Aionis Docs

Appearance

Error Model ​

Use this page to implement client retry, alerting, and incident triage behavior.

Standard response shape ​

json
{
  "error": "string_code",
  "message": "human_readable_message",
  "details": {}
}

Common classes ​

  1. invalid_request (400)
  2. unauthorized / forbidden (401/403)
  3. not_found (404)
  4. rate_limited_* (429)
  5. backend_capability_unsupported (501)

Error example and fix ​

Example error:

json
{
  "error": "rate_limited_write",
  "message": "too many requests",
  "details": { "retry_after_ms": 800 }
}

Fix:

  1. Retry after the provided delay.
  2. Add exponential backoff + jitter in client retry policy.
  3. Reduce burst concurrency for the affected endpoint group.
  1. Retry transient network failures.
  2. Retry 429 using exponential backoff + jitter.
  3. Persist and display request_id for all failures.
  4. Log error + details to preserve incident triage context.

Do not retry blindly ​

Do not blindly retry 400, 401, 403, or schema errors. These usually indicate caller bugs, missing credentials, or policy violations that require input changes instead of transport retry.