Skip to content

Scroll and Pattern lifecycle

To change the score is to name another revision. The admitted Run keeps its first seal.

Within Spellweaver, a Scroll—one immutable Pattern revision—is never silently rewritten. Spellweaver selects one revision from an admitted Intent; the Invocation names that casting, and the canonical Run carries its identity. If the registered executable no longer matches, execution fails instead of substituting another workflow. Pattern identity owns the complete law.

Identity before motion

PatternManifest records a renderer-neutral snapshot: schema version; URL-safe key and revision; opaque reviewed implementation revision; checkpoint-schema identifier; declared entry station; semantic stations with key, label, and kind; permitted edges with declared endpoints; and a deterministic SHA-256 digest. The workflow lookup name is persisted separately, although the current Workflow binding requires it to equal the manifest key.

Executable stations bind one-to-one to Python Graph node types, and the declared entry must bind the actual Graph start node. The manifest is authored beside that Python and fingerprints the declared score; it is not a canonical intermediate representation from which all behavior is compiled. The implementation revision (py.1 for both built-ins) records a reviewed compatibility closure: change it or the public Pattern revision when new behavior cannot safely resume old state. It does not automatically detect source edits. Edge parity and return semantics must still be established by source review and tests.

Each semantic station currently behaves like a legacy inline Spell placement: key, label, kind, implementation closure, and edges are bound only inside its Scroll. There is no independent Spell identity or portable catalogue. A future portable Scroll pins exact authority-qualified Spell contract revisions and digests; a receiver-owned Resolution Lock separately binds each placement to an exact local implementation. Admission reports every unknown, unavailable, incompatible, revoked, or unauthorized placement instead of substituting a similar name or newer revision. A future Loom may show the absence only in an inert resolution report; an unresolved placement never enters an executable Graph.

Boot assembly derives an internal legacy_inline resolution for each executable current station; the declarative terminal remains Graph End, not a Spell. This private lock fingerprints the Python node binding so a comparison can detect change while leaving the frozen v2 manifest snapshot and digest untouched. It is not persisted on Run, portable, configurable from TOML, or exposed by Loom/Orb, and therefore does not pretend that the portable Spell catalogue has shipped.

Future configuration follows the closed spine in ADR 28: TOML selects exact registered identities and bounded policy only. Never place import paths, node classes, graph edges, prompts, source bytes, credentials, or live handles in a Scroll selector. Unknown ids and missing exact revisions are refusal, not cues for package scanning or nearest-name fallback.

The catalogue is immutable after construction. It rejects duplicate (key, revision) pairs, requires an explicit active revision when one name has alternatives, requires explicit non-default route precedence, and names its default. Its two current source-defined manifests are bridge_chat@1, the default with checkpoint schema bridge-chat-state-v1, and delegated_rite@1, the exact /delegate command-token route with delegated-rite-state-v1. Construction performs no package scan.

The fixed registry in source 
def builtin_workflow_registry() -> BuiltinWorkflowRegistry:
    """Build the built-in workflow registry (the sole construction site today)."""
    return BuiltinWorkflowRegistry(
        workflows=(BRIDGE_CHAT, DELEGATED_RITE),
        active_revisions=(
            (BRIDGE_CHAT.name, BRIDGE_CHAT.manifest.revision),
            (DELEGATED_RITE.name, DELEGATED_RITE.manifest.revision),
        ),
        route_precedence=(DELEGATED_RITE.name,),
        default_name=BRIDGE_CHAT.name,

Open the owning registry source

Admission happens once

Admission and ownership orders four operations:

  1. Validate and select the Pattern.
  2. Atomically create the canonical Run and initial durable delivery with the exact manifest snapshot.
  3. Retain the caller-owned initiating record while that delivery is held, then release it.
  4. Publish one queue job keyed by Run and enqueue sequence; relay the same key when unavailable.

Retention failure settles only an exact unreleased held generation. A release call that commits and then raises is resolved from the delivery row and remains admitted; compensation cannot overwrite PENDING. Broker failure leaves the admitted Run queued with its exact pending delivery. PostgreSQL and SAQ are not one distributed transaction; startup reconciliation and the process-owned relay probe and republish the idempotent delivery key without fabricating a new Pattern hop.

Execution and resume look up the exact persisted Pattern key and revision rather than routing the Intent again; a retained revision need not remain active for new admission. Before running, the worker validates checksum, workflow-name binding, implementation revision, and equality with that exact registered manifest. A corrupt checksum, drifted declaration or implementation closure, missing workflow, or unavailable revision fails as pinned Pattern unavailable. No default or newer Pattern replaces it.

Checkpoint and suspension mechanics belong to Stasis and return. AgentJob and provider boundaries belong to Delegated agents.

Human-attested material stops at a Gate

Spellweaver does not infer human authorship from a commit name, account, writing style, or a human approval. A future trusted authoring surface records an Authorship Attestation and Protected Region against the exact artifact revision, stable region id, and content digest. Existing material remains unknown until a human attests its current form.

base artifact + protected-region manifest + candidate
                         |
                  target-owner diff
                    /          \
             no overlap       overlap
                 |               |
          ordinary gates    exact live HitL
                                 |
                         target-owner promotion

An Agent may prepare a replacement in its candidate workspace. It cannot change the active region, remove protection by deleting a source marker, or reuse a verdict from another base or patch. The live review names every touched region, old and replacement digests, and an authorized exact-diff artifact. Base drift, candidate drift, a newly touched region, or a changed target effect invalidates the call and requires another verdict.

Approval and authorship remain separate after promotion. A wholly Agent-written replacement stays agent_generated; retained human text interleaved with Agent work becomes mixed. Unchanged human-attested content may keep its origin through an approved move. Only a fresh human authorship attestation can mark a rewritten exact digest human_attested.

Visible comments, front matter, and Loom decoration are projections of the target-owner manifest, not enforcement. A compliant effectful Agent therefore works through Creation's candidate surface; a direct writer with authority over the active checkout bypasses Spellweaver and cannot claim this protection. No attestation store, protected-region resolver, authoring UI, or exact diff review card is delivered. State of Work owns that boundary.

A new score is a new revision

Keep Invocation choice separate from score change. A new prompt, source ArtifactRef, requested language, target profile, or declared finish boundary is another Invocation of the same Scroll only when its typed schema already admits that value. Adding a station, edge, Spell contract, authority ceiling, or required model changes the score and therefore creates a new immutable revision. Finishing one Invocation at an accepted intermediate boundary and later continuing from its artifact is a new forward Invocation, never resurrection of the terminal Run.

The current catalogue can preserve multiple executable revisions per workflow name. Active routing is explicit; a saved Run continues through its exact registered revision even after activation moves forward. The catalogue is still source-built and preserves no historical Python by itself. Removing old code makes that revision unavailable; automatic compatibility proof, durable publication, migration, drain, or refusal remain future work.

Topology-A local runs are Available and pin manifests. Extension activation is Partial, but no Pattern store or public contribution API is delivered.

Future publication follows the Pattern contribution law: inert Scroll and Spell-contract candidates declare identity, continuity, and evidence; executable Spell implementations register through a separate owner. Validation and review may publish an immutable revision, bind an implementation, and activate a new process generation only as distinct effects. An A2A teaching bundle reaches this path only through Assimilation. The current Workflow type remains internal and pre-v1, not a third-party ABI.

Current code or contribution changes require controlled Evolution and Vessel/catalogue replacement; this is not Reanimation. A future declarative-only Scroll may activate without a restart only after all required implementations already exist and an atomic durable catalogue generation mechanism is proved.

A future scheduler would create an Occurrence and submit it through ordinary revision-pinned admission. Its timer never calls a Graph node, model, or container directly. Scheduling and service classes owns the designed temporal contract; no Occurrence service is implemented.