Pawgrab

Web scraping API. Returns clean Markdown, HTML, text, or structured JSON from any URL.

Features

Scraping

• Single URL scraping with multiple output formats
• Async site crawling (BFS, depth/page limits, Redis job queue)
• Structured extraction via OpenAI, CSS selectors, XPath, or regex

Browser

• Auto JS detection - curl_cffi first, Patchright fallback for JS-heavy pages
• Anti-bot evasion - TLS fingerprint impersonation, stealth browser profiles
• Proxy rotation with health checking

Production

• Per-domain and API-level rate limiting
• Robots.txt compliance
• Unified error responses with machine-readable error codes
• SSE heartbeats for reliable streaming through reverse proxies
• Idempotency keys for safe retries on crawl/batch endpoints
• Request ID correlation and response timing headers
• Docker Compose deployment (API + worker + Redis)

Benchmarks

End-to-end scraping, median of 5 runs, books.toscrape.com, lower is better.

Pawgrab extracts article-quality content via Readability while raw BS4/lxml return ALL page text (nav, footer, ads, etc). Pawgrab also adds TLS fingerprint impersonation + anti-bot evasion.

Install

pip install pawgrab
patchright install chromium

Quickstart

# Start Redis (needed for /crawl)
docker run -d -p 6379:6379 redis:7-alpine

# Configure
cp .env.example .env
# Set PAWGRAB_OPENAI_API_KEY if you need /extract

# Run
pawgrab serve

Or with Docker:

cp .env.example .env
docker compose up

API

All endpoints under /v1.

POST /v1/scrape

curl -X POST http://localhost:8000/v1/scrape \
  -H 'Content-Type: application/json' \
  -d '{"url": "https://example.com"}'

Field	Type	Default	Description
url	string	required	URL to scrape
formats	array	["markdown"]	markdown, html, text, json
wait_for_js	bool/null	null	Force JS (true), skip (false), auto (null)
timeout	int	30000	Timeout in ms

POST /v1/crawl

Returns job ID (HTTP 202). Poll with GET /v1/crawl/{job_id}.

curl -X POST http://localhost:8000/v1/crawl \
  -H 'Content-Type: application/json' \
  -d '{"url": "https://example.com", "max_pages": 5}'

Supports Idempotency-Key header for safe retries.

POST /v1/extract

Requires PAWGRAB_OPENAI_API_KEY.

curl -X POST http://localhost:8000/v1/extract \
  -H 'Content-Type: application/json' \
  -d '{"url": "https://example.com", "prompt": "Extract the main heading"}'

POST /v1/search

Searches the web and scrapes each result in parallel.

curl -X POST http://localhost:8000/v1/search \
  -H 'Content-Type: application/json' \
  -d '{"query": "python web scraping"}'

GET /health

Returns ok, degraded, or unhealthy with per-component checks.

curl http://localhost:8000/health

Error Responses

All errors return a consistent JSON shape:

{
  "success": false,
  "error": "Human-readable message",
  "code": "machine_readable_code",
  "details": null,
  "request_id": "a1b2c3d4e5f6"
}

Error codes: validation_error, invalid_api_key, rate_limited, robots_blocked, resource_not_found, timeout, fetch_failed, browser_unavailable, queue_unavailable, llm_unavailable, extraction_failed, search_failed, internal_error.

Response Headers

Every response includes:

Header	Description
X-Request-ID	Unique request identifier for correlation
X-API-Version	API version
X-Response-Time	Request duration (e.g. 42.3ms)
X-RateLimit-Limit	Requests allowed per minute
X-RateLimit-Remaining	Requests remaining in current window

CLI

pawgrab scrape https://example.com
pawgrab scrape https://example.com --format text
pawgrab extract https://example.com --prompt "Extract the main heading"
pawgrab serve --port 8000 --reload

Configuration

All settings via env vars with PAWGRAB_ prefix. See .env.example for the full list.

Key settings:

Variable	Default	Description
PAWGRAB_API_KEY	empty	API key for Bearer auth (empty = no auth)
PAWGRAB_RATE_LIMIT_RPM	60	Per-domain rate limit (requests/min)
PAWGRAB_API_RATE_LIMIT_RPM	600	API-level rate limit per client (requests/min)
PAWGRAB_REDIS_URL	redis://localhost:6379/0	Redis for job queue and idempotency

License

DBaJ-GPL v69.420