# 1. Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# 2. Pull a capable model (Qwen2.5:7b or Llama 3.1 8B recommended)
ollama pull qwen2.5:7b

# 3. Install Python dependencies
pip install browser-use playwright

# 4. Install Playwright browsers
playwright install chromium

python3 -c "from browser_use import Agent; print('✅ Browser-Use ready')"
playwright --version
ollama list

mkdir -p /opt/browser-agent
vim /opt/browser-agent/search_demo.py

#!/usr/bin/env python3
"""
Browser-Use Quickstart: Let AI search and screenshot
"""
import asyncio
from browser_use import Agent

async def main():
    agent = Agent(
        task="Go to https://www.google.com, search for 'Browser-Use GitHub', "
             "click the first result, wait for the page to load, "
             "and save a screenshot as result.png",
        llm="ollama/qwen2.5:7b",  # using local model
        use_vision=True,           # AI "sees" the page via screenshots
    )
    await agent.run()

if __name__ == "__main__":
    asyncio.run(main())

cd /opt/browser-agent
python3 search_demo.py

#!/usr/bin/env python3
"""
Browser-Use Headless Mode Setup
"""
import asyncio
from browser_use import Agent
from browser_use.browser.browser import Browser, BrowserConfig

async def main():
    # Configure headless browser
    browser = Browser(
        config=BrowserConfig(
            headless=True,                    # no GUI needed
            disable_security=True,            # avoid CORS issues
            window_w=1280, window_h=720,      # viewport size
        )
    )

    agent = Agent(
        task="Log in to https://example.com/admin, "
             "find the 'Today's Data' panel, "
             "screenshot the data table and save as dashboard.png",
        llm="ollama/qwen2.5:7b",
        browser=browser,
        use_vision=True,
    )
    await agent.run()

if __name__ == "__main__":
    asyncio.run(main())

#!/usr/bin/env python3
"""
Production: AI auto-login and report export to CSV
"""
import asyncio
import csv
from pathlib import Path
from browser_use import Agent
from browser_use.browser.browser import Browser, BrowserConfig
from browser_use.controller.service import Controller

# Define custom actions
controller = Controller()

@controller.action("Save data to CSV")
def save_to_csv(data: str, filename: str):
    """Save AI-extracted data to a CSV file"""
    path = Path(f"/opt/browser-agent/output/{filename}")
    path.parent.mkdir(parents=True, exist_ok=True)
    lines = data.strip().split("\n")
    if lines:
        with open(path, "w", newline="", encoding="utf-8") as f:
            writer = csv.writer(f)
            for line in lines:
                writer.writerow(line.split(","))
    return f"Saved to {path}"

async def main():
    browser = Browser(
        config=BrowserConfig(headless=True, window_w=1280, window_h=720)
    )

    agent = Agent(
        task="""
        1. Open https://your-saas-dashboard.com/login
        2. Type 'admin@example.com' into the username field
        3. Type the password into the password field
        4. Click the login button
        5. Wait for the dashboard to fully load
        6. Click 'Reports' in the left navigation
        7. Select date range 'This Month'
        8. Click 'Export Data'
        9. Use save_to_csv to save the data as 'monthly_report.csv'
        10. Take a screenshot to confirm success
        """,
        llm="ollama/qwen2.5:7b",
        browser=browser,
        controller=controller,
        use_vision=True,
    )
    await agent.run()

if __name__ == "__main__":
    asyncio.run(main())

vim /opt/browser-agent/run_daily_report.py

#!/usr/bin/env python3
"""Daily automated report collection"""
import asyncio
from datetime import datetime
from browser_use import Agent
from browser_use.browser.browser import Browser, BrowserConfig

async def main():
    print(f"[{datetime.now()}] Starting daily report collection...")
    
    browser = Browser(
        config=BrowserConfig(headless=True)
    )

    agent = Agent(
        task="""
        1. Log in to https://analytics.example.com
        2. Navigate to 'Traffic Overview'
        3. Take a screenshot of the last 24h data, save as daily_traffic.png
        4. Find the 'Traffic Sources' table, extract Top 10 entries
        5. Save traffic sources using save_to_csv as traffic_sources.csv
        6. Check for anomalies (e.g., sudden traffic drop), note them on the page
        7. Take a final screenshot as daily_report_final.png
        """,
        llm="ollama/qwen2.5:7b",
        browser=browser,
        use_vision=True,
    )
    await agent.run()
    print(f"[{datetime.now()}] Report collection complete")

if __name__ == "__main__":
    asyncio.run(main())

chmod +x /opt/browser-agent/run_daily_report.py

# Run daily at 8 AM
(crontab -l 2>/dev/null; echo "0 8 * * * cd /opt/browser-agent && python3 run_daily_report.py >> /var/log/browser-agent.log 2>&1") | crontab -

# Monitor logs
tail -f /var/log/browser-agent.log

# OpenAI
Agent(
    task="...",
    llm="openai/gpt-4o",
    # Set OPENAI_API_KEY environment variable
)

# Anthropic Claude
Agent(
    task="...",
    llm="anthropic/claude-sonnet-4-20250514",
)

# DeepSeek (ultra-low cost)
Agent(
    task="...",
    llm="deepseek/deepseek-chat",
)

vim /opt/browser-agent/Dockerfile

FROM python:3.11-slim

# System dependencies for Chromium
RUN apt-get update && apt-get install -y \
    curl \
    fonts-liberation \
    libappindicator3-1 \
    libasound2 \
    libatk-bridge2.0-0 \
    libatk1.0-0 \
    libcups2 \
    libdbus-1-3 \
    libgdk-pixbuf2.0-0 \
    libnspr4 \
    libnss3 \
    libx11-xcb1 \
    libxcomposite1 \
    libxdamage1 \
    libxrandr2 \
    xdg-utils \
    && rm -rf /var/lib/apt/lists/*

# Install Python deps
RUN pip install browser-use playwright \
    && playwright install chromium \
    && playwright install-deps

WORKDIR /app
COPY . .

CMD ["python3", "run_daily_report.py"]

vim /opt/browser-agent/docker-compose.yml

version: '3.8'

services:
  browser-agent:
    build: .
    environment:
      - DASHBOARD_PASSWORD=${DASHBOARD_PASSWORD}
      - OPENAI_API_KEY=${OPENAI_API_KEY:-}
    volumes:
      - ./output:/app/output
      - ./screenshots:/app/screenshots
    restart: unless-stopped

# One-click start
cd /opt/browser-agent
docker-compose up -d
docker-compose logs -f

agent_1 = Agent(task="Log in and navigate to the reports page", ...)
await agent_1.run()

# Pass context to next agent
agent_2 = Agent(
    task="Extract all data tables from the current page",
    context=agent_1.context,
    ...
)
await agent_2.run()

from tenacity import retry, stop_after_attempt

@retry(stop=stop_after_attempt(3))
async def run_agent_with_retry(task: str):
    agent = Agent(task=task, ...)
    await agent.run()

# Open multiple tabs simultaneously
agent = Agent(
    task="Open a price comparison page in a new tab "
         "while keeping the login session active in the current tab",
    ...
)

@controller.action("Solve CAPTCHA")
def solve_captcha(image_path: str):
    """AI automatically reads and fills CAPTCHA"""
    # Browser-Use vision mode can read the CAPTCHA image directly
    pass