Pins are server-owned, so they survive paging and follow you between apps

[OutThisLife]

A pin means "keep this reachable." Pins were stored as localStorage strings, which made that promise unenforceable in two directions: the sidebar could only resolve a pin against the rows it happened to have loaded, and a second app on the same gateway had its own separate copy. Ten open PRs are circling one or the other half of that.

The sessions.pinned column already exists and the desktop already mirrors pins into it. What was missing is that nothing ever read it back. This makes the server row authoritative: the list endpoints back-fill pinned conversations past their LIMIT, so a pinned row is always in the page, and the desktop reconciles against it in both directions.

Once the back-fill is in place the read-back needs no new endpoint — the rows the sidebar already fetches carry the flag.

What changes

include_pinned on list_sessions_rich — one bounded extra query for pinned rows the window missed, reusing the page's own WHERE clause. An archived or filtered-out conversation stays out; a pin is not a filter bypass. It runs before compression projection, so a back-filled root surfaces under its live tip like any other row.

The three list endpoints request the back-fill and expose pinned as a JSON boolean. Both merge paths re-window after sorting, so pinned rows are explicitly kept through the cap rather than re-dropped on recency, and the sidebar's "load more" signal discounts them so they can't fake a full page.

pullRemotePins adopts pins made elsewhere and drops local pins the server has released, keyed on the durable lineage root. A write in flight is guarded for the lifetime of its own request, so a list response that predates the PATCH can't silently undo the pin the user just made. A runtime without the flag sends no pinned, which is treated as no opinion.

Verification

Reproduced the fail state first — pin the oldest of 30 sessions, request a 10-row page, confirm it is absent — then confirmed the flag brings it back without disturbing the page. Covered: filters still apply to a back-filled pin, no duplicate when the pin is already on the page, compression roots project to their tip, cross-app adopt/drop, and the stale-page race.

tests/test_hermes_state.py 483 passed. Desktop store + sidebar/session suites 1040 passed. tsc --noEmit clean.

Supersedes

Same root cause, one fix:

• Supersedes feat(desktop): server-synced session pins with live cross-device sync #62716 — server-owned pins with cross-device sync; this is that design, with the paging half it was missing.
• Supersedes feat: sync pinned sessions across multiple GUI apps (#72948) #72953 — the read-back. Its pinned-ids endpoint isn't needed once pinned rows are guaranteed in the page.
• Supersedes fix(desktop): force-include pinned sessions in session list via include_ids #43167 — include_ids. Same insight; the server already knows which rows are pinned, so the client needn't send the list.
• Supersedes fix(desktop): persist pinned sessions across updates #65620, fix: persist pinned sessions to state.db so they survive upgrades #60925, feat(desktop): add file-backed session pins bridge #54299 — parallel pin stores, when state.db already has the column.
• Supersedes fix(desktop): resync pinned sessions from store #60019, fix(desktop): fetch missing pinned sessions on sidebar refresh (#51685) #51700, fix(desktop): restore pinned sessions after restart #39360, fix(desktop): render pinned search sessions #37590 — fetch-the-missing-row patches; there is no missing row now.

Closes #72948

Co-authored-by: Kyzcreig 9063726+Kyzcreig@users.noreply.github.com
Co-authored-by: webtecnica 75556242+webtecnica@users.noreply.github.com
Co-authored-by: hrnbld 260600092+hrnbld@users.noreply.github.com
Co-authored-by: aman-merchant 274313970+aman-merchant@users.noreply.github.com
Co-authored-by: eason2026 209090628+eason2026@users.noreply.github.com
Co-authored-by: konsisumer 11262660+konsisumer@users.noreply.github.com
Co-authored-by: rerdi92 76791321+rerdi92@users.noreply.github.com
Co-authored-by: liuhao1024 11816344+liuhao1024@users.noreply.github.com
Co-authored-by: Tamaz-sujashvili 56168197+Tamaz-sujashvili@users.noreply.github.com
Co-authored-by: ferminquant 14808645+ferminquant@users.noreply.github.com

[github-actions]

૮ >ﻌ< ა ci review

_{ran on 8ce8b70}

ℹ️ Info

Desktop E2E visual evidence · View test artifacts · View job

1 visual diff.

_{inline evidence is publishing...}

[andrejsrna]

nice one ty!