SentinelWebSolver

FastAPI server that extracts OpenAI Sentinel tokens using Camoufox (anti-detection Firefox browser).

How It Works

1. On startup, launches a Context Pool of N Camoufox browser instances, each with a unique randomized fingerprint (OS, user-agent, canvas, WebGL, etc.)
2. Pre-warms pages by navigating to the target URL in advance — when a request arrives, the page is already loaded and ready (~0.8s latency)
3. Resource blocking — images, fonts, media, and analytics requests are aborted to reduce per-page memory (~760 MB idle with 1 browser, down from 2.2 GB without blocking)
4. After serving a request, the pool immediately starts warming a replacement page in the background
5. Round-robin browser rotation ensures consecutive requests produce unique Sentinel tokens

Requirements

• Python 3.11+
• uv

Setup

# Install dependencies
uv sync

# Download Camoufox browser binary
uv run python -m camoufox fetch

Usage

uv run python main.py

Environment Variables

Variable	Default	Description
HOST	0.0.0.0	Server bind address
PORT	8000	Server bind port
POOL_SIZE	1	Number of browser instances in pool
HEADLESS	true	Run browsers in headless mode
MAX_CONCURRENCY	5	Max concurrent token extractions
NAVIGATION_TIMEOUT	30000	Page navigation timeout (ms)
TOKEN_TIMEOUT	30000	SentinelSDK.token() timeout (ms)
PAGE_WAIT	2000	Wait after page load before injection (ms)

Example:

PORT=9000 POOL_SIZE=5 uv run python main.py

API

GET /token

Query Param	Default	Description
timeout	TOKEN_TIMEOUT env value	Override token timeout for this request (ms, 1000-120000)

Response (200):

{
  "p": "...",
  "t": "...",
  "c": "...",
  "id": "uuid"
}

Status	Description
200	Token extracted successfully
502	Browser error
504	Timeout

API docs available at /docs.

License

AGPL-3.0