Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
// Search / sort ordering for the Media Collections grid (#3283).
//
// The grid is dominated by collections nothing created on purpose: every
// Creative Director project, Writers Room render batch, and universe/series
// render bucket auto-files itself as a collection, so a user with a handful of
// real collections scrolls past dozens of empty auto-generated ones to find
// them. These helpers give the page one definition of "auto-generated" and the
// display ordering that keeps real collections on top.
//
// Pure — no React, no I/O. The page and its tests share them.
import { tokenizeQuery, matchHaystack } from './mediaSearch.js';
// Shared name prefixes that auto-creators stamp onto every collection they
// make, paired with the badge label the card renders above the title instead
// of inside it — otherwise every auto-generated card shows the same clipped
// prefix and the trailing project name/date that tells them apart is what gets
// cut. This list drives PRESENTATION (the badge) plus the provenance fallback
// for records an older peer sent without a `source` stamp — the classification
// itself comes from `collection.source` now (#3311), so a new auto-creator only
// has to stamp `source: 'auto'` server-side, not extend this list (it should
// still add its prefix here if it wants the badge). Keep in sync with the
// server-side creators: `universeCollectionNameFor`
// / `seriesCollectionNameFor` in `server/services/mediaCollections.js`,
// `server/services/creativeDirector/projects{DB,File}.js`, and
// `server/services/writersRoom/local.js`.
export const AUTO_NAME_PREFIXES = [
'Creative Director: ',
'Writers Room: ',
'Universe: ',
'Series: ',
].map((prefix) => ({ prefix, badge: prefix.replace(/[:\s]+$/, '') }));
// Descriptions stamped by the auto-creators. A user can't produce these by
// accident through the UI (the create form takes a name only).
const AUTO_DESCRIPTION_PREFIXES = ['Auto-created for project ', 'Auto-generated images for '];
// Deterministic id prefixes for the universe-/series-linked render buckets the
// pipeline files into (see `mediaCollections.js` — `uc-<universeId>` /
// `sc-<seriesId>`).
const AUTO_ID_PREFIXES = ['uc-', 'sc-'];
// One collator for every name comparison — `String.localeCompare` rebuilds
// collation state on each call, and these run per comparison during a sort.
// `sensitivity: 'base'` makes it case- and accent-insensitive, so the sort key
// needs no `toLowerCase()`.
const collator = new Intl.Collator(undefined, { sensitivity: 'base' });
/**
* Split a collection name into its auto-creator badge label and the remaining
* title. Returns `{ badge: null, title: name }` for a user-named collection.
* @param {string} name
* @returns {{ badge: string|null, title: string }}
*/
export function splitCollectionName(name) {
const str = typeof name === 'string' ? name : '';
for (const { prefix, badge } of AUTO_NAME_PREFIXES) {
if (str.startsWith(prefix) && str.length > prefix.length) {
return { badge, title: str.slice(prefix.length) };
}
}
return { badge: null, title: str };
}
/**
* True when a collection was created by an automated flow rather than by the
* user.
*
* The server stamps provenance at mint time (`source: 'auto' | 'user'`, #3311)
* and migration 220 backfilled existing records, so the stamp is authoritative
* whenever it is present — a new auto-creator no longer has to remember to
* extend a client-side list.
*
* An ABSENT `source` is a third state, NOT a synonym for `'user'`. The stamp is
* local to each install (it is stripped from the sync wire — see
* `sanitizeRecordForWire`), so an unstamped record is one this install received
* from a peer after its own migration ran, or restored from a pre-migration
* backup. Those fall back to the marker heuristic below, which is what
* classified every collection before #3311 — so the fallback is never worse
* than the old behavior. Any ONE of the four independent markers is sufficient
* — an install can hold records from before a given marker existed, so this
* must not require all of them to agree.
* @param {object} collection
* @returns {boolean}
*/
export function isAutoCollection(collection) {
if (!collection || collection.synthetic) return false;
if (collection.source === 'auto') return true;
if (collection.source === 'user') return false;
if (splitCollectionName(collection.name).badge) return true;
const description = typeof collection.description === 'string' ? collection.description : '';
if (AUTO_DESCRIPTION_PREFIXES.some((p) => description.startsWith(p))) return true;
const id = typeof collection.id === 'string' ? collection.id : '';
if (AUTO_ID_PREFIXES.some((p) => id.startsWith(p))) return true;
return Boolean(collection.universeId || collection.seriesId);
}
/** Item count for a collection (0 for a malformed/absent items array). */
export function collectionItemCount(collection) {
return Number.isFinite(collection?.itemCount) ? collection.itemCount : Array.isArray(collection?.items) ? collection.items.length : 0;
}
// The sort control's options, in menu order. First entry is the default.
export const COLLECTION_SORTS = [
{ id: 'updated', label: 'Recently updated' },
{ id: 'name', label: 'Name' },
{ id: 'count', label: 'Item count' },
];
export const DEFAULT_COLLECTION_SORT = COLLECTION_SORTS[0].id;
/** Coerce a stored/URL sort id to a known one (unknown → the default). */
export function normalizeCollectionSort(raw) {
return COLLECTION_SORTS.some((s) => s.id === raw) ? raw : DEFAULT_COLLECTION_SORT;
}
// Sort timestamp. `null`/absent means "never stamped" — distinct from a real
// epoch-0 date — and sorts last rather than pretending to be the oldest record.
const updatedTime = (collection) => {
const raw = collection?.updatedAt || collection?.createdAt;
const ms = raw ? Date.parse(raw) : NaN;
return Number.isFinite(ms) ? ms : null;
};
// Derive every sort/search key ONCE per record. Recomputing them inside the
// comparator would re-run the prefix split, the auto-collection markers, and
// Date.parse on both operands of every one of the O(n log n) comparisons.
//
// The name key is the PREFIX-STRIPPED title, so the auto-generated entries
// interleave by their real project names instead of all clumping under "C"/"U"
// — the same reason the prefix moves to a badge in the card.
const decorate = (record) => {
const count = collectionItemCount(record);
return {
record,
// Ordering bucket: synthetic ("Unsorted") is pinned first, then anything
// with items or a user-chosen name, then the auto-generated empties that
// motivated this. Within a bucket the caller's sort applies.
bucket: record?.synthetic ? 0 : ((count === 0 && isAutoCollection(record)) ? 2 : 1),
nameKey: splitCollectionName(record?.name).title,
time: updatedTime(record),
count,
// Same AND-token semantics as the media search, over name + description.
haystack: `${record?.name || ''} ${record?.description || ''}`.toLowerCase(),
};
};
const compareBy = (sort) => (a, b) => {
if (sort === 'name') return collator.compare(a.nameKey, b.nameKey);
if (sort === 'count') return (b.count - a.count) || collator.compare(a.nameKey, b.nameKey);
if (a.time === b.time) return collator.compare(a.nameKey, b.nameKey);
if (a.time === null) return 1;
if (b.time === null) return -1;
return b.time - a.time;
};
/**
* Filter + order the collection grid. "Hide empty" is deliberately NOT applied
* here: the page needs both the full match set and the non-empty slice to say
* how many rows the toggle removed, so it owns that one predicate
* (`collectionItemCount(c) > 0`) rather than having two definitions of empty.
* @param {object[]} collections - Enriched collections (synthetic entry included)
* @param {object} view
* @param {string} [view.query] - Free-text name/description search
* @param {string} [view.sort] - One of COLLECTION_SORTS ids
* @returns {object[]} A new array of the input records; the input is not mutated
*/
export function applyCollectionView(collections, { query = '', sort = DEFAULT_COLLECTION_SORT } = {}) {
const tokens = tokenizeQuery(query);
const comparator = compareBy(normalizeCollectionSort(sort));
return (collections || [])
.map(decorate)
.filter((d) => matchHaystack(d.haystack, tokens))
.sort((a, b) => (a.bucket - b.bucket) || comparator(a, b))
.map((d) => d.record);
}