FAQ

Use this page for the recurring questions that come up during integration and operations.

Why does write succeed but recall return weak context?

Most often nodes were missing in write payload or embedding configuration is not set for semantic retrieval.

Which fields should I always log?

request_id, run_id, decision_id, and commit_uri.

What is the minimum useful write payload?

At minimum, send explicit tenant_id, scope, and at least one recallable node in nodes[]. input_text alone is not enough if you expect later retrieval.

Can I use Aionis without policy loop?

Yes. Start with memory APIs, then add policy APIs when execution governance is needed.

Do I need embeddings on day one?

No. You can start with deterministic writes and basic recall flows. Add production embedding configuration when you need stronger semantic retrieval quality.

When should I introduce replay?

As soon as you have a workflow that matters in production. Replay is most useful once you need incident debugging, release evidence, or decision provenance.

Can I keep auth off in development?

Yes for strictly local evaluation, but turn auth on before shared staging or production. Do not let MEMORY_AUTH_MODE=off leak into any environment used by other people or automation.

What is the difference between API Guide and API Reference?

Use API Guide for integration strategy, auth expectations, retries, and operational behavior. Use API Reference when you need endpoint groups, contracts, and request or response details.

Is sandbox production-ready?

Sandbox APIs are experimental. Gate usage with strict policy and isolated execution settings.

Where should I start if production errors spike?

Run core gate, check request-level error patterns, then replay a failing chain end-to-end.

Where should a new team start?

Start with Overview, then Quickstart, then API Guide or SDK Guide depending on your integration style.