merge: sync upstream through 9c7622dac (11 commits)

[omegent-app]

Adopts upstream through 9c7622dac — the 11 commits after 0a7c662d3.

Important

Merge with a merge commit, not squash. Squashing collapses the second parent and discards
upstream's commits as ancestors — the thing that keeps "commits behind upstream" honest.

Upstream commits entering the product

Commit
9c7622dac	fix(server): handle unborn HEAD in VCS status (pingdotgg#5944)
c14bcca10	fix(server): advertise 256-color TERM on Windows terminals (pingdotgg#5693)
1fa315ea9	fix(server): skip Linux libc detection on Windows/macOS (pingdotgg#5354)
f9730979c	perf(web): skip base64 for oversized image candidates (pingdotgg#5220)
59d0c922e	fix(web): persist sidebar shelf collapse state (pingdotgg#5136)
2ea51bd31	fix(server): OpenCode model parsing drops models with a slash in the JSON body (pingdotgg#5072)
e5c82d79a	fix(mobile): keep chat composer above the Android gesture bar (pingdotgg#5988)
9afef94a6	feat: add three-hour snooze option (pingdotgg#5914)
a7b0366cb	fix(web): account for Windows window controls in PR page header (pingdotgg#6049)
428d9f9f6	Allow Android tablets to rotate (pingdotgg#5613)
1b120f352	fix(ci): extend release publish timeout (pingdotgg#6034)

Resolutions

A quiet batch — three conflicts, one of which needed judgement.

Where	Situation	Resolution
Sidebar.tsx	pingdotgg#5136 persists both shelves. The fork already persisted the settled shelf under its own key (…settled-shelf-expanded:v1, default true), so upstream's declaration would have been a second settledShelfExpanded — a duplicate the build would have rejected. Upstream also re-declared SETTLED_TAIL_* locally, which the fork imports from Sidebar.logic	fork's settled persistence kept (identical default; switching keys would silently reset every existing preference), upstream's snoozed-shelf persistence adopted since the fork lacked it, and the duplicate SETTLED_TAIL_* dropped. Also removed a duplicate useLocalStorage import — both parents import the same module under different specifiers (../hooks/… vs ~/hooks/…)
release.yml	upstream runs on Blacksmith andpingdotgg#6034 extends the publish timeout	fork's ubuntu-24.04 kept, upstream's timeout-minutes: 30 taken — the timeout is what that commit was for, and it matters more on a slower GitHub-hosted runner, not less
ThreadDetailScreen.tsx	fork's StyleSheet vs upstream's useKeyboardState	unioned

Checks I ran rather than assumed

• SNOOZED_SHELF_EXPANDED_KEY is consumed, not an orphan constant — declared at Sidebar.tsx:256, used at :2286. Exactly one settledShelfExpanded state remains.
• No Blacksmith references anywhere in .github/workflows. This fork has been bitten twice by upstream workflows arriving on runners it does not have (web-preview.yml, mobile-fingerprint-check.yml), both times in files that merged without a conflict — so this is now a standing check rather than a reaction.
• feat: add three-hour snooze option pingdotgg/t3code#5914's three-hour snooze reaches every fork surface. It lands in the shared resolveSnoozePresets in client-runtime, and the web sidebar, board and both mobile thread lists all call it, so there is no per-surface parity gap to fix.

Adversarial review

Both reviewers went through all three resolutions and the pingdotgg#5914 coverage question. Neither found a
HIGH or a MEDIUM — gpt-5.6-sol reported "no HIGH, MEDIUM, or LOW merge-weld defects". Two LOWs
between them:

Finding	Disposition
both, LOW — the comment above SNOOZED_SHELF_EXPANDED_KEY claims the key is kept so existing preferences survive a rename. True of the fork's settled key; this merge introduces snoozed persistence, so there is nothing to migrate	Fixed in 01a4cee03. It came across with upstream's constant and was simply inaccurate in the fork's context — and a comment describing a migration nobody is doing is exactly what a future merge reasons from
grok-4.5, LOW — the 30-minute publish timeout is inherited from a Blacksmith-timed job; worth re-tuning empirically once a fork release actually runs on the slower GitHub-hosted runner	Acknowledged, not changed. Guessing a number now is not better than upstream's; worth watching the first release

Both independently confirmed the parts I most wanted checked:

• Snoozed persistence is end-to-end — key → useLocalStorage → toggle → aria-expanded and render — and exactly onesettledShelfExpanded state survives, on the fork's key.
• Dropping upstream's settled key loses nothing: same Boolean schema, same true default, and no upstream migration or reset mechanism was skipped. Keeping the fork's key is what preserves existing preferences.
• SETTLED_TAIL_* now comes only from Sidebar.logic, with no competing local declaration, and the surviving useLocalStorage import resolves to the same hook.
• fix(mobile): keep chat composer above the Android gesture bar pingdotgg/t3code#5988 landed whole on mobile, not half-applied: keyboard visibility rather than composer focus drives the bottom safe-area inset, and that inset reaches both the overlap calculation and ThreadComposer. Both new imports are genuinely used.
• feat: add three-hour snooze option pingdotgg/t3code#5914 reaches all six snooze surfaces — sidebar row, sidebar context, board context, the shared thread action menu, and both mobile paths — because they all build from the shared resolveSnoozePresets. No surface carries a hardcoded preset list that would silently omit it.
• release.yml was the only workflow touched, and there are no remaining blacksmith-* references anywhere.

One out-of-scope observation worth recording: the mobile snoozed shelf still holds its expanded
state in useState(false), so it does not persist. pingdotgg#5136 never touched mobile, so this is an
upstream parity gap rather than anything this merge caused — a candidate for a follow-up, not for a
sync PR.

Verification

• Recursive typecheck clean across all 18 packages.
• vp build of the web app passes.
• 2,755 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• pnpm-lock.yaml regenerated rather than hand-merged.

Co-authored by @patroza

opened by Patrick Roza in chat thread Discord · Discord · T3

[omegent-app]

Adopts upstream through 9c7622dac — the 11 commits after 0a7c662d3.

Important

Merge with a merge commit, not squash. Squashing collapses the second parent and discards
upstream's commits as ancestors — the thing that keeps "commits behind upstream" honest.

Upstream commits entering the product

Commit
9c7622dac	fix(server): handle unborn HEAD in VCS status (pingdotgg#5944)
c14bcca10	fix(server): advertise 256-color TERM on Windows terminals (pingdotgg#5693)
1fa315ea9	fix(server): skip Linux libc detection on Windows/macOS (pingdotgg#5354)
f9730979c	perf(web): skip base64 for oversized image candidates (pingdotgg#5220)
59d0c922e	fix(web): persist sidebar shelf collapse state (pingdotgg#5136)
2ea51bd31	fix(server): OpenCode model parsing drops models with a slash in the JSON body (pingdotgg#5072)
e5c82d79a	fix(mobile): keep chat composer above the Android gesture bar (pingdotgg#5988)
9afef94a6	feat: add three-hour snooze option (pingdotgg#5914)
a7b0366cb	fix(web): account for Windows window controls in PR page header (pingdotgg#6049)
428d9f9f6	Allow Android tablets to rotate (pingdotgg#5613)
1b120f352	fix(ci): extend release publish timeout (pingdotgg#6034)

Resolutions

A quiet batch — three conflicts, one of which needed judgement.

Where	Situation	Resolution
Sidebar.tsx	pingdotgg#5136 persists both shelves. The fork already persisted the settled shelf under its own key (…settled-shelf-expanded:v1, default true), so upstream's declaration would have been a second settledShelfExpanded — a duplicate the build would have rejected. Upstream also re-declared SETTLED_TAIL_* locally, which the fork imports from Sidebar.logic	fork's settled persistence kept (identical default; switching keys would silently reset every existing preference), upstream's snoozed-shelf persistence adopted since the fork lacked it, and the duplicate SETTLED_TAIL_* dropped. Also removed a duplicate useLocalStorage import — both parents import the same module under different specifiers (../hooks/… vs ~/hooks/…)
release.yml	upstream runs on Blacksmith andpingdotgg#6034 extends the publish timeout	fork's ubuntu-24.04 kept, upstream's timeout-minutes: 30 taken — the timeout is what that commit was for, and it matters more on a slower GitHub-hosted runner, not less
ThreadDetailScreen.tsx	fork's StyleSheet vs upstream's useKeyboardState	unioned

Checks I ran rather than assumed

• SNOOZED_SHELF_EXPANDED_KEY is consumed, not an orphan constant — declared at Sidebar.tsx:256, used at :2286. Exactly one settledShelfExpanded state remains.
• No Blacksmith references anywhere in .github/workflows. This fork has been bitten twice by upstream workflows arriving on runners it does not have (web-preview.yml, mobile-fingerprint-check.yml), both times in files that merged without a conflict — so this is now a standing check rather than a reaction.
• feat: add three-hour snooze option pingdotgg/t3code#5914's three-hour snooze reaches every fork surface. It lands in the shared resolveSnoozePresets in client-runtime, and the web sidebar, board and both mobile thread lists all call it, so there is no per-surface parity gap to fix.

Adversarial review

Both reviewers went through all three resolutions and the pingdotgg#5914 coverage question. Neither found a
HIGH or a MEDIUM — gpt-5.6-sol reported "no HIGH, MEDIUM, or LOW merge-weld defects". Two LOWs
between them:

Finding	Disposition
both, LOW — the comment above SNOOZED_SHELF_EXPANDED_KEY claims the key is kept so existing preferences survive a rename. True of the fork's settled key; this merge introduces snoozed persistence, so there is nothing to migrate	Fixed in 01a4cee03. It came across with upstream's constant and was simply inaccurate in the fork's context — and a comment describing a migration nobody is doing is exactly what a future merge reasons from
grok-4.5, LOW — the 30-minute publish timeout is inherited from a Blacksmith-timed job; worth re-tuning empirically once a fork release actually runs on the slower GitHub-hosted runner	Acknowledged, not changed. Guessing a number now is not better than upstream's; worth watching the first release

Both independently confirmed the parts I most wanted checked:

• Snoozed persistence is end-to-end — key → useLocalStorage → toggle → aria-expanded and render — and exactly onesettledShelfExpanded state survives, on the fork's key.
• Dropping upstream's settled key loses nothing: same Boolean schema, same true default, and no upstream migration or reset mechanism was skipped. Keeping the fork's key is what preserves existing preferences.
• SETTLED_TAIL_* now comes only from Sidebar.logic, with no competing local declaration, and the surviving useLocalStorage import resolves to the same hook.
• fix(mobile): keep chat composer above the Android gesture bar pingdotgg/t3code#5988 landed whole on mobile, not half-applied: keyboard visibility rather than composer focus drives the bottom safe-area inset, and that inset reaches both the overlap calculation and ThreadComposer. Both new imports are genuinely used.
• feat: add three-hour snooze option pingdotgg/t3code#5914 reaches all six snooze surfaces — sidebar row, sidebar context, board context, the shared thread action menu, and both mobile paths — because they all build from the shared resolveSnoozePresets. No surface carries a hardcoded preset list that would silently omit it.
• release.yml was the only workflow touched, and there are no remaining blacksmith-* references anywhere.

One out-of-scope observation worth recording: the mobile snoozed shelf still holds its expanded
state in useState(false), so it does not persist. pingdotgg#5136 never touched mobile, so this is an
upstream parity gap rather than anything this merge caused — a candidate for a follow-up, not for a
sync PR.

Verification

• Recursive typecheck clean across all 18 packages.
• vp build of the web app passes.
• 2,755 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• pnpm-lock.yaml regenerated rather than hand-merged.

Co-authored by @patroza

opened by Patrick Roza in chat thread Discord · Discord · T3

[omegent-app]

Adopts upstream through 9c7622dac — the 11 commits after 0a7c662d3.

Important

Merge with a merge commit, not squash. Squashing collapses the second parent and discards
upstream's commits as ancestors — the thing that keeps "commits behind upstream" honest.

Upstream commits entering the product

Commit
9c7622dac	fix(server): handle unborn HEAD in VCS status (pingdotgg#5944)
c14bcca10	fix(server): advertise 256-color TERM on Windows terminals (pingdotgg#5693)
1fa315ea9	fix(server): skip Linux libc detection on Windows/macOS (pingdotgg#5354)
f9730979c	perf(web): skip base64 for oversized image candidates (pingdotgg#5220)
59d0c922e	fix(web): persist sidebar shelf collapse state (pingdotgg#5136)
2ea51bd31	fix(server): OpenCode model parsing drops models with a slash in the JSON body (pingdotgg#5072)
e5c82d79a	fix(mobile): keep chat composer above the Android gesture bar (pingdotgg#5988)
9afef94a6	feat: add three-hour snooze option (pingdotgg#5914)
a7b0366cb	fix(web): account for Windows window controls in PR page header (pingdotgg#6049)
428d9f9f6	Allow Android tablets to rotate (pingdotgg#5613)
1b120f352	fix(ci): extend release publish timeout (pingdotgg#6034)

Resolutions

A quiet batch — three conflicts, one of which needed judgement.

Where	Situation	Resolution
Sidebar.tsx	pingdotgg#5136 persists both shelves. The fork already persisted the settled shelf under its own key (…settled-shelf-expanded:v1, default true), so upstream's declaration would have been a second settledShelfExpanded — a duplicate the build would have rejected. Upstream also re-declared SETTLED_TAIL_* locally, which the fork imports from Sidebar.logic	fork's settled persistence kept (identical default; switching keys would silently reset every existing preference), upstream's snoozed-shelf persistence adopted since the fork lacked it, and the duplicate SETTLED_TAIL_* dropped. Also removed a duplicate useLocalStorage import — both parents import the same module under different specifiers (../hooks/… vs ~/hooks/…)
release.yml	upstream runs on Blacksmith andpingdotgg#6034 extends the publish timeout	fork's ubuntu-24.04 kept, upstream's timeout-minutes: 30 taken — the timeout is what that commit was for, and it matters more on a slower GitHub-hosted runner, not less
ThreadDetailScreen.tsx	fork's StyleSheet vs upstream's useKeyboardState	unioned

Checks I ran rather than assumed

• SNOOZED_SHELF_EXPANDED_KEY is consumed, not an orphan constant — declared at Sidebar.tsx:256, used at :2286. Exactly one settledShelfExpanded state remains.
• No Blacksmith references anywhere in .github/workflows. This fork has been bitten twice by upstream workflows arriving on runners it does not have (web-preview.yml, mobile-fingerprint-check.yml), both times in files that merged without a conflict — so this is now a standing check rather than a reaction.
• feat: add three-hour snooze option pingdotgg/t3code#5914's three-hour snooze reaches every fork surface. It lands in the shared resolveSnoozePresets in client-runtime, and the web sidebar, board and both mobile thread lists all call it, so there is no per-surface parity gap to fix.

Adversarial review

Both reviewers went through all three resolutions and the pingdotgg#5914 coverage question. Neither found a
HIGH or a MEDIUM — gpt-5.6-sol reported "no HIGH, MEDIUM, or LOW merge-weld defects". Two LOWs
between them:

Finding	Disposition
both, LOW — the comment above SNOOZED_SHELF_EXPANDED_KEY claims the key is kept so existing preferences survive a rename. True of the fork's settled key; this merge introduces snoozed persistence, so there is nothing to migrate	Fixed in 01a4cee03. It came across with upstream's constant and was simply inaccurate in the fork's context — and a comment describing a migration nobody is doing is exactly what a future merge reasons from
grok-4.5, LOW — the 30-minute publish timeout is inherited from a Blacksmith-timed job; worth re-tuning empirically once a fork release actually runs on the slower GitHub-hosted runner	Acknowledged, not changed. Guessing a number now is not better than upstream's; worth watching the first release

Both independently confirmed the parts I most wanted checked:

• Snoozed persistence is end-to-end — key → useLocalStorage → toggle → aria-expanded and render — and exactly onesettledShelfExpanded state survives, on the fork's key.
• Dropping upstream's settled key loses nothing: same Boolean schema, same true default, and no upstream migration or reset mechanism was skipped. Keeping the fork's key is what preserves existing preferences.
• SETTLED_TAIL_* now comes only from Sidebar.logic, with no competing local declaration, and the surviving useLocalStorage import resolves to the same hook.
• fix(mobile): keep chat composer above the Android gesture bar pingdotgg/t3code#5988 landed whole on mobile, not half-applied: keyboard visibility rather than composer focus drives the bottom safe-area inset, and that inset reaches both the overlap calculation and ThreadComposer. Both new imports are genuinely used.
• feat: add three-hour snooze option pingdotgg/t3code#5914 reaches all six snooze surfaces — sidebar row, sidebar context, board context, the shared thread action menu, and both mobile paths — because they all build from the shared resolveSnoozePresets. No surface carries a hardcoded preset list that would silently omit it.
• release.yml was the only workflow touched, and there are no remaining blacksmith-* references anywhere.

One out-of-scope observation worth recording: the mobile snoozed shelf still holds its expanded
state in useState(false), so it does not persist. pingdotgg#5136 never touched mobile, so this is an
upstream parity gap rather than anything this merge caused — a candidate for a follow-up, not for a
sync PR.

Verification

• Recursive typecheck clean across all 18 packages.
• vp build of the web app passes.
• 2,755 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• pnpm-lock.yaml regenerated rather than hand-merged.

Co-authored by @patroza

opened by Patrick Roza in chat thread Discord · Discord · T3

[omegent-app]

Adopts upstream through 9c7622dac — the 11 commits after 0a7c662d3.

Important

Merge with a merge commit, not squash. Squashing collapses the second parent and discards
upstream's commits as ancestors — the thing that keeps "commits behind upstream" honest.

Upstream commits entering the product

Commit
9c7622dac	fix(server): handle unborn HEAD in VCS status (pingdotgg#5944)
c14bcca10	fix(server): advertise 256-color TERM on Windows terminals (pingdotgg#5693)
1fa315ea9	fix(server): skip Linux libc detection on Windows/macOS (pingdotgg#5354)
f9730979c	perf(web): skip base64 for oversized image candidates (pingdotgg#5220)
59d0c922e	fix(web): persist sidebar shelf collapse state (pingdotgg#5136)
2ea51bd31	fix(server): OpenCode model parsing drops models with a slash in the JSON body (pingdotgg#5072)
e5c82d79a	fix(mobile): keep chat composer above the Android gesture bar (pingdotgg#5988)
9afef94a6	feat: add three-hour snooze option (pingdotgg#5914)
a7b0366cb	fix(web): account for Windows window controls in PR page header (pingdotgg#6049)
428d9f9f6	Allow Android tablets to rotate (pingdotgg#5613)
1b120f352	fix(ci): extend release publish timeout (pingdotgg#6034)

Resolutions

A quiet batch — three conflicts, one of which needed judgement.

Where	Situation	Resolution
Sidebar.tsx	pingdotgg#5136 persists both shelves. The fork already persisted the settled shelf under its own key (…settled-shelf-expanded:v1, default true), so upstream's declaration would have been a second settledShelfExpanded — a duplicate the build would have rejected. Upstream also re-declared SETTLED_TAIL_* locally, which the fork imports from Sidebar.logic	fork's settled persistence kept (identical default; switching keys would silently reset every existing preference), upstream's snoozed-shelf persistence adopted since the fork lacked it, and the duplicate SETTLED_TAIL_* dropped. Also removed a duplicate useLocalStorage import — both parents import the same module under different specifiers (../hooks/… vs ~/hooks/…)
release.yml	upstream runs on Blacksmith andpingdotgg#6034 extends the publish timeout	fork's ubuntu-24.04 kept, upstream's timeout-minutes: 30 taken — the timeout is what that commit was for, and it matters more on a slower GitHub-hosted runner, not less
ThreadDetailScreen.tsx	fork's StyleSheet vs upstream's useKeyboardState	unioned

Checks I ran rather than assumed

• SNOOZED_SHELF_EXPANDED_KEY is consumed, not an orphan constant — declared at Sidebar.tsx:256, used at :2286. Exactly one settledShelfExpanded state remains.
• No Blacksmith references anywhere in .github/workflows. This fork has been bitten twice by upstream workflows arriving on runners it does not have (web-preview.yml, mobile-fingerprint-check.yml), both times in files that merged without a conflict — so this is now a standing check rather than a reaction.
• feat: add three-hour snooze option pingdotgg/t3code#5914's three-hour snooze reaches every fork surface. It lands in the shared resolveSnoozePresets in client-runtime, and the web sidebar, board and both mobile thread lists all call it, so there is no per-surface parity gap to fix.

Adversarial review

Both reviewers went through all three resolutions and the pingdotgg#5914 coverage question. Neither found a
HIGH or a MEDIUM — gpt-5.6-sol reported "no HIGH, MEDIUM, or LOW merge-weld defects". Two LOWs
between them:

Finding	Disposition
both, LOW — the comment above SNOOZED_SHELF_EXPANDED_KEY claims the key is kept so existing preferences survive a rename. True of the fork's settled key; this merge introduces snoozed persistence, so there is nothing to migrate	Fixed in 01a4cee03. It came across with upstream's constant and was simply inaccurate in the fork's context — and a comment describing a migration nobody is doing is exactly what a future merge reasons from
grok-4.5, LOW — the 30-minute publish timeout is inherited from a Blacksmith-timed job; worth re-tuning empirically once a fork release actually runs on the slower GitHub-hosted runner	Acknowledged, not changed. Guessing a number now is not better than upstream's; worth watching the first release

Both independently confirmed the parts I most wanted checked:

• Snoozed persistence is end-to-end — key → useLocalStorage → toggle → aria-expanded and render — and exactly onesettledShelfExpanded state survives, on the fork's key.
• Dropping upstream's settled key loses nothing: same Boolean schema, same true default, and no upstream migration or reset mechanism was skipped. Keeping the fork's key is what preserves existing preferences.
• SETTLED_TAIL_* now comes only from Sidebar.logic, with no competing local declaration, and the surviving useLocalStorage import resolves to the same hook.
• fix(mobile): keep chat composer above the Android gesture bar pingdotgg/t3code#5988 landed whole on mobile, not half-applied: keyboard visibility rather than composer focus drives the bottom safe-area inset, and that inset reaches both the overlap calculation and ThreadComposer. Both new imports are genuinely used.
• feat: add three-hour snooze option pingdotgg/t3code#5914 reaches all six snooze surfaces — sidebar row, sidebar context, board context, the shared thread action menu, and both mobile paths — because they all build from the shared resolveSnoozePresets. No surface carries a hardcoded preset list that would silently omit it.
• release.yml was the only workflow touched, and there are no remaining blacksmith-* references anywhere.

One out-of-scope observation worth recording: the mobile snoozed shelf still holds its expanded
state in useState(false), so it does not persist. pingdotgg#5136 never touched mobile, so this is an
upstream parity gap rather than anything this merge caused — a candidate for a follow-up, not for a
sync PR.

Verification

• Recursive typecheck clean across all 18 packages.
• vp build of the web app passes.
• 2,755 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• pnpm-lock.yaml regenerated rather than hand-merged.

Co-authored by @patroza

opened by Patrick Roza in chat thread Discord · Discord · T3

[omegent-app]

Adopts upstream through 9c7622dac — the 11 commits after 0a7c662d3.

Important

Merge with a merge commit, not squash. Squashing collapses the second parent and discards
upstream's commits as ancestors — the thing that keeps "commits behind upstream" honest.

Upstream commits entering the product

Commit
9c7622dac	fix(server): handle unborn HEAD in VCS status (pingdotgg#5944)
c14bcca10	fix(server): advertise 256-color TERM on Windows terminals (pingdotgg#5693)
1fa315ea9	fix(server): skip Linux libc detection on Windows/macOS (pingdotgg#5354)
f9730979c	perf(web): skip base64 for oversized image candidates (pingdotgg#5220)
59d0c922e	fix(web): persist sidebar shelf collapse state (pingdotgg#5136)
2ea51bd31	fix(server): OpenCode model parsing drops models with a slash in the JSON body (pingdotgg#5072)
e5c82d79a	fix(mobile): keep chat composer above the Android gesture bar (pingdotgg#5988)
9afef94a6	feat: add three-hour snooze option (pingdotgg#5914)
a7b0366cb	fix(web): account for Windows window controls in PR page header (pingdotgg#6049)
428d9f9f6	Allow Android tablets to rotate (pingdotgg#5613)
1b120f352	fix(ci): extend release publish timeout (pingdotgg#6034)

Resolutions

A quiet batch — three conflicts, one of which needed judgement.

Where	Situation	Resolution
Sidebar.tsx	pingdotgg#5136 persists both shelves. The fork already persisted the settled shelf under its own key (…settled-shelf-expanded:v1, default true), so upstream's declaration would have been a second settledShelfExpanded — a duplicate the build would have rejected. Upstream also re-declared SETTLED_TAIL_* locally, which the fork imports from Sidebar.logic	fork's settled persistence kept (identical default; switching keys would silently reset every existing preference), upstream's snoozed-shelf persistence adopted since the fork lacked it, and the duplicate SETTLED_TAIL_* dropped. Also removed a duplicate useLocalStorage import — both parents import the same module under different specifiers (../hooks/… vs ~/hooks/…)
release.yml	upstream runs on Blacksmith andpingdotgg#6034 extends the publish timeout	fork's ubuntu-24.04 kept, upstream's timeout-minutes: 30 taken — the timeout is what that commit was for, and it matters more on a slower GitHub-hosted runner, not less
ThreadDetailScreen.tsx	fork's StyleSheet vs upstream's useKeyboardState	unioned

Checks I ran rather than assumed

• SNOOZED_SHELF_EXPANDED_KEY is consumed, not an orphan constant — declared at Sidebar.tsx:256, used at :2286. Exactly one settledShelfExpanded state remains.
• No Blacksmith references anywhere in .github/workflows. This fork has been bitten twice by upstream workflows arriving on runners it does not have (web-preview.yml, mobile-fingerprint-check.yml), both times in files that merged without a conflict — so this is now a standing check rather than a reaction.
• feat: add three-hour snooze option pingdotgg/t3code#5914's three-hour snooze reaches every fork surface. It lands in the shared resolveSnoozePresets in client-runtime, and the web sidebar, board and both mobile thread lists all call it, so there is no per-surface parity gap to fix.

Adversarial review

Both reviewers went through all three resolutions and the pingdotgg#5914 coverage question. Neither found a
HIGH or a MEDIUM — gpt-5.6-sol reported "no HIGH, MEDIUM, or LOW merge-weld defects". Two LOWs
between them:

Finding	Disposition
both, LOW — the comment above SNOOZED_SHELF_EXPANDED_KEY claims the key is kept so existing preferences survive a rename. True of the fork's settled key; this merge introduces snoozed persistence, so there is nothing to migrate	Fixed in 01a4cee03. It came across with upstream's constant and was simply inaccurate in the fork's context — and a comment describing a migration nobody is doing is exactly what a future merge reasons from
grok-4.5, LOW — the 30-minute publish timeout is inherited from a Blacksmith-timed job; worth re-tuning empirically once a fork release actually runs on the slower GitHub-hosted runner	Acknowledged, not changed. Guessing a number now is not better than upstream's; worth watching the first release

Both independently confirmed the parts I most wanted checked:

• Snoozed persistence is end-to-end — key → useLocalStorage → toggle → aria-expanded and render — and exactly onesettledShelfExpanded state survives, on the fork's key.
• Dropping upstream's settled key loses nothing: same Boolean schema, same true default, and no upstream migration or reset mechanism was skipped. Keeping the fork's key is what preserves existing preferences.
• SETTLED_TAIL_* now comes only from Sidebar.logic, with no competing local declaration, and the surviving useLocalStorage import resolves to the same hook.
• fix(mobile): keep chat composer above the Android gesture bar pingdotgg/t3code#5988 landed whole on mobile, not half-applied: keyboard visibility rather than composer focus drives the bottom safe-area inset, and that inset reaches both the overlap calculation and ThreadComposer. Both new imports are genuinely used.
• feat: add three-hour snooze option pingdotgg/t3code#5914 reaches all six snooze surfaces — sidebar row, sidebar context, board context, the shared thread action menu, and both mobile paths — because they all build from the shared resolveSnoozePresets. No surface carries a hardcoded preset list that would silently omit it.
• release.yml was the only workflow touched, and there are no remaining blacksmith-* references anywhere.

One out-of-scope observation worth recording: the mobile snoozed shelf still holds its expanded
state in useState(false), so it does not persist. pingdotgg#5136 never touched mobile, so this is an
upstream parity gap rather than anything this merge caused — a candidate for a follow-up, not for a
sync PR.

Verification

• Recursive typecheck clean across all 18 packages.
• vp build of the web app passes.
• 2,755 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• pnpm-lock.yaml regenerated rather than hand-merged.

Co-authored by @patroza

opened by Patrick Roza in chat thread Discord · Discord · T3

[omegent-app]

Adopts upstream through 9c7622dac — the 11 commits after 0a7c662d3.

Important

Merge with a merge commit, not squash. Squashing collapses the second parent and discards
upstream's commits as ancestors — the thing that keeps "commits behind upstream" honest.

Upstream commits entering the product

Commit
9c7622dac	fix(server): handle unborn HEAD in VCS status (pingdotgg#5944)
c14bcca10	fix(server): advertise 256-color TERM on Windows terminals (pingdotgg#5693)
1fa315ea9	fix(server): skip Linux libc detection on Windows/macOS (pingdotgg#5354)
f9730979c	perf(web): skip base64 for oversized image candidates (pingdotgg#5220)
59d0c922e	fix(web): persist sidebar shelf collapse state (pingdotgg#5136)
2ea51bd31	fix(server): OpenCode model parsing drops models with a slash in the JSON body (pingdotgg#5072)
e5c82d79a	fix(mobile): keep chat composer above the Android gesture bar (pingdotgg#5988)
9afef94a6	feat: add three-hour snooze option (pingdotgg#5914)
a7b0366cb	fix(web): account for Windows window controls in PR page header (pingdotgg#6049)
428d9f9f6	Allow Android tablets to rotate (pingdotgg#5613)
1b120f352	fix(ci): extend release publish timeout (pingdotgg#6034)

Resolutions

A quiet batch — three conflicts, one of which needed judgement.

Where	Situation	Resolution
Sidebar.tsx	pingdotgg#5136 persists both shelves. The fork already persisted the settled shelf under its own key (…settled-shelf-expanded:v1, default true), so upstream's declaration would have been a second settledShelfExpanded — a duplicate the build would have rejected. Upstream also re-declared SETTLED_TAIL_* locally, which the fork imports from Sidebar.logic	fork's settled persistence kept (identical default; switching keys would silently reset every existing preference), upstream's snoozed-shelf persistence adopted since the fork lacked it, and the duplicate SETTLED_TAIL_* dropped. Also removed a duplicate useLocalStorage import — both parents import the same module under different specifiers (../hooks/… vs ~/hooks/…)
release.yml	upstream runs on Blacksmith andpingdotgg#6034 extends the publish timeout	fork's ubuntu-24.04 kept, upstream's timeout-minutes: 30 taken — the timeout is what that commit was for, and it matters more on a slower GitHub-hosted runner, not less
ThreadDetailScreen.tsx	fork's StyleSheet vs upstream's useKeyboardState	unioned

Checks I ran rather than assumed

• SNOOZED_SHELF_EXPANDED_KEY is consumed, not an orphan constant — declared at Sidebar.tsx:256, used at :2286. Exactly one settledShelfExpanded state remains.
• No Blacksmith references anywhere in .github/workflows. This fork has been bitten twice by upstream workflows arriving on runners it does not have (web-preview.yml, mobile-fingerprint-check.yml), both times in files that merged without a conflict — so this is now a standing check rather than a reaction.
• feat: add three-hour snooze option pingdotgg/t3code#5914's three-hour snooze reaches every fork surface. It lands in the shared resolveSnoozePresets in client-runtime, and the web sidebar, board and both mobile thread lists all call it, so there is no per-surface parity gap to fix.

Adversarial review

Both reviewers went through all three resolutions and the pingdotgg#5914 coverage question. Neither found a
HIGH or a MEDIUM — gpt-5.6-sol reported "no HIGH, MEDIUM, or LOW merge-weld defects". Two LOWs
between them:

Finding	Disposition
both, LOW — the comment above SNOOZED_SHELF_EXPANDED_KEY claims the key is kept so existing preferences survive a rename. True of the fork's settled key; this merge introduces snoozed persistence, so there is nothing to migrate	Fixed in 01a4cee03. It came across with upstream's constant and was simply inaccurate in the fork's context — and a comment describing a migration nobody is doing is exactly what a future merge reasons from
grok-4.5, LOW — the 30-minute publish timeout is inherited from a Blacksmith-timed job; worth re-tuning empirically once a fork release actually runs on the slower GitHub-hosted runner	Acknowledged, not changed. Guessing a number now is not better than upstream's; worth watching the first release

Both independently confirmed the parts I most wanted checked:

• Snoozed persistence is end-to-end — key → useLocalStorage → toggle → aria-expanded and render — and exactly onesettledShelfExpanded state survives, on the fork's key.
• Dropping upstream's settled key loses nothing: same Boolean schema, same true default, and no upstream migration or reset mechanism was skipped. Keeping the fork's key is what preserves existing preferences.
• SETTLED_TAIL_* now comes only from Sidebar.logic, with no competing local declaration, and the surviving useLocalStorage import resolves to the same hook.
• fix(mobile): keep chat composer above the Android gesture bar pingdotgg/t3code#5988 landed whole on mobile, not half-applied: keyboard visibility rather than composer focus drives the bottom safe-area inset, and that inset reaches both the overlap calculation and ThreadComposer. Both new imports are genuinely used.
• feat: add three-hour snooze option pingdotgg/t3code#5914 reaches all six snooze surfaces — sidebar row, sidebar context, board context, the shared thread action menu, and both mobile paths — because they all build from the shared resolveSnoozePresets. No surface carries a hardcoded preset list that would silently omit it.
• release.yml was the only workflow touched, and there are no remaining blacksmith-* references anywhere.

One out-of-scope observation worth recording: the mobile snoozed shelf still holds its expanded
state in useState(false), so it does not persist. pingdotgg#5136 never touched mobile, so this is an
upstream parity gap rather than anything this merge caused — a candidate for a follow-up, not for a
sync PR.

Verification

• Recursive typecheck clean across all 18 packages.
• vp build of the web app passes.
• 2,755 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• pnpm-lock.yaml regenerated rather than hand-merged.

Co-authored by @patroza

opened by Patrick Roza in chat thread Discord · Discord · T3

[omegent-app]

Adopts upstream through 9c7622dac — the 11 commits after 0a7c662d3.

Important

Merge with a merge commit, not squash. Squashing collapses the second parent and discards
upstream's commits as ancestors — the thing that keeps "commits behind upstream" honest.

Upstream commits entering the product

Commit
9c7622dac	fix(server): handle unborn HEAD in VCS status (pingdotgg#5944)
c14bcca10	fix(server): advertise 256-color TERM on Windows terminals (pingdotgg#5693)
1fa315ea9	fix(server): skip Linux libc detection on Windows/macOS (pingdotgg#5354)
f9730979c	perf(web): skip base64 for oversized image candidates (pingdotgg#5220)
59d0c922e	fix(web): persist sidebar shelf collapse state (pingdotgg#5136)
2ea51bd31	fix(server): OpenCode model parsing drops models with a slash in the JSON body (pingdotgg#5072)
e5c82d79a	fix(mobile): keep chat composer above the Android gesture bar (pingdotgg#5988)
9afef94a6	feat: add three-hour snooze option (pingdotgg#5914)
a7b0366cb	fix(web): account for Windows window controls in PR page header (pingdotgg#6049)
428d9f9f6	Allow Android tablets to rotate (pingdotgg#5613)
1b120f352	fix(ci): extend release publish timeout (pingdotgg#6034)

Resolutions

A quiet batch — three conflicts, one of which needed judgement.

Where	Situation	Resolution
Sidebar.tsx	pingdotgg#5136 persists both shelves. The fork already persisted the settled shelf under its own key (…settled-shelf-expanded:v1, default true), so upstream's declaration would have been a second settledShelfExpanded — a duplicate the build would have rejected. Upstream also re-declared SETTLED_TAIL_* locally, which the fork imports from Sidebar.logic	fork's settled persistence kept (identical default; switching keys would silently reset every existing preference), upstream's snoozed-shelf persistence adopted since the fork lacked it, and the duplicate SETTLED_TAIL_* dropped. Also removed a duplicate useLocalStorage import — both parents import the same module under different specifiers (../hooks/… vs ~/hooks/…)
release.yml	upstream runs on Blacksmith andpingdotgg#6034 extends the publish timeout	fork's ubuntu-24.04 kept, upstream's timeout-minutes: 30 taken — the timeout is what that commit was for, and it matters more on a slower GitHub-hosted runner, not less
ThreadDetailScreen.tsx	fork's StyleSheet vs upstream's useKeyboardState	unioned

Checks I ran rather than assumed

• SNOOZED_SHELF_EXPANDED_KEY is consumed, not an orphan constant — declared at Sidebar.tsx:256, used at :2286. Exactly one settledShelfExpanded state remains.
• No Blacksmith references anywhere in .github/workflows. This fork has been bitten twice by upstream workflows arriving on runners it does not have (web-preview.yml, mobile-fingerprint-check.yml), both times in files that merged without a conflict — so this is now a standing check rather than a reaction.
• feat: add three-hour snooze option pingdotgg/t3code#5914's three-hour snooze reaches every fork surface. It lands in the shared resolveSnoozePresets in client-runtime, and the web sidebar, board and both mobile thread lists all call it, so there is no per-surface parity gap to fix.

Adversarial review

Both reviewers went through all three resolutions and the pingdotgg#5914 coverage question. Neither found a
HIGH or a MEDIUM — gpt-5.6-sol reported "no HIGH, MEDIUM, or LOW merge-weld defects". Two LOWs
between them:

Finding	Disposition
both, LOW — the comment above SNOOZED_SHELF_EXPANDED_KEY claims the key is kept so existing preferences survive a rename. True of the fork's settled key; this merge introduces snoozed persistence, so there is nothing to migrate	Fixed in 01a4cee03. It came across with upstream's constant and was simply inaccurate in the fork's context — and a comment describing a migration nobody is doing is exactly what a future merge reasons from
grok-4.5, LOW — the 30-minute publish timeout is inherited from a Blacksmith-timed job; worth re-tuning empirically once a fork release actually runs on the slower GitHub-hosted runner	Acknowledged, not changed. Guessing a number now is not better than upstream's; worth watching the first release

Both independently confirmed the parts I most wanted checked:

• Snoozed persistence is end-to-end — key → useLocalStorage → toggle → aria-expanded and render — and exactly onesettledShelfExpanded state survives, on the fork's key.
• Dropping upstream's settled key loses nothing: same Boolean schema, same true default, and no upstream migration or reset mechanism was skipped. Keeping the fork's key is what preserves existing preferences.
• SETTLED_TAIL_* now comes only from Sidebar.logic, with no competing local declaration, and the surviving useLocalStorage import resolves to the same hook.
• fix(mobile): keep chat composer above the Android gesture bar pingdotgg/t3code#5988 landed whole on mobile, not half-applied: keyboard visibility rather than composer focus drives the bottom safe-area inset, and that inset reaches both the overlap calculation and ThreadComposer. Both new imports are genuinely used.
• feat: add three-hour snooze option pingdotgg/t3code#5914 reaches all six snooze surfaces — sidebar row, sidebar context, board context, the shared thread action menu, and both mobile paths — because they all build from the shared resolveSnoozePresets. No surface carries a hardcoded preset list that would silently omit it.
• release.yml was the only workflow touched, and there are no remaining blacksmith-* references anywhere.

One out-of-scope observation worth recording: the mobile snoozed shelf still holds its expanded
state in useState(false), so it does not persist. pingdotgg#5136 never touched mobile, so this is an
upstream parity gap rather than anything this merge caused — a candidate for a follow-up, not for a
sync PR.

Verification

• Recursive typecheck clean across all 18 packages.
• vp build of the web app passes.
• 2,755 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• pnpm-lock.yaml regenerated rather than hand-merged.

Co-authored by @patroza

opened by Patrick Roza in chat thread Discord · Discord · T3

[omegent-app]

Adopts upstream through 9c7622dac — the 11 commits after 0a7c662d3.

Important

Merge with a merge commit, not squash. Squashing collapses the second parent and discards
upstream's commits as ancestors — the thing that keeps "commits behind upstream" honest.

Upstream commits entering the product

Commit
9c7622dac	fix(server): handle unborn HEAD in VCS status (pingdotgg#5944)
c14bcca10	fix(server): advertise 256-color TERM on Windows terminals (pingdotgg#5693)
1fa315ea9	fix(server): skip Linux libc detection on Windows/macOS (pingdotgg#5354)
f9730979c	perf(web): skip base64 for oversized image candidates (pingdotgg#5220)
59d0c922e	fix(web): persist sidebar shelf collapse state (pingdotgg#5136)
2ea51bd31	fix(server): OpenCode model parsing drops models with a slash in the JSON body (pingdotgg#5072)
e5c82d79a	fix(mobile): keep chat composer above the Android gesture bar (pingdotgg#5988)
9afef94a6	feat: add three-hour snooze option (pingdotgg#5914)
a7b0366cb	fix(web): account for Windows window controls in PR page header (pingdotgg#6049)
428d9f9f6	Allow Android tablets to rotate (pingdotgg#5613)
1b120f352	fix(ci): extend release publish timeout (pingdotgg#6034)

Resolutions

A quiet batch — three conflicts, one of which needed judgement.

Where	Situation	Resolution
Sidebar.tsx	pingdotgg#5136 persists both shelves. The fork already persisted the settled shelf under its own key (…settled-shelf-expanded:v1, default true), so upstream's declaration would have been a second settledShelfExpanded — a duplicate the build would have rejected. Upstream also re-declared SETTLED_TAIL_* locally, which the fork imports from Sidebar.logic	fork's settled persistence kept (identical default; switching keys would silently reset every existing preference), upstream's snoozed-shelf persistence adopted since the fork lacked it, and the duplicate SETTLED_TAIL_* dropped. Also removed a duplicate useLocalStorage import — both parents import the same module under different specifiers (../hooks/… vs ~/hooks/…)
release.yml	upstream runs on Blacksmith andpingdotgg#6034 extends the publish timeout	fork's ubuntu-24.04 kept, upstream's timeout-minutes: 30 taken — the timeout is what that commit was for, and it matters more on a slower GitHub-hosted runner, not less
ThreadDetailScreen.tsx	fork's StyleSheet vs upstream's useKeyboardState	unioned

Checks I ran rather than assumed

• SNOOZED_SHELF_EXPANDED_KEY is consumed, not an orphan constant — declared at Sidebar.tsx:256, used at :2286. Exactly one settledShelfExpanded state remains.
• No Blacksmith references anywhere in .github/workflows. This fork has been bitten twice by upstream workflows arriving on runners it does not have (web-preview.yml, mobile-fingerprint-check.yml), both times in files that merged without a conflict — so this is now a standing check rather than a reaction.
• feat: add three-hour snooze option pingdotgg/t3code#5914's three-hour snooze reaches every fork surface. It lands in the shared resolveSnoozePresets in client-runtime, and the web sidebar, board and both mobile thread lists all call it, so there is no per-surface parity gap to fix.

Adversarial review

Both reviewers went through all three resolutions and the pingdotgg#5914 coverage question. Neither found a
HIGH or a MEDIUM — gpt-5.6-sol reported "no HIGH, MEDIUM, or LOW merge-weld defects". Two LOWs
between them:

Finding	Disposition
both, LOW — the comment above SNOOZED_SHELF_EXPANDED_KEY claims the key is kept so existing preferences survive a rename. True of the fork's settled key; this merge introduces snoozed persistence, so there is nothing to migrate	Fixed in 01a4cee03. It came across with upstream's constant and was simply inaccurate in the fork's context — and a comment describing a migration nobody is doing is exactly what a future merge reasons from
grok-4.5, LOW — the 30-minute publish timeout is inherited from a Blacksmith-timed job; worth re-tuning empirically once a fork release actually runs on the slower GitHub-hosted runner	Acknowledged, not changed. Guessing a number now is not better than upstream's; worth watching the first release

Both independently confirmed the parts I most wanted checked:

• Snoozed persistence is end-to-end — key → useLocalStorage → toggle → aria-expanded and render — and exactly onesettledShelfExpanded state survives, on the fork's key.
• Dropping upstream's settled key loses nothing: same Boolean schema, same true default, and no upstream migration or reset mechanism was skipped. Keeping the fork's key is what preserves existing preferences.
• SETTLED_TAIL_* now comes only from Sidebar.logic, with no competing local declaration, and the surviving useLocalStorage import resolves to the same hook.
• fix(mobile): keep chat composer above the Android gesture bar pingdotgg/t3code#5988 landed whole on mobile, not half-applied: keyboard visibility rather than composer focus drives the bottom safe-area inset, and that inset reaches both the overlap calculation and ThreadComposer. Both new imports are genuinely used.
• feat: add three-hour snooze option pingdotgg/t3code#5914 reaches all six snooze surfaces — sidebar row, sidebar context, board context, the shared thread action menu, and both mobile paths — because they all build from the shared resolveSnoozePresets. No surface carries a hardcoded preset list that would silently omit it.
• release.yml was the only workflow touched, and there are no remaining blacksmith-* references anywhere.

One out-of-scope observation worth recording: the mobile snoozed shelf still holds its expanded
state in useState(false), so it does not persist. pingdotgg#5136 never touched mobile, so this is an
upstream parity gap rather than anything this merge caused — a candidate for a follow-up, not for a
sync PR.

Verification

• Recursive typecheck clean across all 18 packages.
• vp build of the web app passes.
• 2,755 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• pnpm-lock.yaml regenerated rather than hand-merged.

Co-authored by @patroza

opened by Patrick Roza in chat thread Discord · Discord · T3