Skip to content

Open-source Python library for AI browser agents, plus Browser Use Cloud: stealth browsers and a hosted agent API.

address cubic comments and clean up new example

Alezander9bf15d159da10
diff --git a/browser_use/agent/service.py b/browser_use/agent/service.py
index 8396957a9..432d737fb 100644
--- a/browser_use/agent/service.py
+++ b/browser_use/agent/service.py
@@ -929,7 +929,11 @@ class Agent(Generic[Context, AgentStructuredOutput]):
)
# Call LLM with JudgementResult as output format
- kwargs: dict = {'output_format': JudgementResult, 'request_type': 'judge'}
+ kwargs: dict = {'output_format': JudgementResult}
+
+ # Only pass request_type for ChatBrowserUse (other providers don't support it)
+ if self.judge_llm.provider == 'browser-use':
+ kwargs['request_type'] = 'judge'
try:
response = await self.judge_llm.ainvoke(input_messages, **kwargs)
diff --git a/browser_use/llm/browser_use/chat.py b/browser_use/llm/browser_use/chat.py
index 4298dc0f9..9e97ecbb3 100644
--- a/browser_use/llm/browser_use/chat.py
+++ b/browser_use/llm/browser_use/chat.py
@@ -79,10 +79,14 @@ class ChatBrowserUse(BaseChatModel):
return self.model
@overload
- async def ainvoke(self, messages: list[BaseMessage], output_format: None = None) -> ChatInvokeCompletion[str]: ...
+ async def ainvoke(
+ self, messages: list[BaseMessage], output_format: None = None, request_type: str = 'browser_agent'
+ ) -> ChatInvokeCompletion[str]: ...
@overload
- async def ainvoke(self, messages: list[BaseMessage], output_format: type[T]) -> ChatInvokeCompletion[T]: ...
+ async def ainvoke(
+ self, messages: list[BaseMessage], output_format: type[T], request_type: str = 'browser_agent'
+ ) -> ChatInvokeCompletion[T]: ...
@observe(name='chat_browser_use_ainvoke')
async def ainvoke(
diff --git a/examples/features/judge_trace.py b/examples/features/judge_trace.py
index e23c1b07d..d89673668 100644
--- a/examples/features/judge_trace.py
+++ b/examples/features/judge_trace.py
@@ -19,9 +19,7 @@ from browser_use import Agent, ChatBrowserUse
async def main():
- llm = ChatBrowserUse(base_url='http://localhost:8080')
- # Use gemini flash for judging since it supports vision + structured output
- # judge_llm = ChatGoogle(model='gemini-flash-latest')
+ llm = ChatBrowserUse()
task = "Search Google for 'what is browser automation' and tell me the top 3 results"
agent = Agent(task=task, llm=llm, use_judge=True, judge_llm=llm)
history = await agent.run()