Sylica

Production Readiness

Use this checklist when promoting from staging to production traffic.

A stable integration requires more than a successful test request. Production quality depends on observability, secret hygiene, bounded retries, and workload-specific routing strategy.

  1. Use server-side secrets and rotate keys on a fixed operational cadence.
  2. Set strict connect/read timeouts per request and enforce global request deadlines.
  3. Implement retry policy only for retryable classes (429/502/504/500), with jitter.
  4. Persist x-sylica-request-id in logs, traces, and support tooling.
  5. Track rate-limit, insufficient-credit, and timeout metrics by workload type.
  6. Validate fallback behavior with game-day tests before launch.

Machine-Readable Contract

For automation and AI crawlers, consume /docs/llms.txt. Use these human-readable pages for operational guidance and edge-case behavior.