​

SLI Recipes

​

1. Latency

• Definition: time from request receipt to successful response (p50/p90/p95).
• Implementation: OpenTelemetry spans with http.server.duration, aggregated in Prometheus.
• Dashboard: contrast vs Core Web Vitals (frontend) and API calls.

​

2. Reliability / Error Rate

• Definition: (failed requests / total requests) per service + endpoint.
• Implementation: HTTP status buckets, GraphQL errors, workflow job failures.
• Tip: separate user errors vs platform errors to avoid noisy metrics.

​

3. Freshness / Data Lag

• Definition: time between source event and availability in consumer system.
• Use case: Productivity dashboards fed from automations; Thinki.sh community stats.
• Implementation: event timestamp vs ingest timestamp; alert if > threshold.

​

4. Business Outcomes

• Definition: completion rate of key workflows (productivity ritual, AI agent task, Thinki.sh challenge).
• Implementation: Fire event to Segment/PostHog; compute ratio vs started.
• Note: great for product-led SLOs.

​

5. Experience Metrics

• Frontend: Core Web Vitals (LCP, INP, CLS) via Web-Vitals JS + Analytics.
• Mobile: App start time, error-free sessions (Sentry), offline success rate.

​

6. Storage + Queue Health

• Useful for n8n and automation heavy flows.
• Track backlog size, oldest message age, failure retries.