@taladb/react 0.9.3 → 0.9.4

package/dist/index.d.mts

@@ -1,6 +1,6 @@
 import * as react_jsx_runtime from 'react/jsx-runtime';
 import { ReactNode } from 'react';
-import { CollectionOptions, Document, TalaDB, OpenDBOptions, Collection, Filter } from 'taladb';
+import { CollectionOptions, Document, TalaDB, OpenDBOptions, Collection, Filter, AggregatePipeline, RestSourceOptions, ReplicationSource, CoverageState } from 'taladb';
 
 /**
  * Per-collection options (`schema`, `syncSchema`, `migrateDocument`, …), keyed by
@@ -163,6 +163,40 @@ interface FindOneResult<T> {
  */
 declare function useFindOne<T extends Document>(collection: Collection<T>, filter: Filter<T>): FindOneResult<T>;
 
+interface AggregateResult<R> {
+    /** The current pipeline results. Empty while loading. */
+    data: R[];
+    /** True until the first snapshot has been delivered. */
+    loading: boolean;
+    /** Most recent subscription error, cleared by the next successful snapshot. */
+    error: unknown | null;
+}
+/**
+ * Subscribe to a **live aggregation**. Re-renders whenever the results change.
+ *
+ * This is the paging primitive. `find()` has no sort/skip/limit — only `aggregate`
+ * does — but `aggregate` on its own returns a *dead snapshot*: call it in an effect
+ * and the page sits frozen while a background hydration fills the collection
+ * underneath it. The user watches a spinner finish and the rows never arrive.
+ *
+ * So anything that pages locally subscribes here rather than calling `aggregate`
+ * directly.
+ *
+ * @param collection A `Collection<T>` (memoize it, or take it from `useCollection`).
+ * @param pipeline   Inline arrays are safe — the pipeline is serialised for
+ *                   subscription identity, so a fresh array each render does not
+ *                   re-subscribe.
+ *
+ * @example
+ * const page = useAggregate<Product, Product>(products, [
+ *   { $match: { category: 'kitchen' } },
+ *   { $sort: { price: 1 } },
+ *   { $skip: 100 },
+ *   { $limit: 100 },
+ * ])
+ */
+declare function useAggregate<T extends Document, R extends Document = T>(collection: Collection<T>, pipeline: AggregatePipeline<T>): AggregateResult<R>;
+
 /** Resolved network configuration for one replicated slice. */
 interface ResolvedReplicationConfig {
     /** Base URL; `/push` and `/pull` are appended by {@link HttpSyncAdapter}. */
@@ -183,6 +217,37 @@ interface ResolvedReplicationConfig {
     };
 }
 
+/** When the background hydration walk is allowed to start. */
+type HydrateMode = 
+/** Immediately on mount. */
+'eager'
+/** When the browser is idle (default). Keeps first paint responsive. */
+ | 'idle'
+/** Never automatically — the app calls `hydrate()` itself. */
+ | 'manual';
+interface ReplicateScope<RemoteRow = any, T extends Document = Document> extends Omit<RestSourceOptions<RemoteRow, T>, 'collection'> {
+    /**
+     * Provide a fully custom source instead of the REST defaults. When present,
+     * every other field here is ignored.
+     */
+    source?: ReplicationSource<RemoteRow, T>;
+    hydrate?: HydrateMode;
+    /** Rows per bootstrap page. Default 500. */
+    pageSize?: number;
+    /** Re-check the origin for changes on this interval. `0` disables. */
+    refreshMs?: number;
+    /**
+     * Fetch the current query directly when coverage isn't ready yet, so a cold
+     * start paints immediately. Default `true`.
+     *
+     * Mandatory in practice for a Vite SPA or React Native, which have no server
+     * render to paint behind while the replica fills.
+     */
+    bridge?: boolean;
+}
+/** One entry per local collection. */
+type ReplicateRegistry = Record<string, ReplicateScope<any, any>>;
+
 /** A slice to warm on first run — a collection, optionally on a specific endpoint. */
 type PrefetchSlice = {
     collection: string;
@@ -236,7 +301,15 @@ interface ReplicationConfig {
     /** Max concurrent prefetch pulls — keeps the active page from starving. Default `2`. */
     prefetchConcurrency?: number;
 }
-interface ReplicationProviderProps extends ReplicationConfig {
+interface ReplicationProviderProps extends Partial<ReplicationConfig> {
+    /**
+     * Collections to replicate from a remote origin, keyed by local collection name.
+     *
+     * This is what makes `useQuery` a local read: once a collection is fully
+     * hydrated for its scope, filtering, sorting and paging never touch the network.
+     * See {@link ReplicateRegistry}.
+     */
+    replicate?: ReplicateRegistry;
     children: ReactNode;
 }
 /**
@@ -257,7 +330,7 @@ interface ReplicationProviderProps extends ReplicationConfig {
  * </TalaDBProvider>
  * ```
  */
-declare function ReplicationProvider({ children, ...config }: ReplicationProviderProps): react_jsx_runtime.JSX.Element;
+declare function ReplicationProvider({ children, replicate, ...config }: ReplicationProviderProps): react_jsx_runtime.JSX.Element;
 /**
  * Read the nearest replication config, merged with per-hook overrides.
  * Non-throwing: `config` is `null` when no endpoint is resolvable (valid for
@@ -268,73 +341,134 @@ declare function useReplicationConfig(overrides?: Partial<ReplicationConfig>): {
     pollMs: number;
 };
 
+interface Coverage {
+    /** The raw state machine value. */
+    status: CoverageState['status'];
+    /**
+     * Whether a purely local read is authorized.
+     *
+     * True **only** for `complete`. Notably *not* for `best-effort`, which means we
+     * applied every row the origin gave us but the origin could not pin a snapshot —
+     * so a row that shifted between pages during the walk may never have been seen,
+     * and we cannot prove the replica is whole. Serving that as authoritative would
+     * silently return incomplete results, which is worse than going to the network.
+     */
+    ready: boolean;
+    /** Rows hydrated so far. */
+    rows: number;
+    /** Total rows in scope when supplied by the origin. */
+    total?: number;
+    /** 0–1 when the origin reported a total; otherwise undefined. */
+    progress?: number;
+    /** Present on `error`, `stale` and `best-effort`. */
+    reason?: string;
+}
 /**
- * How a `useQuery` combines the local replica with the remote origin.
- * *(`remote-only` is a planned addition; the durable-replica model makes it a
- * rare escape hatch.)*
+ * How much of a collection is local, and whether it can be trusted for a
+ * network-free read.
+ *
+ * @example
+ * const { ready, progress } = useCoverage('products')
+ * if (!ready) return <ProgressBar value={progress} />
  */
+declare function useCoverage(collection: string): Coverage;
+/** Semantic alias for progress-oriented UIs. */
+declare const useHydrationProgress: typeof useCoverage;
+
 type ReadSource = 'local-first' | 'remote-first' | 'local-only';
 interface UseQueryOptions<T extends Document> extends Partial<Pick<ReplicationConfig, 'endpoint' | 'getAuth' | 'fetch' | 'paths' | 'pollMs'>> {
-    /** Collection name — replicated as a whole; the `filter` narrows locally. */
+    /** The local collection to read. */
     collection: string;
-    /** Live filter over the local collection. Inline objects are safe. */
+    /** Mongo-style filter, applied **locally**. */
     filter?: Filter<T>;
-    /**
-     * - `local-first` *(default)* — serve local immediately; refresh in the
-     *   background, and the live query re-renders when the pull lands.
-     * - `remote-first` — stay `loading` until the first pull completes, then serve.
-     * - `local-only` — never touch the network (no endpoint required).
-     */
+    /** Sort, applied locally. `1` ascending, `-1` descending. */
+    sort?: Partial<Record<keyof T & string, 1 | -1>>;
+    /** 1-based page number. Requires `limit`. */
+    page?: number;
+    /** Rows per page. */
+    limit?: number;
+    /** Skip N rows. Ignored when `page` is set. */
+    skip?: number;
+    /** Skip the query entirely (e.g. while a route param is undefined). */
+    enabled?: boolean;
+    /** Legacy sync-contract read policy. Used when no full-replication scope exists. */
     source?: ReadSource;
 }
 interface QueryResult<T> {
-    /** Current matching documents from the local replica. Reactive. */
+    /** The current page. Reactive: re-renders as rows land. */
     data: T[];
-    /** First local snapshot pending — plus the first pull, for `remote-first`. */
+    /** Total rows in the replicated scope when reported by the origin. */
+    total?: number;
+    /** True until the first local snapshot arrives. */
     loading: boolean;
-    /** Most recent local-read error. */
+    /** Most recent local read error. */
     error: unknown | null;
-    /** A background replication pass is in flight. */
+    /** Most recent cold-start bridge error. */
+    fetchError: unknown | null;
+    /** How much of the collection is local, and whether it is trustworthy. */
+    coverage: Coverage;
+    /**
+     * A cold-start bridge fetch is in flight — we are serving the network because
+     * the replica isn't complete yet.
+     */
+    fetching: boolean;
+    /** Legacy sync-contract pull in progress. */
     syncing: boolean;
-    /** Most recent replication error (the local data is still served). */
+    /** Legacy sync-contract pull error. */
     syncError: unknown | null;
-    /** Trigger a pull now. No-op for `local-only`. */
+    /** Force a delta refresh from the origin. */
     refetch: () => Promise<void>;
 }
 /**
- * Bind a component to a slice of a remote origin, backed by the local replica.
+ * Read a page of a collection.
+ *
+ * ## The point
+ *
+ * Once the collection is **covered** — fully replicated for this scope — this hook
+ * touches the network **zero times**. Filtering, sorting and paging are local
+ * queries against the on-device database. Page 1 → page 2 → a new filter → page 47
+ * → back to page 1: every one is a local read, instant and offline-capable.
+ * Pagination stops being a network concern at all.
+ *
+ * That is the whole reason to put a real database on the device. A cache of API
+ * pages could only ever answer the queries you already asked; a *covered replica*
+ * answers queries nobody has asked yet.
+ *
+ * ## Before coverage lands
  *
- * The read is a live query over the local collection (`useFind`); the network
- * pull writes into that same collection, so the live query re-renders on its
- * own — one-way data flow, no `queryKey`, no `invalidateQueries`. See
- * `docs/scoped-replication.md`.
+ * On a cold start the replica is empty, and a SPA or React Native app has no server
+ * render to paint behind. So the hook **bridges**: it fetches exactly the rows this
+ * query needs and writes them into the same collection, under the same derived ids
+ * the background walk will use. Those rows are not a cache entry to be reconciled
+ * later — they are the replica, arriving early. When the walk reaches them it
+ * overwrites them in place.
  *
  * @example
- * const { data, loading, syncing } = useQuery<Product>({
+ * const { data, coverage } = useQuery<Product>({
  *   collection: 'products',
- *   filter: { category: 'kitchen' },
- *   pollMs: 30_000,
+ *   filter: { category: 'kitchen', price: { $lt: 500 } },
+ *   sort: { price: 1 },
+ *   page: 2,
+ *   limit: 100,
  * })
  */
 declare function useQuery<T extends Document>(options: UseQueryOptions<T>): QueryResult<T>;
 
 /**
- * Run several scoped queries at once — one replication + live query per entry,
- * in parallel. The result array is index-aligned with `queries`.
+ * Several pages at once, from several collections. Index-aligned with `queries`.
  *
- * Each entry is an independent `useQuery`: its own collection, filter, source,
- * and (optionally) endpoint. This is the multi-slice page case — e.g. a
- * dashboard that needs `orders` and `products` from the same origin. There are
- * no cross-collection joins (TalaDB is a document store); compose in the
- * component.
+ * Hooks can't be called in a variable-length loop, so this manages its own
+ * subscriptions rather than calling `useQuery` N times. Behaviour is otherwise
+ * identical: once a collection is covered, its read is purely local.
  *
- * v1 note: entries are typed as `Document`. For a strictly-typed single slice
- * use `useQuery<T>`; per-entry generics (tuple typing) are a follow-up.
+ * TalaDB is a document store with no cross-collection joins, so a page needing
+ * several collections composes them in the component (or denormalises at the
+ * origin). This is the hook for that.
  *
  * @example
- * const [orders, products] = useQueries([
- *   { collection: 'orders', filter: { open: true } },
- *   { collection: 'products' },
+ * const [products, categories] = useQueries([
+ *   { collection: 'products', filter: { category }, sort: { price: 1 }, page, limit: 50 },
+ *   { collection: 'categories' },
  * ])
  */
 declare function useQueries(queries: UseQueryOptions<Document>[]): QueryResult<Document>[];
@@ -392,4 +526,4 @@ interface MutationResult<T extends Document> {
  */
 declare function useMutation<T extends Document>(options: UseMutationOptions): MutationResult<T>;
 
-export { type CollectionRegistry, type CollectionResolver, type FindOneResult, type FindResult, type MutationResult, type PrefetchEntry, type PrefetchMode, type PrefetchSlice, type QueryResult, type ReadSource, type ReplicationConfig, ReplicationProvider, type ReplicationProviderProps, TalaDBProvider, type TalaDBProviderProps, type UseMutationOptions, type UseQueryOptions, type WriteOp, useCollection, useCollectionOptions, useFind, useFindOne, useMutation, useQueries, useQuery, useReplicationConfig, useTalaDB };
+export { type AggregateResult, type CollectionRegistry, type CollectionResolver, type Coverage, type FindOneResult, type FindResult, type HydrateMode, type MutationResult, type PrefetchEntry, type PrefetchMode, type PrefetchSlice, type QueryResult, type ReplicateRegistry, type ReplicateScope, type ReplicationConfig, ReplicationProvider, type ReplicationProviderProps, TalaDBProvider, type TalaDBProviderProps, type UseMutationOptions, type UseQueryOptions, type WriteOp, useAggregate, useCollection, useCollectionOptions, useCoverage, useFind, useFindOne, useHydrationProgress, useMutation, useQueries, useQuery, useReplicationConfig, useTalaDB };

package/dist/index.d.ts

@@ -1,6 +1,6 @@
 import * as react_jsx_runtime from 'react/jsx-runtime';
 import { ReactNode } from 'react';
-import { CollectionOptions, Document, TalaDB, OpenDBOptions, Collection, Filter } from 'taladb';
+import { CollectionOptions, Document, TalaDB, OpenDBOptions, Collection, Filter, AggregatePipeline, RestSourceOptions, ReplicationSource, CoverageState } from 'taladb';
 
 /**
  * Per-collection options (`schema`, `syncSchema`, `migrateDocument`, …), keyed by
@@ -163,6 +163,40 @@ interface FindOneResult<T> {
  */
 declare function useFindOne<T extends Document>(collection: Collection<T>, filter: Filter<T>): FindOneResult<T>;
 
+interface AggregateResult<R> {
+    /** The current pipeline results. Empty while loading. */
+    data: R[];
+    /** True until the first snapshot has been delivered. */
+    loading: boolean;
+    /** Most recent subscription error, cleared by the next successful snapshot. */
+    error: unknown | null;
+}
+/**
+ * Subscribe to a **live aggregation**. Re-renders whenever the results change.
+ *
+ * This is the paging primitive. `find()` has no sort/skip/limit — only `aggregate`
+ * does — but `aggregate` on its own returns a *dead snapshot*: call it in an effect
+ * and the page sits frozen while a background hydration fills the collection
+ * underneath it. The user watches a spinner finish and the rows never arrive.
+ *
+ * So anything that pages locally subscribes here rather than calling `aggregate`
+ * directly.
+ *
+ * @param collection A `Collection<T>` (memoize it, or take it from `useCollection`).
+ * @param pipeline   Inline arrays are safe — the pipeline is serialised for
+ *                   subscription identity, so a fresh array each render does not
+ *                   re-subscribe.
+ *
+ * @example
+ * const page = useAggregate<Product, Product>(products, [
+ *   { $match: { category: 'kitchen' } },
+ *   { $sort: { price: 1 } },
+ *   { $skip: 100 },
+ *   { $limit: 100 },
+ * ])
+ */
+declare function useAggregate<T extends Document, R extends Document = T>(collection: Collection<T>, pipeline: AggregatePipeline<T>): AggregateResult<R>;
+
 /** Resolved network configuration for one replicated slice. */
 interface ResolvedReplicationConfig {
     /** Base URL; `/push` and `/pull` are appended by {@link HttpSyncAdapter}. */
@@ -183,6 +217,37 @@ interface ResolvedReplicationConfig {
     };
 }
 
+/** When the background hydration walk is allowed to start. */
+type HydrateMode = 
+/** Immediately on mount. */
+'eager'
+/** When the browser is idle (default). Keeps first paint responsive. */
+ | 'idle'
+/** Never automatically — the app calls `hydrate()` itself. */
+ | 'manual';
+interface ReplicateScope<RemoteRow = any, T extends Document = Document> extends Omit<RestSourceOptions<RemoteRow, T>, 'collection'> {
+    /**
+     * Provide a fully custom source instead of the REST defaults. When present,
+     * every other field here is ignored.
+     */
+    source?: ReplicationSource<RemoteRow, T>;
+    hydrate?: HydrateMode;
+    /** Rows per bootstrap page. Default 500. */
+    pageSize?: number;
+    /** Re-check the origin for changes on this interval. `0` disables. */
+    refreshMs?: number;
+    /**
+     * Fetch the current query directly when coverage isn't ready yet, so a cold
+     * start paints immediately. Default `true`.
+     *
+     * Mandatory in practice for a Vite SPA or React Native, which have no server
+     * render to paint behind while the replica fills.
+     */
+    bridge?: boolean;
+}
+/** One entry per local collection. */
+type ReplicateRegistry = Record<string, ReplicateScope<any, any>>;
+
 /** A slice to warm on first run — a collection, optionally on a specific endpoint. */
 type PrefetchSlice = {
     collection: string;
@@ -236,7 +301,15 @@ interface ReplicationConfig {
     /** Max concurrent prefetch pulls — keeps the active page from starving. Default `2`. */
     prefetchConcurrency?: number;
 }
-interface ReplicationProviderProps extends ReplicationConfig {
+interface ReplicationProviderProps extends Partial<ReplicationConfig> {
+    /**
+     * Collections to replicate from a remote origin, keyed by local collection name.
+     *
+     * This is what makes `useQuery` a local read: once a collection is fully
+     * hydrated for its scope, filtering, sorting and paging never touch the network.
+     * See {@link ReplicateRegistry}.
+     */
+    replicate?: ReplicateRegistry;
     children: ReactNode;
 }
 /**
@@ -257,7 +330,7 @@ interface ReplicationProviderProps extends ReplicationConfig {
  * </TalaDBProvider>
  * ```
  */
-declare function ReplicationProvider({ children, ...config }: ReplicationProviderProps): react_jsx_runtime.JSX.Element;
+declare function ReplicationProvider({ children, replicate, ...config }: ReplicationProviderProps): react_jsx_runtime.JSX.Element;
 /**
  * Read the nearest replication config, merged with per-hook overrides.
  * Non-throwing: `config` is `null` when no endpoint is resolvable (valid for
@@ -268,73 +341,134 @@ declare function useReplicationConfig(overrides?: Partial<ReplicationConfig>): {
     pollMs: number;
 };
 
+interface Coverage {
+    /** The raw state machine value. */
+    status: CoverageState['status'];
+    /**
+     * Whether a purely local read is authorized.
+     *
+     * True **only** for `complete`. Notably *not* for `best-effort`, which means we
+     * applied every row the origin gave us but the origin could not pin a snapshot —
+     * so a row that shifted between pages during the walk may never have been seen,
+     * and we cannot prove the replica is whole. Serving that as authoritative would
+     * silently return incomplete results, which is worse than going to the network.
+     */
+    ready: boolean;
+    /** Rows hydrated so far. */
+    rows: number;
+    /** Total rows in scope when supplied by the origin. */
+    total?: number;
+    /** 0–1 when the origin reported a total; otherwise undefined. */
+    progress?: number;
+    /** Present on `error`, `stale` and `best-effort`. */
+    reason?: string;
+}
 /**
- * How a `useQuery` combines the local replica with the remote origin.
- * *(`remote-only` is a planned addition; the durable-replica model makes it a
- * rare escape hatch.)*
+ * How much of a collection is local, and whether it can be trusted for a
+ * network-free read.
+ *
+ * @example
+ * const { ready, progress } = useCoverage('products')
+ * if (!ready) return <ProgressBar value={progress} />
  */
+declare function useCoverage(collection: string): Coverage;
+/** Semantic alias for progress-oriented UIs. */
+declare const useHydrationProgress: typeof useCoverage;
+
 type ReadSource = 'local-first' | 'remote-first' | 'local-only';
 interface UseQueryOptions<T extends Document> extends Partial<Pick<ReplicationConfig, 'endpoint' | 'getAuth' | 'fetch' | 'paths' | 'pollMs'>> {
-    /** Collection name — replicated as a whole; the `filter` narrows locally. */
+    /** The local collection to read. */
     collection: string;
-    /** Live filter over the local collection. Inline objects are safe. */
+    /** Mongo-style filter, applied **locally**. */
     filter?: Filter<T>;
-    /**
-     * - `local-first` *(default)* — serve local immediately; refresh in the
-     *   background, and the live query re-renders when the pull lands.
-     * - `remote-first` — stay `loading` until the first pull completes, then serve.
-     * - `local-only` — never touch the network (no endpoint required).
-     */
+    /** Sort, applied locally. `1` ascending, `-1` descending. */
+    sort?: Partial<Record<keyof T & string, 1 | -1>>;
+    /** 1-based page number. Requires `limit`. */
+    page?: number;
+    /** Rows per page. */
+    limit?: number;
+    /** Skip N rows. Ignored when `page` is set. */
+    skip?: number;
+    /** Skip the query entirely (e.g. while a route param is undefined). */
+    enabled?: boolean;
+    /** Legacy sync-contract read policy. Used when no full-replication scope exists. */
     source?: ReadSource;
 }
 interface QueryResult<T> {
-    /** Current matching documents from the local replica. Reactive. */
+    /** The current page. Reactive: re-renders as rows land. */
     data: T[];
-    /** First local snapshot pending — plus the first pull, for `remote-first`. */
+    /** Total rows in the replicated scope when reported by the origin. */
+    total?: number;
+    /** True until the first local snapshot arrives. */
     loading: boolean;
-    /** Most recent local-read error. */
+    /** Most recent local read error. */
     error: unknown | null;
-    /** A background replication pass is in flight. */
+    /** Most recent cold-start bridge error. */
+    fetchError: unknown | null;
+    /** How much of the collection is local, and whether it is trustworthy. */
+    coverage: Coverage;
+    /**
+     * A cold-start bridge fetch is in flight — we are serving the network because
+     * the replica isn't complete yet.
+     */
+    fetching: boolean;
+    /** Legacy sync-contract pull in progress. */
     syncing: boolean;
-    /** Most recent replication error (the local data is still served). */
+    /** Legacy sync-contract pull error. */
     syncError: unknown | null;
-    /** Trigger a pull now. No-op for `local-only`. */
+    /** Force a delta refresh from the origin. */
     refetch: () => Promise<void>;
 }
 /**
- * Bind a component to a slice of a remote origin, backed by the local replica.
+ * Read a page of a collection.
+ *
+ * ## The point
+ *
+ * Once the collection is **covered** — fully replicated for this scope — this hook
+ * touches the network **zero times**. Filtering, sorting and paging are local
+ * queries against the on-device database. Page 1 → page 2 → a new filter → page 47
+ * → back to page 1: every one is a local read, instant and offline-capable.
+ * Pagination stops being a network concern at all.
+ *
+ * That is the whole reason to put a real database on the device. A cache of API
+ * pages could only ever answer the queries you already asked; a *covered replica*
+ * answers queries nobody has asked yet.
+ *
+ * ## Before coverage lands
  *
- * The read is a live query over the local collection (`useFind`); the network
- * pull writes into that same collection, so the live query re-renders on its
- * own — one-way data flow, no `queryKey`, no `invalidateQueries`. See
- * `docs/scoped-replication.md`.
+ * On a cold start the replica is empty, and a SPA or React Native app has no server
+ * render to paint behind. So the hook **bridges**: it fetches exactly the rows this
+ * query needs and writes them into the same collection, under the same derived ids
+ * the background walk will use. Those rows are not a cache entry to be reconciled
+ * later — they are the replica, arriving early. When the walk reaches them it
+ * overwrites them in place.
  *
  * @example
- * const { data, loading, syncing } = useQuery<Product>({
+ * const { data, coverage } = useQuery<Product>({
  *   collection: 'products',
- *   filter: { category: 'kitchen' },
- *   pollMs: 30_000,
+ *   filter: { category: 'kitchen', price: { $lt: 500 } },
+ *   sort: { price: 1 },
+ *   page: 2,
+ *   limit: 100,
  * })
  */
 declare function useQuery<T extends Document>(options: UseQueryOptions<T>): QueryResult<T>;
 
 /**
- * Run several scoped queries at once — one replication + live query per entry,
- * in parallel. The result array is index-aligned with `queries`.
+ * Several pages at once, from several collections. Index-aligned with `queries`.
  *
- * Each entry is an independent `useQuery`: its own collection, filter, source,
- * and (optionally) endpoint. This is the multi-slice page case — e.g. a
- * dashboard that needs `orders` and `products` from the same origin. There are
- * no cross-collection joins (TalaDB is a document store); compose in the
- * component.
+ * Hooks can't be called in a variable-length loop, so this manages its own
+ * subscriptions rather than calling `useQuery` N times. Behaviour is otherwise
+ * identical: once a collection is covered, its read is purely local.
  *
- * v1 note: entries are typed as `Document`. For a strictly-typed single slice
- * use `useQuery<T>`; per-entry generics (tuple typing) are a follow-up.
+ * TalaDB is a document store with no cross-collection joins, so a page needing
+ * several collections composes them in the component (or denormalises at the
+ * origin). This is the hook for that.
  *
  * @example
- * const [orders, products] = useQueries([
- *   { collection: 'orders', filter: { open: true } },
- *   { collection: 'products' },
+ * const [products, categories] = useQueries([
+ *   { collection: 'products', filter: { category }, sort: { price: 1 }, page, limit: 50 },
+ *   { collection: 'categories' },
  * ])
  */
 declare function useQueries(queries: UseQueryOptions<Document>[]): QueryResult<Document>[];
@@ -392,4 +526,4 @@ interface MutationResult<T extends Document> {
  */
 declare function useMutation<T extends Document>(options: UseMutationOptions): MutationResult<T>;
 
-export { type CollectionRegistry, type CollectionResolver, type FindOneResult, type FindResult, type MutationResult, type PrefetchEntry, type PrefetchMode, type PrefetchSlice, type QueryResult, type ReadSource, type ReplicationConfig, ReplicationProvider, type ReplicationProviderProps, TalaDBProvider, type TalaDBProviderProps, type UseMutationOptions, type UseQueryOptions, type WriteOp, useCollection, useCollectionOptions, useFind, useFindOne, useMutation, useQueries, useQuery, useReplicationConfig, useTalaDB };
+export { type AggregateResult, type CollectionRegistry, type CollectionResolver, type Coverage, type FindOneResult, type FindResult, type HydrateMode, type MutationResult, type PrefetchEntry, type PrefetchMode, type PrefetchSlice, type QueryResult, type ReplicateRegistry, type ReplicateScope, type ReplicationConfig, ReplicationProvider, type ReplicationProviderProps, TalaDBProvider, type TalaDBProviderProps, type UseMutationOptions, type UseQueryOptions, type WriteOp, useAggregate, useCollection, useCollectionOptions, useCoverage, useFind, useFindOne, useHydrationProgress, useMutation, useQueries, useQuery, useReplicationConfig, useTalaDB };