superbizon007/antidetect-browser-agent repository overview

# from apps/antidetect_browser_agent/
docker build \
  --build-context a2a_pkg=../../packages/a2a_agent \
  -f Dockerfile.ubuntu \
  -t antidetect-browser-agent:latest \
  .

docker run -d \
  -p 443:443 \
  -e VLM_URL=https://api.openai.com/v1 \
  -e VLM_API_KEY=sk-... \
  -e VLM_MODEL=gpt-4o \
  antidetect-browser-agent:latest

docker run -d \
  -p 443:443 \
  -p 5901:5901 \
  -p 6901:6901 \
  -e VLM_URL=https://api.openai.com/v1 \
  -e VLM_API_KEY=sk-... \
  -e VLM_MODEL=gpt-4o \
  -e CAMOUFOX_HEADLESS=false \
  antidetect-browser-agent:latest

docker run -d \
  -p 443:443 \
  -e VLM_URL=... \
  -e VLM_API_KEY=... \
  -e CAMOUFOX_PROXY_SERVER=http://proxy.example.com:8080 \
  -e CAMOUFOX_PROXY_USERNAME=user \
  -e CAMOUFOX_PROXY_PASSWORD=pass \
  -e CAMOUFOX_GEOIP=true \
  antidetect-browser-agent:latest

docker run -d --cap-add=NET_ADMIN \
  -p 443:443 \
  -e VLM_URL=... \
  -e VLM_API_KEY=... \
  -e SOCKS_PROXY=socks5://user:[email protected]:1080 \
  antidetect-browser-agent:latest

docker run -d \
  -p 443:443 \
  -e VLM_URL=... \
  -e VLM_API_KEY=... \
  -e API_KEY=my-secret-token \
  antidetect-browser-agent:latest

POST   /                         JSON-RPC 2.0 — tasks/send, tasks/sendSubscribe, etc.
GET    /.well-known/agent.json   AgentCard discovery

curl -sk https://localhost:443/ \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "jsonrpc": "2.0",
    "id": "1",
    "method": "message/send",
    "params": {
      "message": {
        "role": "user",
        "parts": [{"text": "Go to example.com and return the page title"}]
      }
    }
  }'

POST   /v1/chat/completions

curl -sk https://localhost:443/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "model": "antidetect-browser-agent",
    "messages": [
      {"role": "user", "content": "Search for 'camoufox' on Google and return the first result title"}
    ]
  }'

GET    /health

curl -sk https://localhost:443/health
# {"status":"ok","agent":"web.browser.antidetect"}

$TRAJECTORY_DIR/[{context_id}/]{timestamp}/
  turn_NNNN/
    screenshot.png            raw screenshot at step start
    screenshot_marked.png     screenshot with numbered element boxes
    screenshot_annotated.png  screenshot with red circle at click target (click actions only)
    elements.json             detected interactive elements
    llm_0_request.json        VLM request payload
    llm_0_response.json       VLM response and parsed action
    locator.json              locator model request + response (only when LOCATOR_URL is set)
  summary.md                  task plan written at completion

docker run -d \
  -p 443:443 \
  -v /tmp/trajectories:/trajectories \
  -e TRAJECTORY_DIR=/trajectories \
  -e VLM_URL=... \
  -e VLM_API_KEY=... \
  antidetect-browser-agent:latest

services:
  antidetect-agent:
    image: antidetect-browser-agent:latest
    ports:
      - "443:443"
      - "5901:5901"
      - "6901:6901"
    environment:
      VLM_URL: https://api.openai.com/v1
      VLM_API_KEY: sk-...
      VLM_MODEL: gpt-4o
      API_KEY: my-secret-token
      CAMOUFOX_HEADLESS: "false"
      CAMOUFOX_GEOIP: "true"
      CAMOUFOX_PROXY_SERVER: http://proxy:8080
      CAMOUFOX_PROXY_USERNAME: user
      CAMOUFOX_PROXY_PASSWORD: pass
      TRAJECTORY_DIR: /trajectories
      VNC_PW: changeme
    volumes:
      - ./trajectories:/trajectories
    cap_add:
      - NET_ADMIN   # only needed for SOCKS_PROXY transparent tunnel