Use Case: Playwright Workers
Problem
Browser-based workers (Playwright, Puppeteer) drive long-lived sessions on behalf of automation tasks. Each worker:
- often holds a logged-in browser session,
- crashes occasionally,
- needs operators to "restart this one", "re-login this one", or "rotate the proxy",
- runs on a laptop, a small VM, or behind NAT — nothing inbound-routable.
Without a control plane, operators SSH around or chase logs.
How Capsule helps
Run each worker (or each pool of workers) as a Capsule Service with an embedded Agent:
- The Agent makes outbound calls to Opstage; no inbound port required.
- Health reflects browser session state (logged in / logged out / crashed).
- Actions:
restartBrowser,relogin,rotateProxy,dumpScreenshot. - Configs report current proxy, current account label (never the password).
Typical Architecture
text
Laptop / small VM:
┌───────────────────────────┐
│ Playwright Worker │
│ ├─ Agent (outbound) │
│ └─ Browser session │
└────────────┬──────────────┘
│ outbound only
▼
Opstage BackendWhat Opstage can show
- Live online/offline status of every worker.
- Health: is the browser session alive and logged in?
- Recent actions and their results (e.g. last
relogin). - Audit of who triggered what.
What Opstage can do
- Restart / re-login / rotate proxy from the console.
- Disable a worker that's misbehaving.
- Compare config across the fleet.
CE scope
- Live online/offline state for every worker, even those running on laptops or behind NAT.
- Health that reflects browser session state (logged in / out / crashed).
- Operator actions:
restartBrowser,relogin,rotateProxy,dumpScreenshot. - Audit of every operator intervention.
Future EE / Cloud enhancements
- Worker-pool views with rollups across many hosts.
- Centralized log/screenshot capture for failed sessions.
- HA Opstage so a control-plane outage never blocks worker recovery.
- Hosted Cloud Opstage for teams running workers on personal devices.