Driftline Analytics

Anonymous analytics service for ATProto app views, hosted on Valtown.

• Anonymous by design: users are identified by pseudonymous IDs derived from DIDs
• Per-app-view isolation: same user gets different IDs across different app views
• Simple event model: accounts, views, and actions
• JSON stats API

Base URL: https://driftline.val.run

GET /

POST /collect
Headers: X-API-Key: <your-api-key>
Content-Type: application/json

Body (single event):
{
  "v": 1,
  "appView": "kipclip.com",
  "env": "prod",
  "ts": "2025-01-15T10:30:00.000Z",
  "uid": "a1b2c3d4e5f6",
  "type": "action",
  "name": "checkin_created",
  "screen": "CheckinScreen",
  "props": { "placeType": "cafe" }
}

Body (batch):
{
  "events": [...]
}

Event types:

• account - Track account creation (once per user)
• view - Track screen impressions
• action - Track user actions

All stats endpoints require the X-API-Key header.

GET /stats/:appView?env=prod
GET /stats/:appView/accounts?env=prod
GET /stats/:appView/users?env=prod
GET /stats/:appView/events?env=prod

See @tijs/driftline-client for the TypeScript client library.

User IDs are derived using SHA-256:

uid = sha256(salt + did).slice(0, 12)

• Each app view uses its own salt
• Same DID produces different UIDs across app views
• Server never sees the original DID

Create API keys (requires ADMIN_SECRET env var):

POST /admin/api-keys
Headers: X-Admin-Secret: <admin-secret>
Body: { "appView": "your-app.com" }

deno task fmt      # Format code
deno task lint     # Lint code
deno task check    # Type check
deno task deploy   # Format, lint, check, and push to Valtown