
vertex_location=settings.VERTEX_LOCATION,
thinking={
"budget_tokens": settings.GEMINI_THINKING_BUDGET,
"type": "enabled" if settings.GEMINI_INCLUDE_THOUGHT else None,
},
),
reasoning_effort=settings.GEMINI_REASONING_EFFORT,
),
)
LLMConfigRegistry.register_config(

1
skyvern/forge/sdk/api/llm/models.py
View File

@@ -16,6 +16,7 @@ class LiteLLMParams(TypedDict, total=False):
model_info: dict[str, Any] | None
vertex_credentials: str | None
vertex_location: str | None
thinking: dict[str, Any] | None
@dataclass(frozen=True)