browser_use.md

browser-use Integration

The browser-use package can be used as a LightAgent tool for browser-based data collection. With browser-use 0.11 and newer, some versions expect the LangChain chat model object to expose a provider attribute. Older langchain_openai.ChatOpenAI objects may not provide that attribute.

LightAgent examples include a small compatibility helper:

def ensure_browser_use_provider(llm, provider: str = "openai"):
    if hasattr(llm, "provider"):
        return llm
    try:
        setattr(llm, "provider", provider)
    except Exception:
        object.__setattr__(llm, "provider", provider)
    return llm

Use it when constructing the browser-use agent:

import os
from browser_use import Agent
from langchain_openai import ChatOpenAI

llm = ensure_browser_use_provider(ChatOpenAI(
    model=os.getenv("BROWSER_USE_MODEL", "gpt-4.1-mini"),
    api_key=os.getenv("OPENAI_API_KEY"),
    base_url=os.getenv("OPENAI_BASE_URL", "https://api.openai.com/v1"),
))

browser_agent = Agent(
    task="Search for the current weather in Shanghai.",
    llm=llm,
    use_vision=False,
)

Notes

• Prefer reading API keys from environment variables in new examples. Existing reusable tools may keep an api_key parameter for backward compatibility.
• Keep browser tasks narrow and explicit. Browser automation is slower and more failure-prone than direct API tools.
• If browser-use changes its LLM adapter requirements, update the helper in the example and any project-specific browser tools that construct ChatOpenAI.