CALIBER
Quickstart
Build & integrate

Error handling and retries

The practical integration policy for CALIBER failures: classify them correctly, use waiters where available, and retry only what is actually safe to retry.

DeveloperHow ToGA
PrerequisitesA CALIBER integration using the SDK or REST API
Reviewed 2026-08-10 · current main branch docs contract

Use this page when you need the practical integration policy for failures: which problems are auth or scope issues, which are transient, and when to use waiters instead of building your own polling loop.

At a glance

ConcernRecommended defaultRelated docs
Authentication and project scopevalidate them firstAuth and project scoping
API errorsprefer typed SDK errors where availableSDK API reference
Long-running operationsuse the SDK waiter/polling pathCLI and async client
Raw HTTP diagnosticscapture request ids and envelopesAuthentication and conventions

1. Classify the failure before retrying

Start by separating:

  • auth or permission failures
  • wrong-project or wrong-scope failures
  • request-shape or validation failures
  • transient transport or dependency failures

Retrying the first three classes usually only creates noise.

2. Use waiters for long-running operations

If the product already exposes a waiter or polling contract, use that rather than inventing an uncontrolled sleep loop in your application.

That keeps retry behavior aligned with the documented SDK surface.

3. Retry safely

For automated retries:

  • prefer idempotent reads first
  • be careful with writes that can change external state
  • capture request identifiers and final error context

CALIBER : Contextual Adaptive Lifecycle for Intelligent Build, Evaluation, and Refinement — this page is generated from the authoritative Markdown sources in docs/ and the repository-level ARCHITECTURE.md.