Operations

Recovery, verification, and drift should be boring.

Use this path when search is stale, failing, unclear, or needs production hardening. Start with status and failed work, then decide whether to retry, reconcile, backfill, reindex, restore, or upgrade.

Sync modes

Accepted work is not the same thing as visible search.

Accepted work is not the same thing as visible search.

:inline

The caller waits for terminal backend success. Good for local workflows and simple apps.

:oban

Durable enqueue only. Good for production throughput, not for pretending the document is visible already.

:manual

Accepted work for operator-controlled follow-up. Good for migrations, imports, and explicit recovery.

First response path

What to run when someone says “search is wrong.”

Start here

  1. Identify the schema and the relevant environment.
  2. Run mix verify.adopter for a quick verification check.
  3. Use mix scrypath.status and mix scrypath.failed.
  4. Use mix scrypath.reconcile before changing state.

When you need more

If the fast checks pass but the app still behaves differently, use the support guide for readiness truth and the intake guide to submit enough evidence for maintainer classification.

Report an integration issue

Run Meilisearch

Use the operations guide for version pinning, keys, storage, backups, upgrades, and task backlog posture through a Scrypath lens.

Open operations guide

SRE view

The library speaks in telemetry and recovery commands.

Observe

Track search latency, hydration misses, sync error rate, task backlog age, failed tasks, disk, and memory pressure.

Recover

Backfill when the contract is still correct. Reindex when the contract changed or drift is untrusted.

Escalate

Bring version, command output, schema metadata, task evidence, and backend health before widening the issue beyond the supported path.

First response

Start with status, then choose the repair.

Use the support guide for supported versions and verification commands. Use the outside-adopter intake guide for evidence submission and maintainer routing. Use the SRE notes for platform signals such as latency, task backlog, disk pressure, and Meilisearch health.

Scope pressure routes through the scope-and-reopen-policy.md and reviewed outside-adopter intake evidence, not implied feature expansion.