Browser Use Service¶

LLM-powered browser automation service for the infrastructure.

Quick Reference¶

Item	Value
Web UI	https://browser-use.rbnk.uk/ui/
REST API	https://browser-use.rbnk.uk/task
API Docs	https://browser-use.rbnk.uk/docs
Health	https://browser-use.rbnk.uk/health
Metrics	https://browser-use.rbnk.uk/metrics
Default Model	gpt-5.1
Stack Location	`/srv/dockerdata/browser-use/`

What It Does¶

Browser Use enables AI agents to control a web browser through natural language commands. Use cases include:

Web Scraping: Extract data from websites
Form Automation: Fill and submit forms
Testing: Automated UI testing
Research: Gather information from multiple sources
Data Entry: Automate repetitive web tasks

API Usage¶

Basic Request¶

curl -X POST https://browser-use.rbnk.uk/task \
  -H "Content-Type: application/json" \
  -H "X-API-Key: YOUR_API_KEY" \
  -d '{
    "task": "Go to example.com and get the page title",
    "max_steps": 10
  }'

Request Parameters¶

Parameter	Type	Default	Description
`task`	string	required	Natural language task
`backend`	string	playwright	`playwright` or `browserless`
`max_steps`	int	50	Max agent steps
`headless`	bool	true	Headless browser mode
`save_screenshot`	bool	false	Save final screenshot
`timeout`	int	300	Timeout in seconds
`model`	string	gpt-5.1	LLM model to use

Response¶

{
  "task_id": "uuid",
  "status": "success",
  "result": "The page title is 'Example Domain'",
  "steps_taken": 3,
  "duration_seconds": 15.5,
  "model_used": "gpt-5.1"
}

n8n Integration¶

Use in n8n workflows with HTTP Request node:

URL: https://browser-use.rbnk.uk/task
Method: POST
Auth: Header Auth with X-API-Key
Body: JSON with task parameters

Operations¶

# Start service
cd /srv/dockerdata/browser-use && docker compose up -d

# View logs
docker compose logs -f browser-use-api

# Restart
docker compose restart browser-use-api

# Check health
curl -s https://browser-use.rbnk.uk/health | jq .

Configuration¶

Environment variables in .env:

BROWSER_USE_API_KEY=...      # API authentication
LITELLM_API_URL=http://litellm:4000  # LLM proxy
LITELLM_API_KEY=...          # LiteLLM virtual key
DEFAULT_MODEL=gpt-5.1        # Default model
MAX_STEPS=50                 # Max steps limit
TASK_TIMEOUT=300             # Timeout seconds

LiteLLM Virtual Keys¶

Browser-use uses LiteLLM virtual keys for secure AI model access: - Budget: $10/month allocation for browser automation tasks - Models: Access to gpt-5.1, claude-4, and other models via LiteLLM proxy - Cost Tracking: Real-time usage monitoring and budget alerts

For virtual key setup and management, see: LiteLLM Virtual Keys

Architecture¶

FastAPI application with Gradio UI
Playwright for browser automation (Chromium)
LiteLLM proxy for LLM routing
Traefik for SSL and routing
Prometheus metrics for monitoring

Full Documentation¶

See /srv/dockerdata/browser-use/README.md for complete documentation including: - Detailed API reference - Example tasks - MCP integration - Troubleshooting guide - Architecture diagrams