Skip to content

Automate browser based workflows with AI

feat(SKY-12788): add Gemini 3.5 Flash Lite and Gemini 3.6 Flash model support (#7617)

Suchintan2c7283f47062
diff --git a/skyvern-frontend/src/components/ModelSelector.test.tsx b/skyvern-frontend/src/components/ModelSelector.test.tsx
index 6d26b825f..9f531311b 100644
--- a/skyvern-frontend/src/components/ModelSelector.test.tsx
+++ b/skyvern-frontend/src/components/ModelSelector.test.tsx
@@ -182,4 +182,20 @@ describe("ModelSelector in a read-only comparison scope", () => {
expect(await screen.findByText("DeepSeek V4 Flash")).toBeTruthy();
expect(screen.getByText("Xiaomi MiMo V2.5")).toBeTruthy();
});
+
+ test("surfaces the new Gemini 3.5 Flash Lite and 3.6 Flash models", async () => {
+ getMock.mockResolvedValueOnce({
+ data: {
+ models: {
+ "gemini-3.5-flash-lite": "Gemini 3.5 Flash Lite",
+ "gemini-3.6-flash": "Gemini 3.6 Flash",
+ },
+ },
+ });
+
+ renderSelector(false, null);
+
+ expect(await screen.findByText("Gemini 3.5 Flash Lite")).toBeTruthy();
+ expect(screen.getByText("Gemini 3.6 Flash")).toBeTruthy();
+ });
});
diff --git a/skyvern/cli/llm_setup.py b/skyvern/cli/llm_setup.py
index 3d8eb96a4..d87841d49 100644
--- a/skyvern/cli/llm_setup.py
+++ b/skyvern/cli/llm_setup.py
@@ -176,8 +176,10 @@ def setup_llm_providers(env_path: Path | str | None = None) -> None:
model_options.extend(
[
"GEMINI_3.1_PRO",
+ "GEMINI_3.6_FLASH",
"GEMINI_3.5_FLASH",
"GEMINI_3.0_FLASH",
+ "GEMINI_3.5_FLASH_LITE",
"GEMINI_3.1_FLASH_LITE",
"GEMINI_2.5_PRO",
"GEMINI_2.5_FLASH",
diff --git a/skyvern/config.py b/skyvern/config.py
index a3df7f249..d697fda59 100644
--- a/skyvern/config.py
+++ b/skyvern/config.py
@@ -796,6 +796,8 @@ class Settings(BaseSettings):
("gemini-3-pro-preview", "VERTEX_GEMINI_3_PRO", "GEMINI_3_PRO", "Gemini 3 Pro (Latest)"),
("gemini-3.0-flash", "VERTEX_GEMINI_3.0_FLASH", "GEMINI_3.0_FLASH", "Gemini 3 Flash"),
("gemini-3.5-flash", "VERTEX_GEMINI_3.5_FLASH", "GEMINI_3.5_FLASH", "Gemini 3.5 Flash"),
+ ("gemini-3.5-flash-lite", "VERTEX_GEMINI_3.5_FLASH_LITE", "GEMINI_3.5_FLASH_LITE", "Gemini 3.5 Flash Lite"),
+ ("gemini-3.6-flash", "VERTEX_GEMINI_3.6_FLASH", "GEMINI_3.6_FLASH", "Gemini 3.6 Flash"),
]
for model_name, vertex_key, gemini_key, label in gemini_models:
mapping[model_name] = {
diff --git a/skyvern/forge/sdk/api/llm/config_registry.py b/skyvern/forge/sdk/api/llm/config_registry.py
index 3552cf52f..5fea70a5e 100644
--- a/skyvern/forge/sdk/api/llm/config_registry.py
+++ b/skyvern/forge/sdk/api/llm/config_registry.py
@@ -1333,6 +1333,19 @@ if settings.ENABLE_GEMINI:
),
),
)
+ LLMConfigRegistry.register_config(
+ "GEMINI_3.6_FLASH",
+ LLMConfig(
+ "gemini/gemini-3.6-flash",
+ ["GEMINI_API_KEY"],
+ supports_vision=True,
+ add_assistant_prefix=False,
+ max_completion_tokens=65536,
+ litellm_params=LiteLLMParams(
+ thinking_level="medium" if settings.GEMINI_INCLUDE_THOUGHT else "minimal",
+ ),
+ ),
+ )
LLMConfigRegistry.register_config(
"GEMINI_3_PRO",
LLMConfig(
@@ -1359,6 +1372,19 @@ if settings.ENABLE_GEMINI:
),
),
)
+ LLMConfigRegistry.register_config(
+ "GEMINI_3.5_FLASH_LITE",
+ LLMConfig(
+ "gemini/gemini-3.5-flash-lite",
+ ["GEMINI_API_KEY"],
+ supports_vision=True,
+ add_assistant_prefix=False,
+ max_completion_tokens=65536,
+ litellm_params=LiteLLMParams(
+ thinking_level="medium" if settings.GEMINI_INCLUDE_THOUGHT else "minimal",
+ ),
+ ),
+ )
# Backward compat alias for non-Vertex Gemini 3 Pro
LLMConfigRegistry.register_config(
"GEMINI_3.1_PRO",
@@ -1746,6 +1772,22 @@ if settings.ENABLE_VERTEX_AI:
),
),
)
+ LLMConfigRegistry.register_config(
+ "VERTEX_GEMINI_3.6_FLASH",
+ LLMConfig(
+ "vertex_ai/gemini-3.6-flash",
+ [],
+ supports_vision=True,
+ add_assistant_prefix=False,
+ max_completion_tokens=65536,
+ litellm_params=LiteLLMParams(
+ api_base=f"{api_base}/gemini-3.6-flash" if api_base else None,
+ vertex_location=settings.VERTEX_LOCATION,
+ thinking_level="medium" if settings.GEMINI_INCLUDE_THOUGHT else "minimal",
+ vertex_credentials=settings.VERTEX_CREDENTIALS,
+ ),
+ ),
+ )
LLMConfigRegistry.register_config(
"VERTEX_GEMINI_3.1_FLASH_LITE",
LLMConfig(
@@ -1762,6 +1804,22 @@ if settings.ENABLE_VERTEX_AI:
),
),
)
+ LLMConfigRegistry.register_config(
+ "VERTEX_GEMINI_3.5_FLASH_LITE",
+ LLMConfig(
+ "vertex_ai/gemini-3.5-flash-lite",
+ [],
+ supports_vision=True,
+ add_assistant_prefix=False,
+ max_completion_tokens=65536,
+ litellm_params=LiteLLMParams(
+ api_base=f"{api_base}/gemini-3.5-flash-lite" if api_base else None,
+ vertex_location=settings.VERTEX_LOCATION,
+ thinking_level="medium" if settings.GEMINI_INCLUDE_THOUGHT else "minimal",
+ vertex_credentials=settings.VERTEX_CREDENTIALS,
+ ),
+ ),
+ )
# Backward compat aliases — both resolve to the canonical VERTEX_GEMINI_3_PRO.
# Bump VERTEX_GEMINI_3_PRO above when Google ships a newer version.
LLMConfigRegistry.register_config(
@@ -1884,6 +1942,25 @@ if settings.ENABLE_VERTEX_AI:
),
),
)
+ LLMConfigRegistry.register_config(
+ "VERTEX_GEMINI_3.5_FLASH_LITE_FLEX",
+ LLMConfig(
+ "vertex_ai/gemini-3.5-flash-lite",
+ [],
+ supports_vision=True,
+ add_assistant_prefix=False,
+ max_completion_tokens=65536,
+ litellm_params=LiteLLMParams(
+ api_base=f"{api_base}/gemini-3.5-flash-lite" if api_base else None,
+ vertex_location=settings.VERTEX_LOCATION,
+ thinking_level="medium" if settings.GEMINI_INCLUDE_THOUGHT else "minimal",
+ vertex_credentials=settings.VERTEX_CREDENTIALS,
+ service_tier="SERVICE_TIER_FLEX",
+ extra_headers={"X-Vertex-AI-LLM-Shared-Request-Type": "flex", "X-Vertex-AI-LLM-Request-Type": "shared"},
+ timeout=FLEX_EXECUTION_TIMEOUT_SECONDS,
+ ),
+ ),
+ )
LLMConfigRegistry.register_config(
"VERTEX_GEMINI_3.1_PRO_FLEX",
LLMConfig(
@@ -1941,6 +2018,25 @@ if settings.ENABLE_VERTEX_AI:
),
),
)
+ LLMConfigRegistry.register_config(
+ "VERTEX_GEMINI_3.6_FLASH_FLEX",
+ LLMConfig(
+ "vertex_ai/gemini-3.6-flash",
+ [],
+ supports_vision=True,
+ add_assistant_prefix=False,
+ max_completion_tokens=65536,
+ litellm_params=LiteLLMParams(
+ api_base=f"{api_base}/gemini-3.6-flash" if api_base else None,
+ vertex_location=settings.VERTEX_LOCATION,
+ thinking_level="medium" if settings.GEMINI_INCLUDE_THOUGHT else "minimal",
+ vertex_credentials=settings.VERTEX_CREDENTIALS,
+ service_tier="SERVICE_TIER_FLEX",
+ extra_headers={"X-Vertex-AI-LLM-Shared-Request-Type": "flex", "X-Vertex-AI-LLM-Request-Type": "shared"},
+ timeout=FLEX_EXECUTION_TIMEOUT_SECONDS,
+ ),
+ ),
+ )
# Register old keys as aliases to prevent breaking existing tasks
LLMConfigRegistry.register_config(
"VERTEX_GEMINI_2.5_FLASH_PREVIEW_09_2025",