Pry Architecture 🟢

 POST /v1/scrape ──▶ API Server (FastAPI)
                        │
                        ▼
                  ┌─────────────┐    miss     ┌──────────────────┐
                  │ Strategy    │────────────▶│ Escalation chain │
                  │ Router      │             │ 1. plain HTTP    │
                  └─────────────┘             │ 2. headless      │
                        │ hit                 │ 3. solver layer  │
                        ▼                     └──────────────────┘
                  Result cache (Redis)              │
                        │                           ▼
                        ▼                     Extraction pipeline
                  Response                (readability → markdown/JSON)

Escalation chain

Pry always tries the cheapest option first:

1. Plain HTTP with tuned headers — fastest, works on most static sites.
2. Headless Chromium (Playwright) — JS rendering, session handling.
3. Solver layer — challenge bypass for protected sites.

Each level reports why it failed; the router records the winning strategy per domain so repeat scrapes skip straight to what works.

Storage

• Redis: result cache + per-domain strategy memory + rate limits.
• Nothing is persisted by default — Pry is a proxy, not an archive. Enable the optional storage backend if you need scrape history.