sync(upstream): merge upstream/main 239ef1c54 into fork/dev

[omegent-app]

Adopts upstream through 239ef1c54 — the seven commits after #377, bringing fork/devlevel with
upstream/main (HEAD..upstream/main is empty).

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
239ef1c54	fix(web): show the correctly matching shortcut in new thread button's tooltip (pingdotgg#5594)
61b51ae0e	feat(web): click the pin icon to unpin a thread (pingdotgg#5578)
bfc69e4b4	Bump mobile app version to 1.0.2 (pingdotgg#5588)
af281c9fc	fix(mobile): repair Clerk auth navigation headers (pingdotgg#5140)
bd422fd8d	fix(mobile): keep Android chat text from showing through the composer (pingdotgg#5582)
33a03c8a7	fix(mobile): pad scroll views above Android nav bar (pingdotgg#5415)
a17459e8a	fix(mobile): avoid iOS terminal reset on clear (pingdotgg#5440)

All seven verified as ancestors of the merge commit individually, not just by range.

Resolutions

Three conflicts, each a real judgement call:

Where	Situation	Resolution
ThreadComposer.tsx	pingdotgg#5582 moves the backdrop gradient offAnimated.View onto a plain View + StyleSheet.absoluteFill, because Reanimated silently drops experimental_backgroundImage there on Android — which is the bug it fixes. The fork's gradient sat on exactly that Animated.View	removed from the animated node; the fork's stronger lower-band stops (0.82/0.96 dark, 0.88/0.98 light at 42%) moved onto upstream's backdrop, so the fork's tuning survives on the node that actually paints
app.config.ts	upstream bumps version to 1.0.2; the fork wraps runtimeVersion in a MOBILE_RUNTIME_VERSION_OVERRIDE escape hatch for OTA-to-installed-binary cases	version bumped inside the fork's IIFE; its default policy is the same fingerprint upstream documents
SidebarV2.tsx	pingdotgg#5594 argues the tooltip must name chat.new, not chat.newLocal	adopted — checked against handleNewThreadClick: the fork's button routes multi-project setups through the palette picker and creates immediately otherwise, i.e. exactly chat.new semantics. The fork's board and command-palette labels stay

Plus one integration fix the merge owns, found by this PR's own CI going red:

Where	Situation	Resolution
apps/web/vite.config.ts	upstream pingdotgg#4479 adds two tests that vi.mock("react") and drive components through reactHookHarness. Upstream runs tests fully isolated; the fork's unit project sets isolate: false to reuse each worker's module graph. Under a shared registry the component graph can already be bound to the real react/compiler-runtime by an earlier file, the mock never applies, and the compiled component reports a memo-cache hit — skipping the very hooks the test asserts on	both files added to the fork's existing isolatedUnitTestFiles, where two other react-mocking tests already live

That is the whole story behind the AddProviderInstanceDialog.environment "flake": not timing (the test is synchronous) but worker scheduling deciding whether the mock lands. It failed this PR's first CI run and had flaked on earlier syncs. Three consecutive full-workspace runs are now identical (1 failed / 2,278 passed — the pre-existing Codex one).

Adversarial review

grok-4.5 and gpt-5.6-sol over the git show --cc surface with worktree access — no defects
from either. Both independently confirmed: the gradient paints once, on the plain View only; the
override still feeds Expo's runtimeVersion; the shortcut label matches the command the button
actually invokes; and no fixture is still pinned to 1.0.1 (both checked the packages/contracts
latestVersion: "1.0.1" and identified it as unrelated Codex advisory fixture data).

Verification

• Full recursive typecheck clean across 17 packages.
• 2,278 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• The AddProviderInstanceDialog.environment flake is root-caused and fixed (see above), not
  retried away: three consecutive full-workspace runs now give identical results.

Co-authored by @patroza

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

[omegent-app]

Adopts upstream through 239ef1c54 — the seven commits after #377, bringing fork/devlevel with
upstream/main (HEAD..upstream/main is empty).

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
239ef1c54	fix(web): show the correctly matching shortcut in new thread button's tooltip (pingdotgg#5594)
61b51ae0e	feat(web): click the pin icon to unpin a thread (pingdotgg#5578)
bfc69e4b4	Bump mobile app version to 1.0.2 (pingdotgg#5588)
af281c9fc	fix(mobile): repair Clerk auth navigation headers (pingdotgg#5140)
bd422fd8d	fix(mobile): keep Android chat text from showing through the composer (pingdotgg#5582)
33a03c8a7	fix(mobile): pad scroll views above Android nav bar (pingdotgg#5415)
a17459e8a	fix(mobile): avoid iOS terminal reset on clear (pingdotgg#5440)

All seven verified as ancestors of the merge commit individually, not just by range.

Resolutions

Three conflicts, each a real judgement call:

Where	Situation	Resolution
ThreadComposer.tsx	pingdotgg#5582 moves the backdrop gradient offAnimated.View onto a plain View + StyleSheet.absoluteFill, because Reanimated silently drops experimental_backgroundImage there on Android — which is the bug it fixes. The fork's gradient sat on exactly that Animated.View	removed from the animated node; the fork's stronger lower-band stops (0.82/0.96 dark, 0.88/0.98 light at 42%) moved onto upstream's backdrop, so the fork's tuning survives on the node that actually paints
app.config.ts	upstream bumps version to 1.0.2; the fork wraps runtimeVersion in a MOBILE_RUNTIME_VERSION_OVERRIDE escape hatch for OTA-to-installed-binary cases	version bumped inside the fork's IIFE; its default policy is the same fingerprint upstream documents
SidebarV2.tsx	pingdotgg#5594 argues the tooltip must name chat.new, not chat.newLocal	adopted — checked against handleNewThreadClick: the fork's button routes multi-project setups through the palette picker and creates immediately otherwise, i.e. exactly chat.new semantics. The fork's board and command-palette labels stay

Plus one integration fix the merge owns, found by this PR's own CI going red:

Where	Situation	Resolution
apps/web/vite.config.ts	upstream pingdotgg#4479 adds two tests that vi.mock("react") and drive components through reactHookHarness. Upstream runs tests fully isolated; the fork's unit project sets isolate: false to reuse each worker's module graph. Under a shared registry the component graph can already be bound to the real react/compiler-runtime by an earlier file, the mock never applies, and the compiled component reports a memo-cache hit — skipping the very hooks the test asserts on	both files added to the fork's existing isolatedUnitTestFiles, where two other react-mocking tests already live

That is the whole story behind the AddProviderInstanceDialog.environment "flake": not timing (the test is synchronous) but worker scheduling deciding whether the mock lands. It failed this PR's first CI run and had flaked on earlier syncs. Three consecutive full-workspace runs are now identical (1 failed / 2,278 passed — the pre-existing Codex one).

Adversarial review

grok-4.5 and gpt-5.6-sol over the git show --cc surface with worktree access — no defects
from either. Both independently confirmed: the gradient paints once, on the plain View only; the
override still feeds Expo's runtimeVersion; the shortcut label matches the command the button
actually invokes; and no fixture is still pinned to 1.0.1 (both checked the packages/contracts
latestVersion: "1.0.1" and identified it as unrelated Codex advisory fixture data).

Verification

• Full recursive typecheck clean across 17 packages.
• 2,278 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• The AddProviderInstanceDialog.environment flake is root-caused and fixed (see above), not
  retried away: three consecutive full-workspace runs now give identical results.

Co-authored by @patroza

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

[omegent-app]

Adopts upstream through 239ef1c54 — the seven commits after #377, bringing fork/devlevel with
upstream/main (HEAD..upstream/main is empty).

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
239ef1c54	fix(web): show the correctly matching shortcut in new thread button's tooltip (pingdotgg#5594)
61b51ae0e	feat(web): click the pin icon to unpin a thread (pingdotgg#5578)
bfc69e4b4	Bump mobile app version to 1.0.2 (pingdotgg#5588)
af281c9fc	fix(mobile): repair Clerk auth navigation headers (pingdotgg#5140)
bd422fd8d	fix(mobile): keep Android chat text from showing through the composer (pingdotgg#5582)
33a03c8a7	fix(mobile): pad scroll views above Android nav bar (pingdotgg#5415)
a17459e8a	fix(mobile): avoid iOS terminal reset on clear (pingdotgg#5440)

All seven verified as ancestors of the merge commit individually, not just by range.

Resolutions

Three conflicts, each a real judgement call:

Where	Situation	Resolution
ThreadComposer.tsx	pingdotgg#5582 moves the backdrop gradient offAnimated.View onto a plain View + StyleSheet.absoluteFill, because Reanimated silently drops experimental_backgroundImage there on Android — which is the bug it fixes. The fork's gradient sat on exactly that Animated.View	removed from the animated node; the fork's stronger lower-band stops (0.82/0.96 dark, 0.88/0.98 light at 42%) moved onto upstream's backdrop, so the fork's tuning survives on the node that actually paints
app.config.ts	upstream bumps version to 1.0.2; the fork wraps runtimeVersion in a MOBILE_RUNTIME_VERSION_OVERRIDE escape hatch for OTA-to-installed-binary cases	version bumped inside the fork's IIFE; its default policy is the same fingerprint upstream documents
SidebarV2.tsx	pingdotgg#5594 argues the tooltip must name chat.new, not chat.newLocal	adopted — checked against handleNewThreadClick: the fork's button routes multi-project setups through the palette picker and creates immediately otherwise, i.e. exactly chat.new semantics. The fork's board and command-palette labels stay

Plus one integration fix the merge owns, found by this PR's own CI going red:

Where	Situation	Resolution
apps/web/vite.config.ts	upstream pingdotgg#4479 adds two tests that vi.mock("react") and drive components through reactHookHarness. Upstream runs tests fully isolated; the fork's unit project sets isolate: false to reuse each worker's module graph. Under a shared registry the component graph can already be bound to the real react/compiler-runtime by an earlier file, the mock never applies, and the compiled component reports a memo-cache hit — skipping the very hooks the test asserts on	both files added to the fork's existing isolatedUnitTestFiles, where two other react-mocking tests already live

That is the whole story behind the AddProviderInstanceDialog.environment "flake": not timing (the test is synchronous) but worker scheduling deciding whether the mock lands. It failed this PR's first CI run and had flaked on earlier syncs. Three consecutive full-workspace runs are now identical (1 failed / 2,278 passed — the pre-existing Codex one).

Adversarial review

grok-4.5 and gpt-5.6-sol over the git show --cc surface with worktree access — no defects
from either. Both independently confirmed: the gradient paints once, on the plain View only; the
override still feeds Expo's runtimeVersion; the shortcut label matches the command the button
actually invokes; and no fixture is still pinned to 1.0.1 (both checked the packages/contracts
latestVersion: "1.0.1" and identified it as unrelated Codex advisory fixture data).

Verification

• Full recursive typecheck clean across 17 packages.
• 2,278 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• The AddProviderInstanceDialog.environment flake is root-caused and fixed (see above), not
  retried away: three consecutive full-workspace runs now give identical results.

Co-authored by @patroza

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

[omegent-app]

Adopts upstream through 239ef1c54 — the seven commits after #377, bringing fork/devlevel with
upstream/main (HEAD..upstream/main is empty).

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
239ef1c54	fix(web): show the correctly matching shortcut in new thread button's tooltip (pingdotgg#5594)
61b51ae0e	feat(web): click the pin icon to unpin a thread (pingdotgg#5578)
bfc69e4b4	Bump mobile app version to 1.0.2 (pingdotgg#5588)
af281c9fc	fix(mobile): repair Clerk auth navigation headers (pingdotgg#5140)
bd422fd8d	fix(mobile): keep Android chat text from showing through the composer (pingdotgg#5582)
33a03c8a7	fix(mobile): pad scroll views above Android nav bar (pingdotgg#5415)
a17459e8a	fix(mobile): avoid iOS terminal reset on clear (pingdotgg#5440)

All seven verified as ancestors of the merge commit individually, not just by range.

Resolutions

Three conflicts, each a real judgement call:

Where	Situation	Resolution
ThreadComposer.tsx	pingdotgg#5582 moves the backdrop gradient offAnimated.View onto a plain View + StyleSheet.absoluteFill, because Reanimated silently drops experimental_backgroundImage there on Android — which is the bug it fixes. The fork's gradient sat on exactly that Animated.View	removed from the animated node; the fork's stronger lower-band stops (0.82/0.96 dark, 0.88/0.98 light at 42%) moved onto upstream's backdrop, so the fork's tuning survives on the node that actually paints
app.config.ts	upstream bumps version to 1.0.2; the fork wraps runtimeVersion in a MOBILE_RUNTIME_VERSION_OVERRIDE escape hatch for OTA-to-installed-binary cases	version bumped inside the fork's IIFE; its default policy is the same fingerprint upstream documents
SidebarV2.tsx	pingdotgg#5594 argues the tooltip must name chat.new, not chat.newLocal	adopted — checked against handleNewThreadClick: the fork's button routes multi-project setups through the palette picker and creates immediately otherwise, i.e. exactly chat.new semantics. The fork's board and command-palette labels stay

Plus one integration fix the merge owns, found by this PR's own CI going red:

Where	Situation	Resolution
apps/web/vite.config.ts	upstream pingdotgg#4479 adds two tests that vi.mock("react") and drive components through reactHookHarness. Upstream runs tests fully isolated; the fork's unit project sets isolate: false to reuse each worker's module graph. Under a shared registry the component graph can already be bound to the real react/compiler-runtime by an earlier file, the mock never applies, and the compiled component reports a memo-cache hit — skipping the very hooks the test asserts on	both files added to the fork's existing isolatedUnitTestFiles, where two other react-mocking tests already live

That is the whole story behind the AddProviderInstanceDialog.environment "flake": not timing (the test is synchronous) but worker scheduling deciding whether the mock lands. It failed this PR's first CI run and had flaked on earlier syncs. Three consecutive full-workspace runs are now identical (1 failed / 2,278 passed — the pre-existing Codex one).

Adversarial review

grok-4.5 and gpt-5.6-sol over the git show --cc surface with worktree access — no defects
from either. Both independently confirmed: the gradient paints once, on the plain View only; the
override still feeds Expo's runtimeVersion; the shortcut label matches the command the button
actually invokes; and no fixture is still pinned to 1.0.1 (both checked the packages/contracts
latestVersion: "1.0.1" and identified it as unrelated Codex advisory fixture data).

Verification

• Full recursive typecheck clean across 17 packages.
• 2,278 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• The AddProviderInstanceDialog.environment flake is root-caused and fixed (see above), not
  retried away: three consecutive full-workspace runs now give identical results.

Co-authored by @patroza

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

[omegent-app]

Adopts upstream through 239ef1c54 — the seven commits after #377, bringing fork/devlevel with
upstream/main (HEAD..upstream/main is empty).

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
239ef1c54	fix(web): show the correctly matching shortcut in new thread button's tooltip (pingdotgg#5594)
61b51ae0e	feat(web): click the pin icon to unpin a thread (pingdotgg#5578)
bfc69e4b4	Bump mobile app version to 1.0.2 (pingdotgg#5588)
af281c9fc	fix(mobile): repair Clerk auth navigation headers (pingdotgg#5140)
bd422fd8d	fix(mobile): keep Android chat text from showing through the composer (pingdotgg#5582)
33a03c8a7	fix(mobile): pad scroll views above Android nav bar (pingdotgg#5415)
a17459e8a	fix(mobile): avoid iOS terminal reset on clear (pingdotgg#5440)

All seven verified as ancestors of the merge commit individually, not just by range.

Resolutions

Three conflicts, each a real judgement call:

Where	Situation	Resolution
ThreadComposer.tsx	pingdotgg#5582 moves the backdrop gradient offAnimated.View onto a plain View + StyleSheet.absoluteFill, because Reanimated silently drops experimental_backgroundImage there on Android — which is the bug it fixes. The fork's gradient sat on exactly that Animated.View	removed from the animated node; the fork's stronger lower-band stops (0.82/0.96 dark, 0.88/0.98 light at 42%) moved onto upstream's backdrop, so the fork's tuning survives on the node that actually paints
app.config.ts	upstream bumps version to 1.0.2; the fork wraps runtimeVersion in a MOBILE_RUNTIME_VERSION_OVERRIDE escape hatch for OTA-to-installed-binary cases	version bumped inside the fork's IIFE; its default policy is the same fingerprint upstream documents
SidebarV2.tsx	pingdotgg#5594 argues the tooltip must name chat.new, not chat.newLocal	adopted — checked against handleNewThreadClick: the fork's button routes multi-project setups through the palette picker and creates immediately otherwise, i.e. exactly chat.new semantics. The fork's board and command-palette labels stay

Plus one integration fix the merge owns, found by this PR's own CI going red:

Where	Situation	Resolution
apps/web/vite.config.ts	upstream pingdotgg#4479 adds two tests that vi.mock("react") and drive components through reactHookHarness. Upstream runs tests fully isolated; the fork's unit project sets isolate: false to reuse each worker's module graph. Under a shared registry the component graph can already be bound to the real react/compiler-runtime by an earlier file, the mock never applies, and the compiled component reports a memo-cache hit — skipping the very hooks the test asserts on	both files added to the fork's existing isolatedUnitTestFiles, where two other react-mocking tests already live

That is the whole story behind the AddProviderInstanceDialog.environment "flake": not timing (the test is synchronous) but worker scheduling deciding whether the mock lands. It failed this PR's first CI run and had flaked on earlier syncs. Three consecutive full-workspace runs are now identical (1 failed / 2,278 passed — the pre-existing Codex one).

Adversarial review

grok-4.5 and gpt-5.6-sol over the git show --cc surface with worktree access — no defects
from either. Both independently confirmed: the gradient paints once, on the plain View only; the
override still feeds Expo's runtimeVersion; the shortcut label matches the command the button
actually invokes; and no fixture is still pinned to 1.0.1 (both checked the packages/contracts
latestVersion: "1.0.1" and identified it as unrelated Codex advisory fixture data).

Verification

• Full recursive typecheck clean across 17 packages.
• 2,278 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• The AddProviderInstanceDialog.environment flake is root-caused and fixed (see above), not
  retried away: three consecutive full-workspace runs now give identical results.

Co-authored by @patroza

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

[omegent-app]

Adopts upstream through 239ef1c54 — the seven commits after #377, bringing fork/devlevel with
upstream/main (HEAD..upstream/main is empty).

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
239ef1c54	fix(web): show the correctly matching shortcut in new thread button's tooltip (pingdotgg#5594)
61b51ae0e	feat(web): click the pin icon to unpin a thread (pingdotgg#5578)
bfc69e4b4	Bump mobile app version to 1.0.2 (pingdotgg#5588)
af281c9fc	fix(mobile): repair Clerk auth navigation headers (pingdotgg#5140)
bd422fd8d	fix(mobile): keep Android chat text from showing through the composer (pingdotgg#5582)
33a03c8a7	fix(mobile): pad scroll views above Android nav bar (pingdotgg#5415)
a17459e8a	fix(mobile): avoid iOS terminal reset on clear (pingdotgg#5440)

All seven verified as ancestors of the merge commit individually, not just by range.

Resolutions

Three conflicts, each a real judgement call:

Where	Situation	Resolution
ThreadComposer.tsx	pingdotgg#5582 moves the backdrop gradient offAnimated.View onto a plain View + StyleSheet.absoluteFill, because Reanimated silently drops experimental_backgroundImage there on Android — which is the bug it fixes. The fork's gradient sat on exactly that Animated.View	removed from the animated node; the fork's stronger lower-band stops (0.82/0.96 dark, 0.88/0.98 light at 42%) moved onto upstream's backdrop, so the fork's tuning survives on the node that actually paints
app.config.ts	upstream bumps version to 1.0.2; the fork wraps runtimeVersion in a MOBILE_RUNTIME_VERSION_OVERRIDE escape hatch for OTA-to-installed-binary cases	version bumped inside the fork's IIFE; its default policy is the same fingerprint upstream documents
SidebarV2.tsx	pingdotgg#5594 argues the tooltip must name chat.new, not chat.newLocal	adopted — checked against handleNewThreadClick: the fork's button routes multi-project setups through the palette picker and creates immediately otherwise, i.e. exactly chat.new semantics. The fork's board and command-palette labels stay

Plus one integration fix the merge owns, found by this PR's own CI going red:

Where	Situation	Resolution
apps/web/vite.config.ts	upstream pingdotgg#4479 adds two tests that vi.mock("react") and drive components through reactHookHarness. Upstream runs tests fully isolated; the fork's unit project sets isolate: false to reuse each worker's module graph. Under a shared registry the component graph can already be bound to the real react/compiler-runtime by an earlier file, the mock never applies, and the compiled component reports a memo-cache hit — skipping the very hooks the test asserts on	both files added to the fork's existing isolatedUnitTestFiles, where two other react-mocking tests already live

That is the whole story behind the AddProviderInstanceDialog.environment "flake": not timing (the test is synchronous) but worker scheduling deciding whether the mock lands. It failed this PR's first CI run and had flaked on earlier syncs. Three consecutive full-workspace runs are now identical (1 failed / 2,278 passed — the pre-existing Codex one).

Adversarial review

grok-4.5 and gpt-5.6-sol over the git show --cc surface with worktree access — no defects
from either. Both independently confirmed: the gradient paints once, on the plain View only; the
override still feeds Expo's runtimeVersion; the shortcut label matches the command the button
actually invokes; and no fixture is still pinned to 1.0.1 (both checked the packages/contracts
latestVersion: "1.0.1" and identified it as unrelated Codex advisory fixture data).

Verification

• Full recursive typecheck clean across 17 packages.
• 2,278 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• The AddProviderInstanceDialog.environment flake is root-caused and fixed (see above), not
  retried away: three consecutive full-workspace runs now give identical results.

Co-authored by @patroza

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

[omegent-app]

Adopts upstream through 239ef1c54 — the seven commits after #377, bringing fork/devlevel with
upstream/main (HEAD..upstream/main is empty).

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
239ef1c54	fix(web): show the correctly matching shortcut in new thread button's tooltip (pingdotgg#5594)
61b51ae0e	feat(web): click the pin icon to unpin a thread (pingdotgg#5578)
bfc69e4b4	Bump mobile app version to 1.0.2 (pingdotgg#5588)
af281c9fc	fix(mobile): repair Clerk auth navigation headers (pingdotgg#5140)
bd422fd8d	fix(mobile): keep Android chat text from showing through the composer (pingdotgg#5582)
33a03c8a7	fix(mobile): pad scroll views above Android nav bar (pingdotgg#5415)
a17459e8a	fix(mobile): avoid iOS terminal reset on clear (pingdotgg#5440)

All seven verified as ancestors of the merge commit individually, not just by range.

Resolutions

Three conflicts, each a real judgement call:

Where	Situation	Resolution
ThreadComposer.tsx	pingdotgg#5582 moves the backdrop gradient offAnimated.View onto a plain View + StyleSheet.absoluteFill, because Reanimated silently drops experimental_backgroundImage there on Android — which is the bug it fixes. The fork's gradient sat on exactly that Animated.View	removed from the animated node; the fork's stronger lower-band stops (0.82/0.96 dark, 0.88/0.98 light at 42%) moved onto upstream's backdrop, so the fork's tuning survives on the node that actually paints
app.config.ts	upstream bumps version to 1.0.2; the fork wraps runtimeVersion in a MOBILE_RUNTIME_VERSION_OVERRIDE escape hatch for OTA-to-installed-binary cases	version bumped inside the fork's IIFE; its default policy is the same fingerprint upstream documents
SidebarV2.tsx	pingdotgg#5594 argues the tooltip must name chat.new, not chat.newLocal	adopted — checked against handleNewThreadClick: the fork's button routes multi-project setups through the palette picker and creates immediately otherwise, i.e. exactly chat.new semantics. The fork's board and command-palette labels stay

Plus one integration fix the merge owns, found by this PR's own CI going red:

Where	Situation	Resolution
apps/web/vite.config.ts	upstream pingdotgg#4479 adds two tests that vi.mock("react") and drive components through reactHookHarness. Upstream runs tests fully isolated; the fork's unit project sets isolate: false to reuse each worker's module graph. Under a shared registry the component graph can already be bound to the real react/compiler-runtime by an earlier file, the mock never applies, and the compiled component reports a memo-cache hit — skipping the very hooks the test asserts on	both files added to the fork's existing isolatedUnitTestFiles, where two other react-mocking tests already live

That is the whole story behind the AddProviderInstanceDialog.environment "flake": not timing (the test is synchronous) but worker scheduling deciding whether the mock lands. It failed this PR's first CI run and had flaked on earlier syncs. Three consecutive full-workspace runs are now identical (1 failed / 2,278 passed — the pre-existing Codex one).

Adversarial review

grok-4.5 and gpt-5.6-sol over the git show --cc surface with worktree access — no defects
from either. Both independently confirmed: the gradient paints once, on the plain View only; the
override still feeds Expo's runtimeVersion; the shortcut label matches the command the button
actually invokes; and no fixture is still pinned to 1.0.1 (both checked the packages/contracts
latestVersion: "1.0.1" and identified it as unrelated Codex advisory fixture data).

Verification

• Full recursive typecheck clean across 17 packages.
• 2,278 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• The AddProviderInstanceDialog.environment flake is root-caused and fixed (see above), not
  retried away: three consecutive full-workspace runs now give identical results.

Co-authored by @patroza

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

[omegent-app]

Adopts upstream through 239ef1c54 — the seven commits after #377, bringing fork/devlevel with
upstream/main (HEAD..upstream/main is empty).

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
239ef1c54	fix(web): show the correctly matching shortcut in new thread button's tooltip (pingdotgg#5594)
61b51ae0e	feat(web): click the pin icon to unpin a thread (pingdotgg#5578)
bfc69e4b4	Bump mobile app version to 1.0.2 (pingdotgg#5588)
af281c9fc	fix(mobile): repair Clerk auth navigation headers (pingdotgg#5140)
bd422fd8d	fix(mobile): keep Android chat text from showing through the composer (pingdotgg#5582)
33a03c8a7	fix(mobile): pad scroll views above Android nav bar (pingdotgg#5415)
a17459e8a	fix(mobile): avoid iOS terminal reset on clear (pingdotgg#5440)

All seven verified as ancestors of the merge commit individually, not just by range.

Resolutions

Three conflicts, each a real judgement call:

Where	Situation	Resolution
ThreadComposer.tsx	pingdotgg#5582 moves the backdrop gradient offAnimated.View onto a plain View + StyleSheet.absoluteFill, because Reanimated silently drops experimental_backgroundImage there on Android — which is the bug it fixes. The fork's gradient sat on exactly that Animated.View	removed from the animated node; the fork's stronger lower-band stops (0.82/0.96 dark, 0.88/0.98 light at 42%) moved onto upstream's backdrop, so the fork's tuning survives on the node that actually paints
app.config.ts	upstream bumps version to 1.0.2; the fork wraps runtimeVersion in a MOBILE_RUNTIME_VERSION_OVERRIDE escape hatch for OTA-to-installed-binary cases	version bumped inside the fork's IIFE; its default policy is the same fingerprint upstream documents
SidebarV2.tsx	pingdotgg#5594 argues the tooltip must name chat.new, not chat.newLocal	adopted — checked against handleNewThreadClick: the fork's button routes multi-project setups through the palette picker and creates immediately otherwise, i.e. exactly chat.new semantics. The fork's board and command-palette labels stay

Plus one integration fix the merge owns, found by this PR's own CI going red:

Where	Situation	Resolution
apps/web/vite.config.ts	upstream pingdotgg#4479 adds two tests that vi.mock("react") and drive components through reactHookHarness. Upstream runs tests fully isolated; the fork's unit project sets isolate: false to reuse each worker's module graph. Under a shared registry the component graph can already be bound to the real react/compiler-runtime by an earlier file, the mock never applies, and the compiled component reports a memo-cache hit — skipping the very hooks the test asserts on	both files added to the fork's existing isolatedUnitTestFiles, where two other react-mocking tests already live

That is the whole story behind the AddProviderInstanceDialog.environment "flake": not timing (the test is synchronous) but worker scheduling deciding whether the mock lands. It failed this PR's first CI run and had flaked on earlier syncs. Three consecutive full-workspace runs are now identical (1 failed / 2,278 passed — the pre-existing Codex one).

Adversarial review

grok-4.5 and gpt-5.6-sol over the git show --cc surface with worktree access — no defects
from either. Both independently confirmed: the gradient paints once, on the plain View only; the
override still feeds Expo's runtimeVersion; the shortcut label matches the command the button
actually invokes; and no fixture is still pinned to 1.0.1 (both checked the packages/contracts
latestVersion: "1.0.1" and identified it as unrelated Codex advisory fixture data).

Verification

• Full recursive typecheck clean across 17 packages.
• 2,278 tests pass; the only failure is the pre-existing CodexTextGeneration launch-args one,
  which reproduces on unmerged fork/dev.
• The AddProviderInstanceDialog.environment flake is root-caused and fixed (see above), not
  retried away: three consecutive full-workspace runs now give identical results.

Co-authored by @patroza

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