@noy-db/hub 0.3.0-pre.10 → 0.3.0-pre.12

package/dist/kernel/collection-config.d.ts

@@ -0,0 +1,789 @@
+/**
+ * Collection construction config — the PURE half of the `Collection`
+ * constructor.
+ *
+ * `resolveCollectionConfig` takes the raw `CollectionOpts` the Vault threads in
+ * and resolves every `?? default`, every normalized/derived field, and runs the
+ * pure construction-time validations (embeddings-on-CRDT, money-field paths,
+ * deterministic-risk acknowledgement). It holds **no `this`** and performs no
+ * external side effect — the only objects it allocates (`Set`s, `VectorSet`,
+ * the CEK `Lru`) are the same instances the collection then owns by reference.
+ *
+ * The `this`-dependent construction (the search-index store whose persisted
+ * callback closes over `this.searchContext()`, the `RecordCodec`, and the
+ * SyncEngine conflict-resolver registration) stays in the constructor — those
+ * genuinely capture the live instance (and the conflict-resolver closures close
+ * over `conflictPolicy: ConflictPolicy<T>`, whose invariant-in-T custom-merge
+ * signature can't be exposed via a method without breaking `Collection<T>`
+ * assignability to `Collection<unknown>`).
+ */
+import type { NoydbStore, ConflictPolicy, CollectionConflictResolver, HistoryConfig, TierMode, CrossTierAccessEvent, VdigFieldPolicy } from './types.js';
+import type { EnclaveKey } from './enclave/index.js';
+import type { UnlockedKeyring } from '../with-party/team/keyring.js';
+import type { NoydbEventEmitter } from './events.js';
+import type { WriteQueueTracker } from './write-queue.js';
+import type { WriteHookRegistry } from '../port/with/write-hooks.js';
+import type { ServiceBus } from '../port/with/service-bus.js';
+import type { SchemaUpdateGate } from '../with-shape/schema-update/gate.js';
+import type { SchemaFenceController } from '../with-shape/schema-update/fence-controller.js';
+import type { StandardSchemaV1 } from './schema.js';
+import type { LedgerStore } from '../with-commit/history/ledger/index.js';
+import type { CrdtMode } from '../with-commit/crdt/crdt.js';
+import type { CrdtStrategy } from './types.js';
+import { type HistoryStrategy } from '../with-commit/history/strategy.js';
+import { type I18nStrategy } from '../port/with/i18n-strategy.js';
+import { type SyncStrategy } from '../with-party/team/sync-strategy.js';
+import { type BlobStrategy } from '../port/with/blob-strategy.js';
+import { type AggregateStrategy } from '../with-lookup/aggregate/strategy.js';
+import { type TiersStrategy } from '../with-audit/tiers/strategy.js';
+import { type SearchStrategy } from '../with-lookup/search/strategy.js';
+import type { ObjectProjection } from '../with-shape/blobs/object-projection.js';
+import type { BlobFieldsConfig } from '../with-shape/blobs/blob-compaction.js';
+import type { IndexStrategy } from '../with-lookup/indexing/strategy.js';
+import type { IndexDef } from '../with-lookup/indexing/eager-indexes.js';
+import type { I18nTextDescriptor, DictKeyDescriptor, StaticDictDescriptor, DictionaryHandle } from '../port/with/i18n-strategy.js';
+import type { LookupDescriptor, MaterializedBacking } from '../port/with/lookup-strategy.js';
+import type { ComputedFields, ComputedFn, ComputedFieldEntry } from '../with-formula/computed/index.js';
+import type { RollupDeleteIntent } from './via/dispatch.js';
+import '../port/with/computed-strategy.js';
+import { type ClassifiedEntry, type ResolvedClassified, type ClassifiedGuardCtx, type ClassifiedStrategy, type ClassifiedViaConfig } from '../port/with/classified-strategy.js';
+import type { FieldRef, Grain } from './via/graph.js';
+import type { FieldMeta } from '../with-shape/introspection/field-meta.js';
+import type { CollectionMeta } from '../with-shape/introspection/meta.js';
+import type { RefDescriptor } from './refs.js';
+import type { JoinableSource } from './query/index.js';
+import { VectorSet, type EmbeddingDescriptor } from '../with-lookup/embeddings/index.js';
+import { Lru } from './cache/index.js';
+import type { ReadOnlyVaultFacade } from '../with-audit/guards/types.js';
+import type { DerivationRegistry } from '../with-formula/derivations/registry.js';
+import type { TxContext } from '../with-commit/tx/transaction.js';
+import type { MaterializedViewRegistry } from '../with-formula/materialized-views/registry.js';
+import type { MVQueryContext } from '../with-formula/materialized-views/types.js';
+import type { Collection, OnDirtyCallback, CacheOptions } from './collection.js';
+import type { LazyStrategy } from '../port/with/lazy-strategy.js';
+import { ViaPipeline } from './via/pipeline.js';
+import { type ViaBinding, type ViaDescriptor } from './via/index.js';
+import { type ViaFieldSpec } from './via/compose.js';
+/**
+ * Raw options handed to the {@link Collection} constructor by the Vault.
+ * One named type shared by the constructor and the pure resolver.
+ */
+export interface CollectionOpts<T> {
+    adapter: NoydbStore;
+    vault: string;
+    name: string;
+    keyring: UnlockedKeyring;
+    encrypted: boolean;
+    /**
+     * Opt-in: keep the working set encrypted in RAM, decrypting on read (future phase).
+     * Default false — the working set is plaintext.
+     */
+    ramCiphertext?: boolean;
+    emitter: NoydbEventEmitter;
+    /**
+     * Vault-level in-flight write tracker. When present,
+     * `put`/`delete` run inside `writeQueue.track()` so `hub.writeQueue`
+     * reflects outstanding writes. Optional so direct Collection
+     * construction in tests still works untracked.
+     */
+    writeQueue?: WriteQueueTracker | undefined;
+    /** Per-collection schema-update gate; `put`/`delete` await it. */
+    schemaUpdateGate?: SchemaUpdateGate | undefined;
+    /** Vault-level fence controller; `put`/`delete` consult it. */
+    schemaFence?: SchemaFenceController | undefined;
+    /** Hub-level write-hook registry; fired around put/delete. */
+    writeHooks?: WriteHookRegistry | undefined;
+    /** The observe bus, threaded from Noydb. */
+    subsystemBus?: ServiceBus | undefined;
+    /** Active transaction id supplier (null outside a transaction). */
+    activeTxId?: (() => string | null) | undefined;
+    getDEK: (collectionName: string) => Promise<EnclaveKey>;
+    historyConfig?: HistoryConfig | undefined;
+    /**
+     * When `true`, the caller explicitly provided `historyConfig` rather than
+     * inheriting the vault-wide default. Used by `getConfig()` to decide
+     * whether to surface `history: true` in the schema dump.
+     */
+    historyConfigExplicit?: boolean | undefined;
+    onDirty?: OnDirtyCallback | undefined;
+    /**
+     * #693: live check for whether multi-tab coordination is active at all
+     * (`Noydb._tabCoordinationActive`). When true, the shared local store may be
+     * written out-of-band by another tab, so the #606 marker-id-set gate falls
+     * back to an unconditional store read on re-create.
+     */
+    tabCoordinated?: (() => boolean) | undefined;
+    /**
+     * tree-shake seam. When omitted, `collection.blob(id)` throws
+     * with a pointer at the `@noy-db/hub/blobs` subpath. When set (via
+     * `createNoydb({ blobStrategy: blobs() })`), blob storage is live.
+     * `@internal` by virtue of `BlobStrategy` being `@internal`.
+     */
+    blobStrategy?: BlobStrategy | undefined;
+    objectStore?: ObjectProjection | undefined;
+    blobFields?: BlobFieldsConfig | undefined;
+    aggregateStrategy?: AggregateStrategy | undefined;
+    crdtStrategy?: CrdtStrategy | undefined;
+    /**
+     * tree-shake seam — strategy for optional history/ledger/
+     * time-machine. When omitted, history snapshots and ledger appends
+     * become silent no-ops (data still writes); the read APIs
+     * (`history`, `getVersion`, `revert`, `diff`, `clearHistory`,
+     * `pruneRecordHistory`) throw with a pointer at `@noy-db/hub/history`.
+     */
+    historyStrategy?: HistoryStrategy | undefined;
+    i18nStrategy?: I18nStrategy | undefined;
+    syncStrategy?: SyncStrategy | undefined;
+    /**
+     * tree-shake seam. When omitted, indexing is off for this
+     * collection — every `.lazyQuery()` call throws, `.rebuildIndexes()`
+     * is a no-op, and `indexes: [...]` declarations are ignored. Enable
+     * by passing `withIndexing()` from `@noy-db/hub/indexing` at
+     * `createNoydb` time.
+     */
+    indexStrategy?: IndexStrategy | undefined;
+    indexes?: IndexDef[] | undefined;
+    /**
+     * Auto-reconcile behavior for persisted-index drift on lazy-mode
+     * collections. Defaults to `'off'` — operators call
+     * `collection.reconcileIndex(field)` explicitly.
+     *
+     *   - `'off'` (default): no implicit work. Same semantics as.
+     *   - `'dry-run'`: on first lazy-mode query, run
+     *     `reconcileIndex(field, { dryRun: true })` per declared field
+     *     and emit `index:reconciled` with the diff. Nothing is written.
+     *   - `'auto'`: same walk as `'dry-run'` but with `dryRun: false`.
+     *     Drift is repaired in-place and the fix count surfaces on the
+     *     event.
+     *
+     * Unattended long-lived processes (Workers, Node services with no
+     * human operator) should set `'auto'`. Attended desktop apps should
+     * leave it `'off'` and surface a manual "rebuild indexes" button.
+     */
+    reconcileOnOpen?: 'off' | 'dry-run' | 'auto';
+    /**
+     * Hydration mode. `'eager'` (default) loads everything into memory on
+     * first access — matches behavior exactly. `'lazy'` defers loads
+     * to per-id `get()` calls and bounds memory via the `cache` option.
+     */
+    prefetch?: boolean;
+    /**
+     * LRU cache options. Only meaningful when `prefetch: false`. At least
+     * one of `maxRecords` or `maxBytes` must be set in lazy mode — an
+     * unbounded lazy cache defeats the purpose.
+     */
+    cache?: CacheOptions | undefined;
+    /**
+     * Lazy service seam (#267) — supplies the bounded-LRU working-set cache
+     * when `prefetch: false`. Omitted → the deprecated IMPLICIT_LAZY
+     * back-compat default (identical behavior + one-time warn).
+     */
+    lazyStrategy?: LazyStrategy | undefined;
+    /**
+     * Optional Standard Schema v1 validator (Zod, Valibot, ArkType,
+     * Effect Schema, etc.). When set, every `put()` is validated before
+     * encryption and every read is validated after decryption. See the
+     * `schema` field docstring for the error semantics.
+     */
+    schema?: StandardSchemaV1<unknown, T> | undefined;
+    /** Declares this collection a satellite of `satelliteOf` (spec #591). */
+    satelliteOf?: string | undefined;
+    /** Satellite routing table — the fields owned by this satellite (required with satelliteOf). */
+    fields?: readonly string[] | undefined;
+    /** Registers the full-record joined handle under this name (optional; see vault.joined()). */
+    joined?: string | undefined;
+    /**
+     * Optional reference to the compartment's hash-chained ledger.
+     * When present, successful mutations append a ledger entry via
+     * `LedgerStore.append()`. Constructed at the Vault level and
+     * threaded through — see the Vault.collection() source for
+     * the wiring.
+     */
+    ledger?: LedgerStore | undefined;
+    /**
+     * Optional back-reference to the owning compartment's ref
+     * enforcer`).
+     * Collection.put calls `enforceRefsOnPut` before the adapter
+     * write; Collection.delete calls `enforceRefsOnDelete` before
+     * its own adapter delete. See the `refEnforcer` field docstring
+     * for the full protocol.
+     */
+    refEnforcer?: {
+        enforceRefsOnPut(collectionName: string, record: unknown): Promise<void>;
+        enforceRefsOnDelete(collectionName: string, id: string): Promise<void>;
+    } | undefined;
+    /**
+     * Optional back-reference to the owning compartment's join
+     * resolver. When present, `query()` builds a
+     * `JoinContext` so `.join(field)` can resolve through the
+     * existing `ref()` declaration into the target collection.
+     * Absent in tests that construct a Collection directly without
+     * a vault; production usage always has one because
+     * Vault.collection() passes `this` through.
+     */
+    joinResolver?: {
+        resolveSource(collectionName: string): JoinableSource | null;
+        resolveRef(leftCollection: string, field: string): RefDescriptor | null;
+        resolveDictSource?: (leftCollection: string, field: string) => JoinableSource | null;
+    } | undefined;
+    /** — i18nText field descriptors for locale-aware reads. */
+    i18nFields?: Record<string, I18nTextDescriptor> | undefined;
+    /** — embedding config for write-time vector derivation + semantic retrieval. */
+    embeddings?: EmbeddingDescriptor | undefined;
+    /** — string fields exposed to client-side `retrieve()`. */
+    textIndexes?: readonly string[] | undefined;
+    /** — pre-build the lexical index on open (eager-only). */
+    warmIndexOnOpen?: boolean | undefined;
+    /** — persist the lexical index as an opaque encrypted blob at `_ftindex/<name>`. */
+    textIndexPersist?: boolean | undefined;
+    /** — dictKey field descriptors for label resolution on reads. */
+    dictKeyFields?: Record<string, DictKeyDescriptor | StaticDictDescriptor> | undefined;
+    /** — consumer-neutral per-field descriptors. Read via getFieldMeta(). */
+    fieldMeta?: Record<string, FieldMeta> | undefined;
+    /** — collection-level descriptive metadata. Read via getMeta(). */
+    meta?: CollectionMeta | undefined;
+    moneyFields?: Record<string, ViaDescriptor> | undefined;
+    /** — declare via() composed fields; grouped by `_viaBrand` and merged with the money/i18n sugar keys above (a field in both throws). */
+    viaFields?: Record<string, ViaFieldSpec> | undefined;
+    /** — outbound ref declarations (snapshot from vault refRegistry). Used by describe(). */
+    declaredRefs?: Record<string, RefDescriptor> | undefined;
+    computed?: ComputedFields | undefined;
+    /** — declare classified() sensitive-field descriptors (sealed + riders + projections). */
+    classifiedFields?: Record<string, ClassifiedEntry> | undefined;
+    /**
+     * The forget-cascade subject key for this collection (from
+     * `withForgetCascade({ subjects })`), plumbed by the Vault. Consumed by the
+     * classified refusal matrix (R4: a digest-only field cannot be the subject key).
+     */
+    subjectKeyField?: string | undefined;
+    /** — tree-shake seam for `collection.reveal()`. Defaults to `NO_CLASSIFIED`. */
+    classifiedStrategy?: ClassifiedStrategy | undefined;
+    /**
+     * async callback that resolves a dict key to its label
+     * for a given locale. Provided by the Vault.
+     */
+    dictLabelResolver?: ((dictName: string, key: string, locale: string, fallback?: string | readonly string[]) => Promise<string | undefined>) | undefined;
+    /**
+     * Async callback to open a dynamic dictionary handle.
+     * Provided by the Vault for dynamic-dict label-map resolution in
+     * the search index. Static dicts bypass this.
+     */
+    getDictionary?: ((name: string) => Promise<DictionaryHandle>) | undefined;
+    /**
+     * synchronous callback that validates i18nText fields
+     * on put. Provided by the Vault. Throws MissingTranslationError.
+     */
+    i18nPutValidator?: ((record: unknown) => void) | undefined;
+    /** — declare lookup()/enumOf()/dict() fields (#650 Task 2 sugar key, mirrors dictKeyFields), merged with `via(lookup(...))` entries. */
+    lookupFields?: Record<string, LookupDescriptor> | undefined;
+    /**
+     * async label resolver for the `'lookup'` binding's `'static'`/`'reserved'`
+     * tiers. Provided by the Vault — the SAME closure as `dictLabelResolver`
+     * (static table first, else the `vault.dictionary()` handle), so a
+     * native `dict()`/`lookup(static)` field resolves through identical
+     * label data as its `dictKey()`/`staticDict()` alias.
+     */
+    lookupLabelResolver?: ((dimension: string, key: string, locale: string, fallback?: string | readonly string[]) => Promise<string | undefined>) | undefined;
+    /** — the matrix (`backing:'collection'`) tier's present-time row accessor, keyed by the full
+     *  descriptor so it can resolve by `descriptor.key`, not the backing row's PUT-id (#651 Task 3).
+     *  Provided by the Vault. */
+    getLookupBacking?: ((descriptor: LookupDescriptor) => (key: string) => Promise<Record<string, unknown> | undefined>) | undefined;
+    /** — closed-vocabulary write-time membership test (#650 Task 3). Provided by the Vault. */
+    membership?: ((field: string, key: string) => boolean | Promise<boolean>) | undefined;
+    /** — altKeys `ingest` normalization source (#650 Task 3). Provided by the Vault. */
+    getAltIndex?: ((desc: LookupDescriptor) => MaterializedBacking | undefined) | undefined;
+    /**
+     * — sync materialized `key -> row` rows for a lookup descriptor (#650
+     * Task 6, spec §5's snapshot+locale seam; matrix-tier routing added Task
+     * 7). Provided by the Vault; feeds the `'lookup'` binding's
+     * `compareForOrder`/`resolveOrderLabel` and this collection's
+     * `presentForJoin` hook (`resolveCollectionConfig` builds the latter via
+     * `buildPresentForJoin`, below).
+     */
+    snapshotFor?: ((descriptor: LookupDescriptor) => ReadonlyMap<string, Record<string, unknown>> | undefined) | undefined;
+    /**
+     * translator callback from Noydb. When present, missing
+     * translations for `autoTranslate: true` i18nText fields are generated
+     * before the i18n validator runs.
+     */
+    autoTranslateHook?: ((text: string, from: string, to: string, field: string, collection: string) => Promise<string>) | undefined;
+    /**
+     * vault-default locale, inherited from
+     * `openVault(name, { locale })` or `vault.setLocale()`.
+     */
+    defaultLocale?: string | undefined;
+    /**
+     * collection-level conflict resolution policy.
+     * Overrides the db-level `conflict` option for this collection only.
+     */
+    conflictPolicy?: ConflictPolicy<T> | undefined;
+    /**
+     * callback to register an envelope-level resolver with the
+     * SyncEngine. Provided by the Vault (wired from the SyncEngine).
+     */
+    onRegisterConflictResolver?: ((name: string, resolver: CollectionConflictResolver) => void) | undefined;
+    /**
+     * CRDT mode for this collection. When set, `put()` stores
+     * CRDT state in the envelope and `get()` returns the resolved snapshot.
+     * `getRaw(id)` returns the full CRDT state for merge operations.
+     */
+    crdt?: CrdtMode | undefined;
+    /**
+     * optional remote/sync adapter. When present, `presence()`
+     * writes heartbeats to this adapter so other devices can read them.
+     * If the adapter implements pub/sub, presence updates are real-time.
+     */
+    syncAdapter?: NoydbStore | undefined;
+    /**
+     * called by the collection after every successful
+     * `get` / `put` / `delete`. The Vault installs a callback that
+     * appends a consent-audit entry when `withConsent` is active;
+     * outside a consent scope the callback is a no-op. Awaited so a
+     * thrown audit write surfaces to the caller.
+     */
+    onAccess?: (op: 'get' | 'put' | 'delete' | 'reveal' | 'verify' | 'find', id: string) => Promise<void>;
+    /**
+     * invoked by `put`/`delete` before any adapter
+     * write. Receives the prior envelope timestamp + decrypted
+     * record (or `null` if no prior) and the incoming record (or
+     * `null` for delete). Throws `PeriodClosedError` to abort.
+     */
+    /**
+     * opt-in deterministic-encryption index.
+     *
+     * Field names listed here get a deterministic AES-GCM ciphertext
+     * attached to every envelope's `_det` map, which enables blind
+     * equality search via `collection.findByDet(field, value)`.
+     *
+     * **Leaks equality.** Two records with the same value in a
+     * deterministic field produce identical ciphertexts, so anyone
+     * with store access can tell which records share a value without
+     * learning the value itself. This is the textbook trade-off of
+     * deterministic encryption — strictly opt-in for that reason.
+     *
+     * Declaring any field here without also passing
+     * `acknowledgeDeterministicRisk: true` throws at construction,
+     * so the risk must be explicitly acknowledged.
+     */
+    deterministicFields?: readonly string[] | undefined;
+    /**
+     * gate for `deterministicFields`. Must be `true` when
+     * any deterministic field is declared. Any other value throws.
+     */
+    acknowledgeDeterministicRisk?: boolean | undefined;
+    /**
+     * gate for the classified `equatable` knob (R8 double door). Must be `true`
+     * when any classified field declares `equatable: true`; otherwise construction
+     * throws. One-directional — setting it with zero equatable members is a
+     * silent no-op (mirrors `acknowledgeDeterministicRisk`).
+     */
+    acknowledgeEquatableRisk?: boolean | undefined;
+    /**
+     * Structural group-encryption. Fields listed here are
+     * encrypted into their own `_sealed[field]` envelope slot — each under
+     * an HKDF-derived per-field key — instead of sitting inside the open
+     * `_data` blob. Default-off: with no `sensitive` fields the envelope is
+     * byte-identical to today. Read merges them back inline (the
+     * `Sealed<V>`/`reveal()` access restriction is a separate follow-up).
+     *
+     * **Incompatible with `perRecordKeys`/forget-cascade:** sealed
+     * field keys derive off the *collection* DEK, not the per-record CEK, so
+     * crypto-shredding a record does not erase its sealed fields.
+     */
+    sensitive?: readonly string[] | undefined;
+    /**
+     * Per-record content-encryption keys. When `true`, every record body
+     * (and every history version of it) is encrypted under a fresh
+     * per-record CEK, AES-KW-wrapped under the collection DEK and stored
+     * on the envelope's `_cek`. Off by default. Foundation for per-record
+     * erasure and record-scoped sealing. `_det` slots stay
+     * keyed to the collection DEK regardless.
+     */
+    perRecordKeys?: boolean | undefined;
+    /**
+     * Per-record provenance tracking. When `true`, `put()` calls that
+     * supply a `source` option stamp `_source` (opaque source id) and
+     * `_sourceTs` (ISO-8601 timestamp) onto the unencrypted envelope
+     * metadata. Off by default — zero cost for collections that don't
+     * need lineage tracking.
+     */
+    provenance?: boolean | undefined;
+    /**
+     * declared tiers this collection supports. An
+     * undefined or empty list disables the hierarchical-tier surface
+     * on this collection (`putAtTier`, `getAtTier`, `elevate`, `demote`
+     * throw). Tier 0 is implicit and always available.
+     */
+    tiers?: readonly number[] | undefined;
+    /**
+     * tree-shake seam — strategy for the collection-level tier operations
+     * (`putAtTier`/`getAtTier`/`listAtTier`/`elevate`/`demote`). When omitted,
+     * every tier operation throws `TiersNotEnabledError`. Enable by passing
+     * `withTiers()` from `@noy-db/hub/tiers` at `createNoydb` time.
+     */
+    tiersStrategy?: TiersStrategy | undefined;
+    /**
+     * Search / retrieval capability strategy. When omitted, a collection's
+     * `search`/`retrieve`/`similarTo`/`warmIndex`/`flushIndex` methods and the
+     * embedding write-hook throw `SearchNotEnabledError`. Enable by passing
+     * `withSearch()` from `@noy-db/hub` at `createNoydb` time.
+     */
+    searchStrategy?: SearchStrategy | undefined;
+    /**
+     * what a lower-tier caller sees for above-tier
+     * records. Default `'invisibility'`.
+     */
+    tierMode?: TierMode | undefined;
+    /**
+     * optional callback fired on every cross-tier access.
+     * Provided by the Vault; collects notification events and writes
+     * to the ledger.
+     */
+    onCrossTierAccess?: ((event: CrossTierAccessEvent) => void) | undefined;
+    /**
+    /**
+     * Optional back-reference to the owning vault's derivation
+     * registry + collection accessor. When present, successful
+     * `put()` dispatches registered derivation strategies for the
+     * source collection.
+     */
+    derivationSource?: {
+        registry(): DerivationRegistry;
+        getCollection(name: string): Collection<Record<string, unknown>>;
+        /**
+         * Read-only vault facade handed to `derive(source, ctx)` so a
+         * derivation can fetch sibling records. Same shape and
+         * instance the guards service uses for `check(incoming, ctx)`.
+         */
+        getReadOnlyFacade(): ReadOnlyVaultFacade;
+        /**
+         * Read access to the owning Noydb's currently-active multi-record
+         * transaction context, or `null` when no transaction is running.
+         * `dispatchDerivations` consults this so a recursive derived-output
+         * write can register its pre-write envelope onto `ctx._executed`
+         * and roll back alongside the source op on mid-batch failure.
+         */
+        getActiveTxContext(): TxContext | null;
+        /**
+         * Construct a transient TxContext bound to the owning Noydb. Used
+         * by `Collection.putManyAtomic` to publish an active context for
+         * its Phase 2 loop.
+         */
+        createTxContext(): TxContext;
+        /** Publish a TxContext for the duration of a bulk-atomic loop. */
+        setActiveTxContext(ctx: TxContext): void;
+        /** Drop a previously-published TxContext. */
+        clearActiveTxContext(ctx: TxContext): void;
+    } | undefined;
+    /**
+     * Vault-internal hook for materialized-view dispatch.
+     * Parallel to `derivationSource`. When set, `Collection.put` fires
+     * registered MV `onSourceWrite` after the standard derivation
+     * dispatch.
+     */
+    materializedViewSource?: {
+        registry(): MaterializedViewRegistry;
+        getCollection(name: string): Collection<any>;
+        getActiveTxContext(): TxContext | null;
+        getQueryContext(): MVQueryContext;
+    } | undefined;
+    /**
+     * #638 Task 4 — thin collector hook wired by the Vault (`Vault._collectGraphTouch`). Every
+     * `sync-apply`/`cutover`/`restore` mutation origin calls `collect(this.name, id)`; a no-op
+     * when no batch is open (`_beginGraphBatch()` wasn't called) or the vault has no graph.
+     * `collectDelete` (#640) is the sync-apply delete socket — the resolved rollup-parent intents
+     * for a deleted record, captured pre-invalidation.
+     */
+    graphDispatch?: {
+        collect(collection: string, id: string): void;
+        collectDelete(collection: string, id: string, intents: readonly RollupDeleteIntent[]): void;
+    } | undefined;
+}
+/** One normalized `ComputedFields` entry — a plain `ComputedFn` or a `ComputedFieldEntry`
+ *  reduced to its parts, `mode` always defaulted (#638 Task 7). Exported so
+ *  `via/graph-wiring.ts`'s reconcile-path guard can check `mode` without duplicating
+ *  the plain-fn-vs-object-form normalization. */
+export interface ComputedEntryParts {
+    readonly fn: ComputedFn;
+    readonly deps?: readonly string[];
+    readonly mode: 'materialized' | 'virtual';
+}
+export declare function computedEntryParts(entry: ComputedFn | ComputedFieldEntry): ComputedEntryParts;
+/**
+ * Compile a collection's declared config into the ordered list of `ViaBinding`s
+ * for its `ViaPipeline`.
+ *
+ * money then i18n then lookup then classified then blob then computed — order
+ * pinned for pipeline parity with the hand-wired baseline this replaces: money
+ * encode ran before the i18n write stages, and money decode ran before i18n
+ * locale/dict-label resolution on read. Lookup compiles right after i18n
+ * (#650 Task 2) — a separate binding for native lookup()/enumOf()/dict()
+ * fields (`dictKeyFields`/`dictKey()`/`staticDict()` stay on the i18n
+ * binding above, unchanged); its `present` hook only adds `<field>Label`,
+ * same shape of read-time addition as i18n's own dict-label dressing, and
+ * declares no write-pipeline hooks yet (Task 3 adds `ingest`/`enforceWrite`).
+ * Classified compiles after those (#629 Task 6): its
+ * `encodeAtRest`/`decodeAtRest` hooks make the pipeline's `hasAtRestHooks`
+ * true, retiring the codec's inline `sensitiveFields` seal path
+ * (record-codec.ts) for any collection that declares `classifiedFields` —
+ * `classifiedGuardCtx` is the SAME `ClassifiedGuardCtx`
+ * `resolveCollectionConfig` already built for door 1's
+ * `guardClassifiedCompat` call, threaded in by its one caller below. Blob
+ * compiles next (#629 Task 7) but its position is inert: the blob binding
+ * declares NO write/read pipeline hooks (blob content is out-of-band
+ * `BlobSet` side-collections — it must never flip `hasAtRestHooks`), only
+ * `erase`/`describeFragment`. Computed compiles LAST (#638 Task 7) — its
+ * `present` hook must run AFTER money's own `present` so a virtual field's
+ * `deps` read the decoded (quantized) amount, not the raw stored form; at
+ * present-time i18n/lookup's DRESSING now runs AFTER computed instead (#665
+ * `_presentOrder`, `via/pipeline.ts` — compile order here is unchanged).
+ * `via/graph-wiring.ts#applyTaintOverlay` appends the `taint`
+ * binding after WHATEVER this function returns, so taint's present-time
+ * redaction always runs after computed's regardless of this ordering.
+ * {@link Collection._applyMoneyFields} PREPENDS money for the same reason
+ * on its own (MV-precreation reconcile) path — see its docstring;
+ * {@link Collection._applyClassifiedFields} APPENDS classified on that same
+ * reconcile path (blobFields/computed have no late-attach reconcile door —
+ * `viaFields`, like i18nFields/dictKeyFields, is construction-only).
+ *
+ * `eraseCfgOut` (#629 Task 10, optional out-param) — this function runs
+ * before the owning `Collection` exists (`this.codec` isn't built yet), so
+ * the classified binding's `classifySealedShred` closure can't be wired
+ * here. When supplied, `eraseCfgOut.classified` is set to the SAME cfg
+ * object instance handed to `viaBinder('classified')`, so a caller
+ * (`resolveCollectionConfig`) can thread it out to the `Collection`
+ * constructor, which mutates `classifySealedShred` in place once
+ * `this.codec` exists. Additive — every existing caller omits it and keeps
+ * getting a plain `ViaBinding[]`.
+ */
+export declare function compileViaBindings<T>(opts: CollectionOpts<T>, classifiedGuardCtx: ClassifiedGuardCtx, eraseCfgOut?: {
+    classified?: ClassifiedViaConfig;
+}): ViaBinding[];
+/**
+ * The money∩virtual-mode-computed field-name intersection (#669) — shared by
+ * `compileViaBindings` above (fresh construction, `isVirtual` closes over the just-built
+ * `virtualFields` Map) and {@link Collection._applyMoneyFields} (late-attach, `isVirtual`
+ * closes over the already-compiled pipeline's `computed` binding instead — virtual-mode
+ * computed fields have no late-attach door of their own (declaring one on a reconcile call
+ * throws), so by the time money reconciles onto an existing collection, any virtual field
+ * is already sitting in `bindings`, never in `this.computed`: `resolveCollectionConfig`
+ * deliberately excludes virtual-mode entries from the map it assigns to `this.computed`,
+ * since that field feeds ONLY the write-time `evalComputedFields` path — see
+ * `materializedComputed` above). Both callers get the identical intersection semantics from
+ * one function, so money's `presentLate` MAJOR-UNITS dressing (`via/money/binding.ts`) can
+ * never silently disagree between the fresh and late-attach paths.
+ */
+export declare function resolveVirtualMoneyFields(moneyFieldNames: Iterable<string>, isVirtual: (field: string) => boolean): Set<string>;
+/** One `ViaGraph.registerDerived` call's worth of edge data — target + sources,
+ *  `kind` chosen by the caller (#638 Task 2 edge extraction). `grain` defaults to
+ *  `'record'` when absent; `resolveComputedEdges` (#638 Task 7) sets it to `'virtual'`
+ *  for a `mode: 'virtual'` computed field's edge. */
+export interface GraphEdge {
+    readonly target: FieldRef;
+    readonly sources: readonly FieldRef[];
+    readonly grain?: Grain;
+}
+/**
+ * The known-field-name universe builder (#638 Task 2 fix wave 2) — used by
+ * {@link resolveCollectionConfig} to validate `resolveViaBindingDepsEdges`'s `deps`
+ * entries, and by {@link resolveComputedEdges}'s classified-collection dep-name check
+ * (Finding I2ii's "shared, so the two paths cannot silently drift apart" rationale).
+ * #638 Task 7 dropped `resolveComputedEdges`'s dep-name check entirely (legalizing a
+ * plain-field dep on a non-classified collection); the Task 7 review's CRITICAL finding
+ * (a typo'd dep on a CLASSIFIED collection reopens the #636 leak) restored a
+ * classified-only slice of it, scoped through this SAME helper rather than a
+ * hand-rolled second universe — exported so `via/graph-wiring.ts`'s reconcile path can
+ * build its own call-scoped knownFields the identical way the fresh path does.
+ */
+export declare function collectKnownFieldNames(parts: {
+    readonly moneyFields?: Record<string, unknown> | undefined;
+    readonly i18nFields?: Record<string, unknown> | undefined;
+    readonly dictKeyFields?: Record<string, unknown> | undefined;
+    readonly classifiedFields?: Record<string, unknown> | undefined;
+    readonly computed?: Record<string, unknown> | undefined;
+    readonly lookupFields?: Record<string, unknown> | undefined;
+}): Set<string>;
+/**
+ * Validate each `computed` entry's `deps` well-formedness and resolve them into graph
+ * edges (#638 Task 2; #638 Task 7 folded the formerly-separate `computedDeps` sibling
+ * option into each entry's own `{ fn, deps?, mode? }` shape — see
+ * `with-formula/computed/index.ts#ComputedFieldEntry`). `computed` is the RAW
+ * user-declared map — `unifyComputedFields`'s union of the `computed:` sugar option and
+ * `via(computed(...))` entries, NEVER `mergedComputed`: a classified preset's
+ * `riderComputed` companions are a sanctioned, already-vetted classified→computed channel
+ * and are deliberately NOT subject to this guard. A depsless entry is fine UNLESS the
+ * collection also declares classified fields (`hasClassifiedFields`), in which case an
+ * opaque computed function could silently copy a classified field's plaintext into an
+ * ordinary, unredacted field (the #636 leak) — refused at declare time, regardless of
+ * `mode` (a virtual field's read-time redaction, `via/taint-binding.ts`, only fires for a
+ * field the graph actually taints — a depsless one never is).
+ *
+ * A `deps` entry may name ANY field — including a PLAIN field with no via feature
+ * declared on it at all (#638 Task 7; Task 2's original design only ever validated a
+ * `deps` entry against `moneyFields`/`i18nFields`/`dictKeyFields`/`classifiedFields`/
+ * `computed`'s own declared names — `collection-config.ts` has no schema-introspection
+ * API to check a `deps` entry against the record's FULL field set, since
+ * `StandardSchemaV1` — Zod/Valibot/ArkType/Effect, deliberately schema-library-agnostic —
+ * exposes no "list of field names" capability). This is safe on a NON-classified
+ * collection: a dep with no registered `ViaGraph` node contributes `DEFAULT_POSTURE`
+ * when folded (`ViaGraph._contribution`'s `?? DEFAULT_POSTURE` fallback) — i.e. nothing,
+ * exactly like any other untainted source.
+ *
+ * Task 7 review CRITICAL fix (empirically confirmed leak reopening): on a collection
+ * that DOES declare classified fields, an UNKNOWN `deps` entry (a typo — e.g.
+ * `deps: ['sssn']` instead of `['ssn']`) used to fold to `DEFAULT_POSTURE` exactly like
+ * the "harmless plain field" case above, silently reopening the #636 leak — construction
+ * didn't throw, and the derived field was written/read UNSEALED even though its `fn`
+ * actually read a classified field. `knownFields` (built by {@link collectKnownFieldNames}
+ * the SAME way at every call site — never a hand-rolled second universe, the exact drift
+ * that caused a Task 2 bug) restores the "every dep must name a known field" check, but
+ * ONLY when `hasClassifiedFields` — a non-classified collection keeps the Task 7 freedom
+ * (any string is a legal dep, per the paragraph above) since an untainted fold there is
+ * always safe regardless of typos.
+ *
+ * KNOWN LIMIT (pre-existing since Task 2, phase-E territory — pinned by
+ * `via/taint.test.ts`'s "KNOWN LIMIT" test): this only checks that a `deps` entry names
+ * SOME known field, not that it names the RIGHT one. A `deps` entry naming a real,
+ * declared-but-WRONG field (e.g. `fn` reads classified field `ssn` but `deps: ['amount']`
+ * — a genuine, known, non-classified field) still passes this check and still leaks:
+ * the graph edge folds from `amount`'s posture, not `ssn`'s, so the derived field is
+ * folded/sealed as if it read `amount`, while its actual output is `ssn`'s plaintext.
+ * There is no schema-introspection capability (see above) to verify a `deps` entry
+ * actually corresponds to what `fn` reads — closing this fully would need runtime
+ * read-tracking or a schema-aware capability, out of this fix's scope.
+ */
+export declare function resolveComputedEdges(collectionName: string, computed: ComputedFields | undefined, hasClassifiedFields: boolean, knownFields?: ReadonlySet<string>): readonly GraphEdge[];
+/**
+ * Extract graph edges from `ViaBinding.deps` (#638 Task 2 — `deps` goes from
+ * inert to validated). For any compiled binding declaring `deps`, every field
+ * it `covers()` (tested against `knownFields`) becomes a derived target whose
+ * sources are `deps`; an unknown source field throws declare-time
+ * `ValidationError`. No shipped binding declares `deps` today (money/i18n/
+ * classified/blob don't) — this is the general path a future derive-bearing
+ * binding (phase C Task 7's `computed` via-binding) plugs into.
+ */
+export declare function resolveViaBindingDepsEdges(collectionName: string, bindings: readonly ViaBinding[], knownFields: ReadonlySet<string>): readonly GraphEdge[];
+/**
+ * Resolve the raw {@link CollectionOpts} into the concrete field values the
+ * {@link Collection} constructor assigns — every `?? default`, every derived
+ * `Set`/subset, plus the three pure construction-time validations. Pure: no
+ * `this`, no external side effect. The allocated `Set`/`VectorSet`/`Lru`
+ * instances become the collection's own by reference.
+ */
+export declare function resolveCollectionConfig<T>(opts: CollectionOpts<T>): {
+    adapter: NoydbStore;
+    vault: string;
+    name: string;
+    keyring: UnlockedKeyring;
+    storeCiphertext: boolean;
+    ramCiphertext: boolean;
+    emitter: NoydbEventEmitter;
+    writeQueue: WriteQueueTracker | undefined;
+    schemaUpdateGate: SchemaUpdateGate | undefined;
+    schemaFence: SchemaFenceController | undefined;
+    writeHooks: WriteHookRegistry | undefined;
+    subsystemBus: ServiceBus | undefined;
+    activeTxId: (() => string | null) | undefined;
+    blobStrategy: BlobStrategy;
+    objectStore: ObjectProjection | undefined;
+    blobFields: BlobFieldsConfig | undefined;
+    aggregateStrategy: AggregateStrategy;
+    crdtStrategy: CrdtStrategy;
+    historyStrategy: HistoryStrategy;
+    i18nStrategy: I18nStrategy;
+    syncStrategy: SyncStrategy;
+    reconcileOnOpen: "auto" | "off" | "dry-run";
+    getDEK: (collectionName: string) => Promise<EnclaveKey>;
+    onDirty: OnDirtyCallback | undefined;
+    tabCoordinated: (() => boolean) | undefined;
+    historyConfig: HistoryConfig;
+    historyConfigExplicit: boolean;
+    schema: StandardSchemaV1<unknown, T> | undefined;
+    ledger: LedgerStore | undefined;
+    refEnforcer: {
+        enforceRefsOnPut(collectionName: string, record: unknown): Promise<void>;
+        enforceRefsOnDelete(collectionName: string, id: string): Promise<void>;
+    } | undefined;
+    joinResolver: {
+        resolveSource(collectionName: string): JoinableSource | null;
+        resolveRef(leftCollection: string, field: string): RefDescriptor | null;
+        resolveDictSource?: (leftCollection: string, field: string) => JoinableSource | null;
+    } | undefined;
+    i18nFields: Record<string, I18nTextDescriptor> | undefined;
+    textIndexes: readonly string[] | undefined;
+    embeddings: EmbeddingDescriptor | undefined;
+    vectorSet: VectorSet | undefined;
+    dictKeyFields: Record<string, StaticDictDescriptor<string> | DictKeyDescriptor<string>> | undefined;
+    lookupFields: Record<string, LookupDescriptor<string>> | undefined;
+    presentForJoin: ((record: unknown, locale: string) => unknown) | undefined;
+    fieldMeta: Record<string, FieldMeta> | undefined;
+    meta: CollectionMeta | undefined;
+    _refs: Record<string, RefDescriptor>;
+    via: ViaPipeline | undefined;
+    classifiedEraseCfg: ClassifiedViaConfig | undefined;
+    moneyFields: Record<string, ViaDescriptor> | undefined;
+    classified: ResolvedClassified | undefined;
+    classifiedGuardCtx: ClassifiedGuardCtx;
+    classifiedStrategy: ClassifiedStrategy;
+    computed: ComputedFields | undefined;
+    computedEdges: readonly GraphEdge[];
+    viaDepsEdges: readonly GraphEdge[];
+    computedFieldNames: string[];
+    dictLabelResolver: ((dictName: string, key: string, locale: string, fallback?: string | readonly string[]) => Promise<string | undefined>) | undefined;
+    getDictionary: ((name: string) => Promise<DictionaryHandle>) | undefined;
+    i18nPutValidator: ((record: unknown) => void) | undefined;
+    autoTranslateHook: ((text: string, from: string, to: string, field: string, collection: string) => Promise<string>) | undefined;
+    defaultLocale: string | undefined;
+    crdtMode: CrdtMode | undefined;
+    syncAdapter: NoydbStore | undefined;
+    onAccess: ((op: "get" | "put" | "delete" | "reveal" | "verify" | "find", id: string) => Promise<void>) | undefined;
+    derivationSource: {
+        registry(): DerivationRegistry;
+        getCollection(name: string): Collection<Record<string, unknown>>;
+        /**
+         * Read-only vault facade handed to `derive(source, ctx)` so a
+         * derivation can fetch sibling records. Same shape and
+         * instance the guards service uses for `check(incoming, ctx)`.
+         */
+        getReadOnlyFacade(): ReadOnlyVaultFacade;
+        /**
+         * Read access to the owning Noydb's currently-active multi-record
+         * transaction context, or `null` when no transaction is running.
+         * `dispatchDerivations` consults this so a recursive derived-output
+         * write can register its pre-write envelope onto `ctx._executed`
+         * and roll back alongside the source op on mid-batch failure.
+         */
+        getActiveTxContext(): TxContext | null;
+        /**
+         * Construct a transient TxContext bound to the owning Noydb. Used
+         * by `Collection.putManyAtomic` to publish an active context for
+         * its Phase 2 loop.
+         */
+        createTxContext(): TxContext;
+        /** Publish a TxContext for the duration of a bulk-atomic loop. */
+        setActiveTxContext(ctx: TxContext): void;
+        /** Drop a previously-published TxContext. */
+        clearActiveTxContext(ctx: TxContext): void;
+    } | undefined;
+    materializedViewSource: {
+        registry(): MaterializedViewRegistry;
+        getCollection(name: string): Collection<any>;
+        getActiveTxContext(): TxContext | null;
+        getQueryContext(): MVQueryContext;
+    } | undefined;
+    graphDispatch: {
+        collect(collection: string, id: string): void;
+        collectDelete(collection: string, id: string, intents: readonly RollupDeleteIntent[]): void;
+    } | undefined;
+    tiers: Set<number> | null;
+    tiersStrategy: TiersStrategy;
+    searchStrategy: SearchStrategy;
+    tierMode: TierMode;
+    onCrossTierAccess: ((event: CrossTierAccessEvent) => void) | undefined;
+    deterministicFields: ReadonlySet<string> | null;
+    sensitiveFields: ReadonlySet<string>;
+    vdigFields: ReadonlyMap<string, VdigFieldPolicy> | null;
+    perRecordCek: boolean;
+    cekCache: Lru<string, CryptoKey> | null;
+    provenance: boolean;
+};