feat(sync): derive shared-schedule labels for Zoom and Google Calendar from the linked family

[tunile943]

@coderabbitai summary

Description

Second PR of the linked meeting modes work, and the first consumer of PR 1's linkedToMid model. A meeting the group runs as two co-existing weekly schedules now carries one name across both external services — "One Day at a Time - Hybrid Mon-Fri - Zoom Only Sat" — on its shared Zoom meeting and on every family member's Google Calendar event. The label is not Zoom-specific, so it is built once and both services call it, rather than the segment logic living twice.

Behavior-neutral for every existing single-schedule meeting (the almost-all case is asserted byte-for-byte in both services' tests), and immediately correct for the three legacy pairs PR 1 just backfilled.

• frontend/util/meetings/linkedSchedules.ts: the shared label builder and its supporting pieces.
  • buildLinkedScheduleLabel(baseTitle, meeting, family, singleScheduleSuffix?) — pure text, no knowledge of Zoom, Google or zid. Segments follow LINKED_SCHEDULE_MODES' fixed [Hybrid, In Person, Remote] order (never row-creation order, so the name is stable no matter which member triggered the write), each rendered as ${modeLabel} ${formatDayColumn(pattern)} reusing the exports' own Day-column formatter. LINKED_SCHEDULE_MODE_LABEL is the one mode→label map both services now share.
  • resolveFamilyRows(meeting, family) — reconciles the in-flight row with its pre-edit copy still sitting in the DB-loaded family, so a caller may pass either the whole family or only the siblings.
  • isDetachedSplitChild(row) — a "this occurrence" split-off child is a one-off, not a schedule; it is excluded from the label, so one detached child whose mode was later edited can't inject a bogus "- Zoom Only One-time" segment into every sharing row's name.
  • getZoomScheduleFamily(tx, mid, zid) — the rows the family's single Zoom meeting must account for, unioned by mid from two sources that are each incomplete alone: the linkedToMid family (the only way to reach a Zoom-free In-Person member) and every other live row sharing the zid (scoped-edit split children and any legacy zid group the backfill didn't adopt — dropping them would re-narrow the union schedule fixed in Editing one row of a shared Zoom meeting narrows the unioned Zoom recurrence #513).
  • linkedFamilyLoader(tx, mid) — a once-per-request reader caching the in-flight promise (a scoped edit runs its parent and child after() syncs concurrently), so the Zoom write and the Google Calendar writes in one request read the same rows and cannot disagree.
• frontend/services/zoom.ts: zoomTopicFor(meeting, family) keeps its pinned-zoomTopic early return and otherwise delegates to the shared builder with ZOOM_SINGLE_TOPIC_SUFFIX (Hybrid/Remote only — an in-person meeting has no Zoom meeting to name). buildZoomMeetingBody/updateZoomMeeting/createZoomMeeting take the full family instead of "sibling rows sharing this zid", and buildZoomRecurrence/nextOccurrenceStart filter it through isZoomBearing internally — a standalone correctness hardening, since unioning an In-Person member's weekdays into weekly_days would advertise Zoom occurrences for a schedule that never meets online. Nothing writes the derived topic back into Meeting.zoomTopic: a null column keeps meaning "auto, recompute from the current family", which is what makes the topic recomputable rather than pin-once.
• frontend/services/googleCalendar.ts: buildEventTitle(meeting, family) calls the same builder, retiring the private modeTitleSuffix map. buildEventBody/createCalendarEvent/updateCalendarEvent/reconcileMeetingCalendars gain a trailing, defaulted family parameter — trailing because most call sites (resume, suspension, delete rewrites) have no family concept and an extra positional argument they'd all pass [] to is only a chance to get the order wrong. Google has no zoomTopic equivalent, so there is no verbatim-name escape hatch; every member's event gets the full union title, mirroring the one shared Zoom meeting even though each schedule keeps its own event with its own dates and RRULE.
• frontend/app/api/update/meeting/route.ts, update/meeting/sync/route.ts, write/meeting/route.ts: one linkedFamilyLoader per request, replacing the two zid-only findMany calls, and its result feeds both the Zoom write and the calType/Zoom-Room calendar writes in that same request. handleScopedEdit shares a single loader between syncScopedParentCalendar and syncSplitMeeting, which start concurrently.
• frontend/app/api/update/meeting/resume/route.ts, delete/meeting/route.ts, util/meetings/suspension.ts: the paths that rewrite or recreate calendar events without an accompanying edit — resume, pre-created resume series, and a scoped delete's survivor rewrite — now load the family too. Without it, those writes would quietly rename a family member's event back to its own mode alone, and it would stay wrong until the row was next fully edited. createPendingResumeSeries uses the lazy loader so its "no upcoming occurrence" early returns stay genuine no-ops, query included.
• frontend/app/api/retrieve/meeting/[id]/route.ts: zoomScheduleDiverged's inline detached-child filter now calls isDetachedSplitChild, so the divergence signal and the family label can never disagree about what counts as a detached one-off. Same predicate, no behavior change.
• frontend/scripts/zoom-scan.ts: the monthly horizon-extension scan passes the representative's family rather than its raw zid group, so the scan can never rename a linked family's Zoom meeting back to a single schedule's name.
• frontend/tests/unit/linkedSchedules.test.ts, zoom.test.ts, googleCalendar.test.ts: 29 new tests across the three files — fixed segment order regardless of family order, all three family shapes (Hybrid+Remote / Hybrid+In Person / In Person+Remote), Daily collapsing, the in-flight-vs-stored row reconciliation, detached split children excluded but recurring tail splits kept, the In-Person exclusion from Zoom's recurrence union, pinned zoomTopic short-circuit, and single-schedule output asserted byte-identical to today in both services. Integration suites' Zoom/Calendar mocks and fixtures updated for the new signatures.
• docs/02-handoff/technical-decisions.md: the adopted-legacy-Zoom bullet now describes the derived topic as coming from the linked family, and records that the Zoom topic and calendar title are the same string by construction.

Known gaps, deliberately deferred to PR 3

Both are carried forward from this branch's review and marked TODO(linked-schedules PR3) at buildEventTitle:

• Calendar fan-out. The union title is recomputed only for the row being written, so adding or removing a family member leaves the other members' calendar events on their previous title until each is itself edited or retry-synced. PR 3's after(...) must republish every member's events alongside its family-wide updateZoomMeeting.
• Title de-sync between members. The label uses each row's own title as its base, so two members' names agree only while their title columns do. PR 3 derives the linked row's title from the anchor and decides there whether a direct edit of a linked row's title is rejected, propagated, or accepted.

Testing

• cd frontend && yarn lint — clean.
• cd frontend && yarn typecheck — clean (this is what catches every un-updated call site of the six changed signatures).
• cd frontend && yarn test:unit — green, including the 29 new label/topic/title tests.
• cd frontend && yarn test:integration — green; the route suites exercise the new family lookups against a real Postgres.
• Reviewed against the pre-existing suites for assertions on the old contract (pr_convention.md sweep): tests/unit/zoom.test.ts's suffix-only topic expectations, write-meeting-route.test.ts's createZoomMeeting call assertions, and the updateZoomMeeting sibling-array assertions in update-meeting-route.test.ts / update-meeting-sync-route.test.ts all updated to the family signature rather than left asserting the old one.
• Reviewer check: git diff feat/linked-schedules-model...feat/linked-zoom-topic -- frontend/services — confirm no code path writes a derived topic back into Meeting.zoomTopic, which is what keeps NULL meaning "auto".

Area(s) Touched

Product areas

• auth — sign-in, sessions, NextAuth, role-based access
• admin — /admin shell (Diagnostics, Users, Import, Export tabs)
• docs — /docs Resources page (rendering, navigation, search)
• ui/ux — frontend layout/styling not tied to a specific integration

Integrations

• google-calendar — Google Calendar sync
• zoom-api — Zoom meeting/host sync

Engineering

• security — auth hardening, data exposure, dependency/security scanning
• testing — test suite (Jest/Playwright) changes
• github-actions — workflows, Dependabot config, other .github/ CI tooling
• cleanup — refactor or dead-code removal, no behavior change
• documentation — docs/ or README changes only

Pre-merge Checklist

• Code follows current formatting conventions and passes lint (yarn lint).
• Comments are appropriate — only where genuinely non-obvious, not restating what the code already says.
• All test suites pass, both run locally and green in GitHub CI. Do not merge if any test suite is failing.
• A test suite was added or updated for the feature/fix in this PR. Double-check this if you change a feature and did not update the test suites.
• Only files relevant to this change are committed — no stray or unrelated files.
• If this branch has a merge conflict with master, master was merged into this branch first (not the other way around).
• The rest of this PR description (Description, Testing) is filled out, not left as template placeholders.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
ithaca-recovery	Ready	Preview	Aug 23, 2026 5:52pm

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 7a4458e4-1b02-4b81-ae32-af986b3f6fa2

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}