Manifesto

Appearance

@manifesto-ai/lineage ​

Optional history and restore extension for the runtime.

Overview ​

@manifesto-ai/lineage is an optional protocol extension that adds history, restore, audit, branch/head queries, and sealed records to a Manifesto app definition. It is not part of the base runtime ontology.

Current Contract Note: This page describes the Lineage v5 action handle surface. The package contract is packages/lineage/docs/lineage-SPEC.md.

Use this package when you want:

  • withLineage(createManifesto(...), config).activate()
  • seal-aware action.<name>.submit(...args)
  • lineage LineageSubmissionResult values with world and report
  • head, branch, world, and restore APIs on the activated runtime
  • getWorldSnapshot(worldId) for stored snapshot inspection by history id
  • getLineage() for DAG inspection
  • direct access to @manifesto-ai/lineage/provider for low-level persistence and tooling

Runtime Surface ​

ts
import { createManifesto } from "@manifesto-ai/sdk";
import { createInMemoryLineageStore, withLineage } from "@manifesto-ai/lineage";
import TodoMel from "./domain/todo.mel";
import type { TodoDomain } from "./domain/todo.domain";

const lineage = withLineage(
  createManifesto<TodoDomain>(TodoMel, effects),
  { store: createInMemoryLineageStore() },
).activate();

What This Package Owns ​

  • withLineage() and LineageConfig
  • activated LineageInstance<T>
  • lineage-aware action.<name>.submit(...args) that seals before publication
  • lineage-owned LineageSubmissionResult reports
  • inherited action handle reads such as available(), check(), and preview()
  • restore, getWorld, getWorldSnapshot, getLineage, getLatestHead, getHeads, getBranches, getActiveBranch, switchActiveBranch, createBranch
  • history ownership plus the provider surface

Those inherited legality queries keep the base SDK meaning:

  • availability is checked before dispatchability
  • check() returns the first failing layer, so an unavailable action yields an availability failure and does not evaluate dispatchable

Runtime Meaning ​

action.<name>.submit(...args) on a lineage runtime is not the base SDK write path.

It means:

  1. admit the typed action submission
  2. execute the intent through Host/Core
  3. seal the resulting terminal snapshot into lineage
  4. publish only the snapshot that becomes the new visible head

If seal commit fails, submit rejects with SubmissionFailedError and the new snapshot does not become visible.

Settled lineage results include before, after, world, outcome, and an optional report. Failed domain outcomes can still settle and seal a failed world; settlement failures reject instead of fabricating a lineage result.

Relationship to SDK ​

text
@manifesto-ai/sdk
  -> createManifesto(...)
  -> withLineage(...)
  -> activate()

SDK owns the base activation boundary. Lineage owns history once that boundary is decorated.

On a lineage runtime:

  • snapshot() is the app-facing runtime read
  • inspect.canonicalSnapshot() is the current visible full internal snapshot
  • getWorldSnapshot(worldId) is the stored sealed full snapshot
  • base dispatch verbs and historical lineage commit verbs are no longer part of the runtime surface

Lineage-Backed Inspection ​

Use getWorldSnapshot(worldId) when a tool needs to inspect a stored world without changing the live visible runtime. Pair that stored full snapshot with @manifesto-ai/sdk/extensions for app-facing projection, explanation, and read-only simulation.

Replay tooling should also read the seal attempt's computeEnvelope. The envelope stores the exact submitted intent and full Core Context used for the transition. It is replay/accountability metadata and does not participate in world identity hashes.

Use restore(worldId) only when the product is intentionally resuming the visible runtime from that world.

The full Studio/tooling recipe is in Runtime Tooling Surface.