Loop Engineering Profile

Aionis fits Loop Engineering as the memory governance layer, not as the loop executor.

Your Agent host owns the loop:


plan -> act -> validate -> observe -> revise

Aionis owns the memory and governance around that loop:


observe loop evidence -> guide next iteration -> attribute feedback -> measure -> snapshot

What Aionis Adds To A Loop

Loop need	Aionis surface
Carry state across iterations	Execution memory and handoff evidence.
Avoid failed retries	Failed branches become `do_not_use` or inspect-first evidence.
Reuse successful procedure	Stable workflow candidates can become procedure memory.
Measure whether memory helped	`/v1/measure` and effect reports.
Debug incidents	Operator snapshot and Flight Recorder.

Minimal Iteration Shape


await aionis.execution.observeStep({
  agent_id: "worker-1",
  run_id,
  task_signature,
  title: "Iteration 2 verifier result",
  summary: "Unit tests passed but integration verifier failed.",
  outcome: "failed",
  target_files: ["src/checkout/adapter.ts"],
});
 
const guide = await aionis.execution.guideForRole({
  agent_id: "worker-1",
  role: "worker",
  run_id,
  task_signature,
  query_text: "Revise the implementation using the latest verifier result.",
  context_mode: "compact_agent",
});

The host still decides what tools to run. Aionis decides how previous loop evidence may influence the next context.

Promotion Rule

One successful iteration should not become global authority. Treat loop evidence as candidate experience until repeated outcome attribution supports promotion. This keeps Aionis from becoming an automatic rule writer.

Run The Profile


npm run -s runtime:e2e:loop-engineering-profile

Guide: AIONIS_LOOP_ENGINEERING.md .