@gotcos/glasses-server 6.46.1 → 6.47.0

package/server/lib/imported-library-rows.ts

@@ -0,0 +1,616 @@
+/**
+ * Rows for the imported library, and the supersession rule (6.47.0, WS5).
+ *
+ * WHAT SUPERSESSION IS. A merge is ADDITIVE: the Fireflies import, the G2
+ * recordings and the merged record all exist on disk afterwards, because that is
+ * what makes Undo possible (D12). But a person scrolling their meetings must see
+ * ONE row for one meeting, so the LIST drops the inputs a derived record already
+ * holds. Nothing is deleted; the sources stay reachable from the merged record's
+ * detail through its `sources[]`.
+ *
+ * WHY A FILTER RATHER THAN A MERGE RULE. `mergeMeetingSources` dedupes by
+ * sessionId, and 6.46.1 leans on its group ORDER so an operations copy beats the
+ * store copy of the same capture. Teaching it about derived records would have
+ * meant a second precedence rule inside the first. Worse, a split piece has NO
+ * sessionId (it is a span of a recording, not a capture), so two pieces of one
+ * recording would have collapsed onto each other. Supersession therefore runs
+ * BEFORE the merge, as its own pass, and derived rows dedupe on `recordId`.
+ *
+ * COST. Every helper here is free until a derived record exists: the scan reads
+ * the imports root's `.derived.json` sidecars, and when there are none, each
+ * function returns its input untouched without opening a meeting file. That
+ * matters because `probeMeetings` sums day counts across every month on every
+ * morning brief.
+ */
+
+import { existsSync, readdirSync, readFileSync } from 'node:fs'
+import { join } from 'node:path'
+import { dataPath } from './data-dir.js'
+import {
+  cosOperationsMeetingsConfigured,
+  discoverMeetingDomains,
+  findCosOperationsMeetingBySessionId,
+  listCosOperationsMeetingDays,
+  listDirectLibraryMeetingDays,
+  mergedScribeSessions,
+  resolveCosOperationsDir,
+  resolveMeetingLibrary,
+  sidecarSessionId,
+} from './cos-operations-meetings.js'
+import { g2RecordingsReachOperations } from './g2-ops-handoff.js'
+import {
+  type ImportedMeetingLibrary,
+  type ImportedRecordKind,
+  IMPORTED_DOMAIN,
+  IMPORTED_FILENAME_PATTERN,
+  IMPORTED_MONTH_PATTERN,
+  getImportedMeetingLibrary,
+  importHash,
+  importRecordId,
+} from './imported-meeting-library.js'
+import { getMeetingStore, type MeetingMeta, type MeetingStore } from './meeting-store.js'
+
+/** The merged-scribe markers live with the scribe reader; re-exported for callers. */
+export {
+  G2_SESSION_MARKER,
+  MERGE_ACTION_MARKER,
+  mergedScribeActionId,
+  mergedScribeSessions,
+} from './cos-operations-meetings.js'
+
+const DAY_FILE_PREFIX = /^(\d{4}-(?:0[1-9]|1[0-2])-(?:0[1-9]|[12]\d|3[01]))_/
+/** The month folder a day-prefixed filename belongs to. */
+const MONTH_FROM_DAY_FILE = /^(\d{4}-(?:0[1-9]|1[0-2]))-(?:0[1-9]|[12]\d|3[01])_/
+
+/** One derived record's identity and inputs, read from its `.derived.json`. */
+export interface DerivedRecordSummary {
+  recordId: string
+  month: string
+  filename: string
+  kind: 'merge' | 'split'
+  actionId?: string
+  /** G2 captures the record holds, in the sidecar's own order (start order). */
+  g2SessionIds: string[]
+  /** Fireflies meetings the record holds: the primary first, then alternates. */
+  firefliesIds: string[]
+  pieceIndex?: number
+  /** The capture whose content defined a split piece, when it names one. */
+  sourceSessionId?: string
+}
+
+/** Everything a derived record has taken over. Empty until one exists. */
+export interface SupersededInputs {
+  g2Sessions: Set<string>
+  /** `imported:fireflies:<h16>` record ids of the imports a derived record holds. */
+  importRecordIds: Set<string>
+  isEmpty: boolean
+}
+
+function asRecord(value: unknown): Record<string, unknown> | null {
+  return value && typeof value === 'object' && !Array.isArray(value) ? value as Record<string, unknown> : null
+}
+
+function stringField(source: Record<string, unknown>, key: string): string | undefined {
+  const value = source[key]
+  return typeof value === 'string' && value.length > 0 ? value : undefined
+}
+
+/**
+ * Read one `.derived.json` into the fields the surfaces need.
+ *
+ * Tolerant on purpose: a sidecar from a newer version, or one that lost a field,
+ * must degrade to "this record exists" rather than take the whole list down with
+ * it. The one field it cannot do without is `inputs`, because that is what
+ * supersession is computed from.
+ */
+export function summarizeDerivedSidecar(
+  sidecar: unknown,
+  ref: { recordId: string; month: string; filename: string; kind: ImportedRecordKind },
+): DerivedRecordSummary | null {
+  const doc = asRecord(sidecar)
+  if (!doc) return null
+  const inputs = Array.isArray(doc.inputs) ? doc.inputs : []
+  const g2SessionIds: string[] = []
+  const firefliesIds: string[] = []
+  for (const entry of inputs) {
+    const input = asRecord(entry)
+    if (!input) continue
+    const id = stringField(input, 'id')
+    if (!id) continue
+    if (input.kind === 'g2') g2SessionIds.push(id)
+    else if (input.kind === 'fireflies') firefliesIds.push(id)
+  }
+  const pieces = Array.isArray(doc.pieces) ? doc.pieces : []
+  const firstPiece = asRecord(pieces[0])
+  const pieceIndex = typeof firstPiece?.index === 'number' && Number.isInteger(firstPiece.index)
+    ? firstPiece.index
+    : undefined
+  const pieceSessions = Array.isArray(firstPiece?.sessionIds) ? firstPiece.sessionIds : []
+  const sourceSessionId = typeof pieceSessions[0] === 'string' ? pieceSessions[0] as string : undefined
+  const actionId = stringField(doc, 'actionId')
+
+  return {
+    recordId: ref.recordId,
+    month: ref.month,
+    filename: ref.filename,
+    kind: ref.kind === 'piece' ? 'split' : 'merge',
+    ...(actionId ? { actionId } : {}),
+    g2SessionIds,
+    firefliesIds,
+    ...(ref.kind === 'piece' && pieceIndex !== undefined ? { pieceIndex } : {}),
+    ...(ref.kind === 'piece' && sourceSessionId ? { sourceSessionId } : {}),
+  }
+}
+
+/** Derived record filenames on disk, without opening anything. */
+function derivedFilenames(
+  library: ImportedMeetingLibrary,
+): Array<{ month: string; filename: string; kind: ImportedRecordKind; hash: string }> {
+  const found: Array<{ month: string; filename: string; kind: ImportedRecordKind; hash: string }> = []
+  let months: string[]
+  try {
+    months = readdirSync(library.root).filter(name => IMPORTED_MONTH_PATTERN.test(name))
+  } catch {
+    return found
+  }
+  for (const month of months) {
+    let names: string[]
+    try {
+      names = readdirSync(join(library.root, month))
+    } catch {
+      continue
+    }
+    for (const filename of names) {
+      const match = filename.match(IMPORTED_FILENAME_PATTERN)
+      if (!match) continue
+      const kind = match[2] as ImportedRecordKind
+      if (kind === 'fireflies') continue
+      found.push({ month, filename, kind, hash: match[3] })
+    }
+  }
+  return found
+}
+
+/**
+ * Every derived record in the imports library.
+ *
+ * Reads only the derived sidecars — never a markdown file — through the
+ * library's own bounded reader, so a 32 MiB sidecar is capped and an iCloud
+ * conflict copy is refused by the filename pattern before it is opened.
+ */
+export function readDerivedRecords(
+  library: ImportedMeetingLibrary = getImportedMeetingLibrary(),
+): DerivedRecordSummary[] {
+  const records: DerivedRecordSummary[] = []
+  for (const { month, filename, kind, hash } of derivedFilenames(library)) {
+    const summary = summarizeDerivedSidecar(library.readSidecar(month, filename), {
+      recordId: importRecordId(kind, hash),
+      month,
+      filename,
+      kind,
+    })
+    if (summary) records.push(summary)
+  }
+  return records
+}
+
+export function supersededInputsOf(records: readonly DerivedRecordSummary[]): SupersededInputs {
+  const g2Sessions = new Set<string>()
+  const importRecordIds = new Set<string>()
+  for (const record of records) {
+    for (const sessionId of record.g2SessionIds) g2Sessions.add(sessionId)
+    for (const firefliesId of record.firefliesIds) {
+      importRecordIds.add(importRecordId('fireflies', importHash(firefliesId)))
+    }
+  }
+  return { g2Sessions, importRecordIds, isEmpty: g2Sessions.size === 0 && importRecordIds.size === 0 }
+}
+
+/**
+ * Drop the rows a derived record already holds.
+ *
+ * A row is superseded when its session is one a derived record holds, or when it
+ * IS the import a derived record holds. A derived row is never superseded: pieces
+ * of one long recording each stand alone, and grouping (WS3) guarantees one
+ * merged record per capture.
+ */
+export function dropSupersededRows(rows: readonly MeetingMeta[], superseded: SupersededInputs): MeetingMeta[] {
+  if (superseded.isEmpty) return [...rows]
+  return rows.filter(row => {
+    if (row.derivedKind) return true
+    if (row.recordId && superseded.importRecordIds.has(row.recordId)) return false
+    if (row.sessionId && superseded.g2Sessions.has(row.sessionId)) return false
+    return true
+  })
+}
+
+/** Derived rows dedupe on `recordId`, because a split piece has no session. */
+export function dedupeDerivedRows(rows: readonly MeetingMeta[]): MeetingMeta[] {
+  const seen = new Set<string>()
+  const kept: MeetingMeta[] = []
+  for (const row of rows) {
+    const key = row.recordId ?? `${row.month}:${row.filename}`
+    if (seen.has(key)) continue
+    seen.add(key)
+    kept.push(row)
+  }
+  return kept
+}
+
+export function withDerivedIdentity(row: MeetingMeta, record: DerivedRecordSummary | undefined): MeetingMeta {
+  if (!record) return row
+  const identified: MeetingMeta = { ...row }
+  if (record.actionId) identified.actionId = record.actionId
+  if (record.kind === 'merge') {
+    // The earliest capture, so the row that replaced it can still open the
+    // speaker panel. `g2SessionIds` keeps the rest reachable.
+    if (record.g2SessionIds.length > 0) {
+      identified.sessionId = record.g2SessionIds[0]
+      identified.g2SessionIds = [...record.g2SessionIds]
+    }
+  } else {
+    // A piece is a SPAN, not a capture. A sessionId here would make two pieces of
+    // one recording collapse onto each other in mergeMeetingSources.
+    delete identified.sessionId
+    if (record.pieceIndex !== undefined) identified.pieceIndex = record.pieceIndex
+    if (record.sourceSessionId) identified.sourceSessionId = record.sourceSessionId
+  }
+  return identified
+}
+
+/**
+ * The imported library's rows, split into imports and derived records, with the
+ * derived rows carrying their action and their inputs.
+ *
+ * The limit is doubled before it reaches the library because supersession runs
+ * after this: a superseded import and the merged record that replaced it are two
+ * rows on the same date, and asking for exactly `limit` would let the pair push
+ * an older meeting out of a page that ends up holding only one of them.
+ */
+export function listImportedLibraryRows(
+  options: { limit?: number; domain?: string; month?: string; day?: string } = {},
+  library: ImportedMeetingLibrary = getImportedMeetingLibrary(),
+  records: readonly DerivedRecordSummary[] = readDerivedRecords(library),
+): { imports: MeetingMeta[]; derived: MeetingMeta[] } {
+  const byRecordId = new Map(records.map(record => [record.recordId, record]))
+  const imports: MeetingMeta[] = []
+  const derived: MeetingMeta[] = []
+  const rows = library.list({
+    ...options,
+    ...(typeof options.limit === 'number' ? { limit: options.limit * 2 } : {}),
+  })
+  for (const row of rows) {
+    if (row.librarySource === 'imported') {
+      imports.push(row as MeetingMeta)
+      continue
+    }
+    derived.push(withDerivedIdentity(row as MeetingMeta, byRecordId.get(row.recordId)))
+  }
+  return { imports, derived: dedupeDerivedRows(derived) }
+}
+
+/** Month folders of the imported library, newest first. */
+export function importedLibraryMonths(
+  library: ImportedMeetingLibrary = getImportedMeetingLibrary(),
+): string[] {
+  try {
+    return readdirSync(library.root).filter(name => IMPORTED_MONTH_PATTERN.test(name)).sort().reverse()
+  } catch {
+    return []
+  }
+}
+
+/**
+ * A derived record's inputs, each with the record id that opens it.
+ *
+ * This is how "merged automatically" stays honest: the sources are not gone, and
+ * the detail names exactly where each one still lives. A G2 input's id is
+ * resolved against the trees rather than assumed, because the same capture is
+ * `ops:...` on one Mac and `standalone:...` on another, and a link built from the
+ * wrong assumption opens nothing.
+ */
+export function derivedSourcesFor(
+  record: DerivedRecordSummary,
+  store: MeetingStore = getMeetingStore(),
+): NonNullable<MeetingMeta['sources']> {
+  const sources: NonNullable<MeetingMeta['sources']> = []
+  for (const id of record.firefliesIds) {
+    sources.push({ kind: 'fireflies', id, recordId: importRecordId('fireflies', importHash(id)) })
+  }
+  for (const id of record.g2SessionIds) {
+    const recordId = g2RecordIdFor(id, store)
+    sources.push({ kind: 'g2', id, ...(recordId ? { recordId } : {}) })
+  }
+  return sources
+}
+
+function g2RecordIdFor(sessionId: string, store: MeetingStore): string | undefined {
+  if (cosOperationsMeetingsConfigured()) {
+    const operations = findCosOperationsMeetingBySessionId(sessionId)
+    if (operations) return `ops:${operations.domain}:${operations.month}:${operations.filename}`
+  }
+  try {
+    return store.findBySessionId(sessionId) ? `standalone:${sessionId}` : undefined
+  } catch {
+    return undefined
+  }
+}
+
+/** The derived record behind a `blended:` row, or null when there is none. */
+export function findDerivedRecord(
+  recordId: string,
+  library: ImportedMeetingLibrary = getImportedMeetingLibrary(),
+): DerivedRecordSummary | null {
+  if (!/^blended:[0-9a-f]{16}$/.test(recordId)) return null
+  return readDerivedRecords(library).find(record => record.recordId === recordId) ?? null
+}
+
+/**
+ * Sessions the rows in hand already declare they hold.
+ *
+ * Used for the operations tree, where a merged scribe carries its inputs as
+ * `<!-- g2-session -->` markers rather than a `.derived.json`: the list has the
+ * markers in hand, so the filter costs nothing beyond what it already read.
+ */
+export function supersededFromRows(rows: readonly MeetingMeta[]): SupersededInputs {
+  const g2Sessions = new Set<string>()
+  for (const row of rows) {
+    if (!row.derivedKind) continue
+    for (const sessionId of row.g2SessionIds ?? []) g2Sessions.add(sessionId)
+    if (row.sessionId) g2Sessions.add(row.sessionId)
+  }
+  return { g2Sessions, importRecordIds: new Set<string>(), isEmpty: g2Sessions.size === 0 }
+}
+
+/** Day counts from filenames, summed across groups. Shared with the list. */
+export function mergeDayCounts(
+  groups: Array<Array<{ date: string; count: number }>>,
+): Array<{ date: string; count: number }> {
+  const counts = new Map<string, number>()
+  for (const group of groups) {
+    for (const { date, count } of group) {
+      counts.set(date, (counts.get(date) ?? 0) + count)
+    }
+  }
+  return [...counts.entries()]
+    .filter(([, count]) => count > 0)
+    .sort((a, b) => a[0].localeCompare(b[0]))
+    .map(([date, count]) => ({ date, count }))
+}
+
+/**
+ * Day counts for the imported library in one month.
+ *
+ * An import a derived record has taken over is dropped, so the calendar dot and
+ * the list agree. For `all` and `imported` nothing is opened: the record id is a
+ * pure function of the filename's hash. A filter on the meeting's OWN domain has
+ * to read each record's sidecar head, which is what `list()` already does.
+ */
+export function importedLibraryDayCounts(
+  month: string,
+  superseded: SupersededInputs,
+  library: ImportedMeetingLibrary = getImportedMeetingLibrary(),
+  domain = 'all',
+): Array<{ date: string; count: number }> {
+  const counts = new Map<string, number>()
+  if (domain !== 'all' && domain !== IMPORTED_DOMAIN) {
+    for (const row of library.list({ month, domain })) {
+      if (row.librarySource === 'imported' && superseded.importRecordIds.has(row.recordId)) continue
+      counts.set(row.date, (counts.get(row.date) ?? 0) + 1)
+    }
+  } else {
+    let names: string[]
+    try {
+      names = readdirSync(join(library.root, month))
+    } catch {
+      return []
+    }
+    for (const filename of names) {
+      const match = filename.match(IMPORTED_FILENAME_PATTERN)
+      if (!match) continue
+      const kind = match[2] as ImportedRecordKind
+      if (kind === 'fireflies' && superseded.importRecordIds.has(importRecordId('fireflies', match[3]))) continue
+      counts.set(match[1], (counts.get(match[1]) ?? 0) + 1)
+    }
+  }
+  return [...counts.entries()].sort((a, b) => a[0].localeCompare(b[0])).map(([date, count]) => ({ date, count }))
+}
+
+// ── Day counts with supersession ─────────────────────────────────────────────
+//
+// The calendar dot and the list must agree, and the list drops superseded rows,
+// so the count has to drop them too. The existing day-count helpers are FILENAME
+// scans and are uncapped, which is why they can describe a whole month the
+// 50-row list cannot; that property is kept. What is added is a subtraction, and
+// the subtraction is the only part that opens a file — a 4 KB sidecar head per
+// candidate — so it is gated on there being anything to subtract at all.
+
+export type SupersededLayout = 'direct' | 'multi_domain' | 'standalone'
+
+/**
+ * The session a meeting file declares, with the same fallback the LIST uses.
+ *
+ * The operations tree gitignores `.g2-chunks.json`, so a recording whose markdown arrived
+ * through git or iCloud sits there with no sidecar beside it. `listCosOperationsMeetings`
+ * handles that by looking under the server's own recordings root by the same stem; this did
+ * not, so a row the list DROPPED was still counted, and the calendar dot said two where the
+ * list showed one. A count that can disagree with the list it counts is the bug being fixed.
+ */
+function droppedSessionId(directory: string, filename: string, recordingsRoot: string): string | undefined {
+  const direct = sidecarSessionId(directory, filename)
+  if (direct) return direct
+  const month = filename.match(MONTH_FROM_DAY_FILE)?.[1]
+  if (!month) return undefined
+  const fallbackDir = join(recordingsRoot, month)
+  if (!existsSync(fallbackDir)) return undefined
+  return sidecarSessionId(fallbackDir, filename)
+}
+
+/** One `-1` per file in `directories` whose chunk sidecar names a dropped session. */
+function sessionDropCounts(
+  directories: readonly string[],
+  sessions: ReadonlySet<string>,
+  recordingsRoot: string = dataPath('recordings'),
+): Array<{ date: string; count: number }> {
+  if (sessions.size === 0) return []
+  const drops: Array<{ date: string; count: number }> = []
+  for (const directory of directories) {
+    let names: string[]
+    try {
+      names = readdirSync(directory).filter(name => name.endsWith('.md'))
+    } catch {
+      continue
+    }
+    for (const filename of names) {
+      const date = filename.match(DAY_FILE_PREFIX)?.[1]
+      if (!date) continue
+      const sessionId = droppedSessionId(directory, filename, recordingsRoot)
+      if (!sessionId || !sessions.has(sessionId)) continue
+      drops.push({ date, count: -1 })
+    }
+  }
+  return drops
+}
+
+/** Month folders of the operations tree that the domain filter admits. */
+function operationsMonthDirs(month: string, domainFilter: string): string[] {
+  const operationsDir = resolveCosOperationsDir()
+  if (!operationsDir) return []
+  const discovered = discoverMeetingDomains(operationsDir)
+  const domains = domainFilter === 'all'
+    ? discovered
+    : discovered.includes(domainFilter) ? [domainFilter] : []
+  return domains.map(domain => join(operationsDir, domain, 'meetings', month))
+}
+
+
+/**
+ * Sessions declared by merged scribes in the operations tree for one month.
+ *
+ * READS EVERY SCRIBE IN THE MONTH, deliberately. A cheaper gate was tried first
+ * — scan only where the server has written a pipeline decision file — and it was
+ * wrong: the ROW filter reads these markers unconditionally, so a tree holding a
+ * merged scribe with no decision file beside it showed one row while the
+ * calendar dot said two. A count that can disagree with the list it counts is
+ * worse than a count that costs a read, and in this layout the scribe is the
+ * only place the merge is recorded.
+ *
+ * NOT WHEN THE CALLER ALREADY KNOWS. `/api/meetings` has just read every scribe
+ * in this month to build its rows, and each merged one told it which sessions it
+ * holds. Reading the same files again to learn the same thing doubled the cost
+ * of every month request on a pipeline Mac. The caller passes what it declared;
+ * only a caller with no list of its own — `probeMeetings` — pays for the scan.
+ *
+ * The markers sit before `## Transcript`, past the summary, decisions, action
+ * items and attendees, so there is no head-sized prefix that reliably holds
+ * them: when this does read, it reads the file.
+ */
+function operationsMergedSessions(
+  directories: readonly string[],
+  declared: ReadonlySet<string> | undefined,
+): Set<string> {
+  if (declared) return new Set(declared)
+  const sessions = new Set<string>()
+  for (const directory of directories) {
+    let names: string[]
+    try {
+      names = readdirSync(directory).filter(name => name.endsWith('.md'))
+    } catch {
+      continue
+    }
+    for (const filename of names) {
+      let content = ''
+      try {
+        content = readFileSync(join(directory, filename), 'utf8')
+      } catch {
+        continue
+      }
+      for (const sessionId of mergedScribeSessions(content)) sessions.add(sessionId)
+    }
+  }
+  return sessions
+}
+
+/**
+ * Day counts for one month across every source the list shows, with supersession
+ * applied. The list and `probeMeetings` share it so a dot and a row can never
+ * disagree about how many meetings a day holds.
+ */
+export function supersededDayCounts(
+  month: string,
+  layout: SupersededLayout,
+  options: {
+    domain?: string
+    store?: MeetingStore
+    library?: ImportedMeetingLibrary
+    /** Whether a COS pipeline files this Mac's recordings. Read live by default. */
+    pipeline?: boolean
+    /**
+     * Sessions the CALLER's rows already declare they hold.
+     *
+     * `/api/meetings` has just read every scribe in this month; passing what it found here
+     * is the difference between one read per file and two. A caller with no list of its own
+     * omits it and the scan runs.
+     */
+    declaredSessions?: ReadonlySet<string>
+    /** The server's own recordings root, for the sidecar fallback. Tests point it at a temp dir. */
+    recordingsRoot?: string
+  } = {},
+): Array<{ date: string; count: number }> {
+  if (!IMPORTED_MONTH_PATTERN.test(month)) return []
+  const domain = options.domain ?? 'all'
+  const store = options.store ?? getMeetingStore()
+  const library = options.library ?? getImportedMeetingLibrary()
+  const superseded = supersededInputsOf(readDerivedRecords(library))
+
+  if (layout === 'multi_domain' && (options.pipeline ?? g2RecordingsReachOperations())) {
+    // Operations rows only, as the list shows them. The imports root is not read
+    // into this layout at all: on a pipeline Mac the operations tree is the one
+    // library, and a derived record here would be a second row for one meeting.
+    const directories = operationsMonthDirs(month, domain)
+    return mergeDayCounts([
+      listCosOperationsMeetingDays(month, domain),
+      sessionDropCounts(directories, operationsMergedSessions(directories, options.declaredSessions), options.recordingsRoot),
+    ])
+  }
+
+  const groups: Array<Array<{ date: string; count: number }>> = []
+  const sessionDirectories: string[] = []
+
+  if (layout !== 'standalone' && cosOperationsMeetingsConfigured()) {
+    groups.push(listCosOperationsMeetingDays(month, domain))
+    sessionDirectories.push(...operationsMonthDirs(month, domain))
+  }
+  if (layout === 'direct' && (domain === 'all' || domain === 'library')) {
+    groups.push(listDirectLibraryMeetingDays(month))
+    const root = resolveMeetingLibrary().root
+    if (root) sessionDirectories.push(join(root, month))
+  }
+  // DOMAIN-SCOPED. Every other group here honours the filter; this one did not, so a
+  // `domain=quilt` count added every personal recording in the store to it.
+  groups.push(store.listDayCounts(month, domain))
+  sessionDirectories.push(join(store.root, month))
+  groups.push(importedLibraryDayCounts(month, superseded, library, domain))
+
+  if (layout === 'multi_domain') {
+    // 6.46.1: an operations copy of a capture wins over the store copy, and the
+    // day count skips the store row it covers. Unchanged, and now uncapped.
+    const covered = new Set<string>()
+    for (const directory of operationsMonthDirs(month, domain)) {
+      let names: string[]
+      try {
+        names = readdirSync(directory).filter(name => name.endsWith('.md'))
+      } catch {
+        continue
+      }
+      for (const filename of names) {
+        const sessionId = sidecarSessionId(directory, filename)
+        if (sessionId) covered.add(sessionId)
+      }
+    }
+    groups.push(sessionDropCounts([join(store.root, month)], covered, options.recordingsRoot))
+  }
+
+  groups.push(sessionDropCounts(sessionDirectories, superseded.g2Sessions, options.recordingsRoot))
+  return mergeDayCounts(groups)
+}