Skip to content

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

Update LLM integration in evaluate_tasks.py to use ChatGoogle instead of ChatOpenAI for agent and judge initialization

Magnus Müller7a463675ffc0
diff --git a/tests/ci/evaluate_tasks.py b/tests/ci/evaluate_tasks.py
index b4b4736fd..034de133c 100644
--- a/tests/ci/evaluate_tasks.py
+++ b/tests/ci/evaluate_tasks.py
@@ -17,7 +17,7 @@ import anyio
import yaml
from pydantic import BaseModel
-from browser_use import Agent, AgentHistoryList, BrowserProfile, BrowserSession, ChatOpenAI
+from browser_use import Agent, AgentHistoryList, BrowserProfile, BrowserSession, ChatGoogle
from browser_use.llm.messages import UserMessage
# --- CONFIG ---
@@ -56,8 +56,8 @@ async def run_single_task(task_file):
print(f'[DEBUG] Task: {task[:100]}...', file=sys.stderr)
print(f'[DEBUG] Max steps: {max_steps}', file=sys.stderr)
- agent_llm = ChatOpenAI(model='gpt-4.1-mini')
- judge_llm = ChatOpenAI(model='gpt-4.1-mini')
+ agent_llm = ChatGoogle(model='gemini-flash-latest')
+ judge_llm = ChatGoogle(model='gemini-flash-latest')
print('[DEBUG] LLMs initialized', file=sys.stderr)
# Each subprocess gets its own profile and session