Review runtime

Review-oriented runtime paths are how Aionis surfaces continuity review material, evolution review material, and replay repair review inside the public runtime.

Review-oriented surfaces

These endpoints are not generic comments or annotations. They package runtime state into review-ready structures so a human or host system can evaluate continuity quality, evolution quality, or replay repair decisions.

Continuity review packEvolution review packEvolution inspectReplay repair reviewGoverned subset

Continuity

Can this task resume safely?

Review the latest handoff, resume state, and recovery contract before trusting a pause/resume path.

/v1/memory/continuity/review-pack

Evolution

Is the runtime learning the right thing?

Review workflow, pattern, and recommendation signals before treating them as trustworthy guidance.

/v1/memory/evolution/review-pack

Replay repair

Should this repaired playbook advance?

Use the Lite governed subset to approve, reject, or shadow-validate repaired playbook versions.

/v1/memory/replay/playbooks/repair/review

trusted continuitycandidate evolutioncontested patterngoverned decisionshadow validateReview surfaces exist to decide whether runtime state is trustworthy enough to reuse.

Public review methods

SDK method	Route	Purpose
memory.reviewPacks.continuity(...)	POST /v1/memory/continuity/review-pack	Build a continuity review pack from handoff/recovery context
memory.reviewPacks.evolution(...)	POST /v1/memory/evolution/review-pack	Build an evolution review pack from kickoff and learning context
memory.replay.playbooks.repairReview(...)	POST /v1/memory/replay/playbooks/repair/review	Approve or reject a replay playbook repair in the Lite governed subset

Reading rule

Read review runtime in one question: what is being trusted or rejected here? Continuity review is about resume trust, evolution review is about learning trust, and replay repair review is about workflow trust. If you keep that decision boundary in mind, the three surfaces stop feeling like unrelated endpoints.

What each surface is for

Continuity review pack

Use this when you want to review whether a task can be resumed safely.

The response centers on:

• latest handoff
• latest resume
• latest terminal run
• recovered handoff payload
• a review contract with target files, next action, acceptance checks, and rollback expectations

Evolution review pack

Use this when you want to review whether the runtime is learning the right workflow and pattern signals from prior execution.

The response centers on:

• selected tool and recommended file path
• stable and promotion-ready workflows
• trusted, contested, and suppressed patterns
• learning summary and recommendations
• evolution_inspect
• policy review and governance contract
• optional governance apply payload or result

Replay repair review

Use this when a playbook repair needs an explicit decision before promotion or further validation.

This is the review gate for replay repair in Lite's governed subset.

Minimal continuity review example

const continuityPack = await aionis.memory.reviewPacks.continuity({
  tenant_id: "default",
  scope: "repair-flow",
  anchor: "task:export-repair",
});

Read these fields first:

1. continuity_review_pack.review_contract
2. continuity_review_pack.latest_handoff
3. continuity_review_pack.latest_resume
4. continuity_review_pack.recovered_handoff

Minimal evolution review example

const evolutionPack = await aionis.memory.reviewPacks.evolution({
  tenant_id: "default",
  scope: "repair-flow",
  query_text: "repair export route serialization bug",
  context: {
    goal: "repair export route serialization bug",
    task_kind: "bugfix",
  },
  tool_candidates: ["read", "edit", "test"],
});

Read these fields first:

1. evolution_review_pack.review_contract
2. evolution_review_pack.learning_summary
3. evolution_review_pack.learning_recommendations
4. evolution_review_pack.evolution_inspect
5. evolution_review_pack.policy_governance_contract

If you are reviewing self-evolving policy state, this is the page where evolution_inspect becomes visible in the public runtime.

Minimal replay repair review example

await aionis.memory.replay.playbooks.repairReview({
  tenant_id: "default",
  scope: "repair-flow",
  actor: "reviewer",
  playbook_id: "repair-export",
  version: 2,
  action: "approve",
  note: "repair looks safe for shadow validation",
  auto_shadow_validate: true,
  shadow_validation_mode: "execute_sandbox",
  target_status_on_approve: "shadow",
});

Why this matters

These surfaces matter because continuity is not only about starting or replaying work. It is also about deciding whether the runtime state is trustworthy enough to reuse. These surfaces matter because continuity includes deciding whether runtime state is trustworthy enough to reuse.

That is what review runtime paths are for:

1. continuity review checks whether pause/resume state is trustworthy
2. evolution review checks whether learning signals are trustworthy
3. replay repair review checks whether repaired playbooks are trustworthy

Trust boundary

Task start, handoff, and replay create continuity artifacts. Review runtime is the layer that decides whether those artifacts deserve reuse. That is why this surface belongs in the public runtime: self-evolving systems still need explicit trust checkpoints.

Review in Lite today

In Lite, these paths are present through:

• public review packs
• replay repair review
• policy and evolution review surfaces

Related docs

1. Memory
2. Policy Memory and Evolution
3. Replay and Playbooks
4. Handoff
5. Automation