Feature request: Option to run a native browser instead of playwright

[Eloitor]

I tried this and it worked. It would be nice if there was an option to enable this without modifying the code.

diff --git a/operator_agent.py b/operator_agent.py
index 3ce6401..1bc7f73 100644
--- a/operator_agent.py
+++ b/operator_agent.py
@@ -5,11 +5,14 @@ import httpx
 import asyncio
 from langchain_openai import ChatOpenAI
 from langchain_community.llms import Ollama
-from browser_use import Agent
+from browser_use import Agent, Browser, BrowserConfig
 # from dotenv import load_dotenv
 
 # load_dotenv()
 from api_secrets import DEEPSEEK_API_KEY, OPENROUTER_API_KEY
+browser = Browser(
+    config=BrowserConfig(chrome_instance_path='/usr/bin/chromium')
+)
 
 class OllamaOperator:
     def __init__(self, model="deepseek-r1:1.5b"):
@@ -22,7 +25,7 @@ class OllamaOperator:
         
     async def execute_task(self, task, max_steps=20):
         try:
-            agent = Agent(task=task, llm=self.client)
+            agent = Agent(task=task, llm=self.client, browser=browser)
             result = await agent.run(max_steps=max_steps)
             return f"Execution Result:\n{result}"
         except Exception as e:

[tomjuggler]

That makes a lot of sense. I will have a look - the only issue I can see for now is maybe Windows support..
Let's ask Deepseek and see

[tomjuggler]

So I'm going to leave this open for now. Having your own browser running opens up more possibilities so yeah it's a great idea.

To do it properly would take a lot of work I think. Right now this is just a demo, which works as-is. If I actually start using it for real things this is one of the first things I will look at.

Feel free to sort it out and add a pull request. DeepSeek has this to say:

Let's analyze the browser integration changes and cross-platform requirements:

1 Current Cross-Platform Issues:

• Hardcoded Chromium path (/usr/bin/chromium) is Linux-specific
• No platform detection logic
• Browser configuration is global rather than per-instance
• No apparent support for different browser types

2 Multi-Browser Support Requirements:

• Would need browser detection/selection logic
• Requires different driver binaries for each browser
• Path management for different OS browser locations
• Browser-specific configuration parameters

Recommended Changes:

1 Cross-Platform Improvements:

Use platform detection and standard browser locations

from sys import platform

def get_default_browser_path(browser_name):
if platform == "linux" or platform == "linux2":
paths = {
'chrome': '/usr/bin/google-chrome',
'chromium': '/usr/bin/chromium',
'firefox': '/usr/bin/firefox'
}
elif platform == "darwin":
paths = {
'chrome': '/Applications/Google Chrome.app/Contents/MacOS/Google Chrome',
'firefox': '/Applications/Firefox.app/Contents/MacOS/firefox'
}
elif platform == "win32":
paths = {
'chrome': r'C:\Program Files\Google\Chrome\Application\chrome.exe',
'edge': r'C:\Program Files (x86)\Microsoft\Edge\Application\msedge.exe'
}
return paths.get(browser_name)

2 Multi-Browser Support Implementation:

Modified BrowserConfig with browser type selection

browser = Browser(
config=BrowserConfig(
browser_type='chrome', # Default, can be chrome|firefox|edge|chromium
instance_path=get_default_browser_path('chrome'),
headless=True # Add common options
)
)

Add to argument parser:

parser.add_argument("--browser", choices=["chrome", "firefox", "edge", "chromium"],
default="chrome", help="Browser to use for automation")

Implementation Difficulty:

1 Cross-Platform (Medium Effort):

• Requires testing on all target OSes
• Need fallback paths/error handling
• Environment variable overrides for custom paths
• Consider using webdriver-manager package for automatic driver management

2 Multi-Browser Support (High Effort):

• Different browser automation quirks
• Varied DevTools protocol support
• Memory/performance characteristics per browser
• Consistent behavior across browsers would need abstraction

Suggested Implementation Path:

1 Make browser configuration dynamic:

In init() of your operators:

def init(self, model=..., browser_type='chrome'):
self.browser = Browser(
config=BrowserConfig(
browser_type=browser_type,
instance_path=get_default_browser_path(browser_type)
)
)

2 Use environment variables for paths:

Prefer environment-specified paths if available

instance_path = os.getenv('BROWSER_PATH') or get_default_browser_path(browser_type)

3 Add browser management to CLI arguments:

parser.add_argument("--browser", help="Browser type to use")
parser.add_argument("--browser-path", help="Custom browser executable path")