Conversation
Deploying with
|
| Status | Name | Latest Commit | Preview URL | Updated (UTC) |
|---|---|---|---|---|
| ✅ Deployment successful! View logs |
skymap | 789da09 | Commit Preview URL Branch Preview URL |
Oct 04 2026, 08:57 AM |
The focus tween, the follow seed, the home pose and the lon/lat fly-to now all take their body distance from focusFraming, so a followed body honours focusDistanceRadii instead of re-deriving the screen-fill distance.
Each row's read returns its share of one LinkIntent; linkIntentFrom merges them and owns precedence (a takeover beats focus and pose, focus plus pose is one focus view), so table order fixes only the write layout. The applyLinkIntent shim emits exactly the actions the rows used to.
requestFocus({ id, transition }) hands its transition to
updateSelectionFocus as meta; the focus tween runs only on 'fly'. A 'cut'
still spends a parked #pose=, since it is the arrival focus. Every existing
caller passes 'fly'.
bodySelectionRow, bodyHomePose and the lon/lat fly-to build a body's SelectionRow through bodyRowAt instead of three literals; the fly-to frames only off the focused body. Trims tests that could not fail and tightens three doc lines.
One path from a LinkIntent to the screen: t and orientation first, then the subject, then one camera commit. A cut focus frames its row with framingPose in the same tick the row lands; a fly focus tweens as before. The write half canonicalizes on the arrival settling instead of hashArrivalApplied, and the applyLinkIntent shim goes.
The boot read hands its LinkIntent to arrivalSaga, which cuts to the subject once the camera exists, waits two frames and puts arrived; an id still unresolved once loads go quiet, or the 30 s backstop, lands home as failed. wireInput commits only the neutral home base. The urlPose park and selectHasSelectionIntent go with it.
The arrival puts goLive before any framing when the link has no t, so a plain boot frames Earth against the wall-clock sun; startLoop's guarded goLive goes. A request retired while deferring is dropped on resolve, and a failed arrival clears the selection before landing home. The backstop starts once the camera exists, an engine error fails the arrival, and the unknown-id check re-reads load progress after its hop. The perf hook's setPose waits on afterTwoFrames.
Three read-only hash rows name a takeover subject; navigateSaga validates the id against its registry before dispatching anything. An exhibit cut commits its fitted pose in place of the fly, and its copy shows at once (openExhibit carries entry: Transition). A tour or clip arrives once its first clip has started, so an opening snap lands behind the veil.
An opaque veil with the splash's load progress, shown while the arrival is pending on a boot that carried a deep link; it fades once the subject is on screen. A plain boot keeps the splash. Mounted outside cinema mode only.
A cut tour or clip that the arrival gives up on is stopped (exitTakeover / stopClip) by the saga that started it, so it cannot begin over home later. A cut clip clears the seeded home selection. Tests: unknown ids for every takeover kind, a player that starts inside the dispatch, the copy delay.
whenStablyReady now folds arrival.status into the shared "measure-ready" predicate: it holds for arrived, rejects on failed with the reason, and otherwise debounces as before. captureScene drops the boot waitSettled, the POST_ESC_WAIT_MS re-settle, and the pose re-apply-and-verify via readLiveCameraState — arrival already owns that wait, so a capture.pose override lands once and stays. Exhibit cards boot on #exhibit=<id> instead of hand-merging the registry's settings and pose. waitSettled, readLiveCameraState and the now-unread poseMismatch are deleted outright. No "clip frame-0 guess" was found in tools/record/record.ts to delete — its clip takes already resolve frame 0 through the recorder hook's own seenActive latch, independent of arrival; left untouched.
#exhibit=, #tour= and #clip= joined #focus= as read-only deep-link rows; the README's feature list named only the one it had before them. No stale urlPose/applyUrlPose/POST_ESC_WAIT_MS mentions remain outside this plan's own history.
… seeded wireInput seeds the runtime in the store's orientation; the arrival used to set a linked #orientation= only after waiting for that runtime, so frame 1 re-encoded the already-linked-frame commit. The arrival now lands t and orientation on the boot read and hands navigateSaga only the view.
A takeover link's rows write nothing back, so the settled body after #exhibit= differed from the URL and the write pushed; Back then reopened the exhibit and pushed again. applyNavigation now signals the burst, and the write half replaces for it, as it does for the arrival.
A cut #exhibit= starts the held exhibit's slow spin, and an exhibit card has no capture pose to stop it, so the thumbnail caught a drifting camera. The last step's dispatches also went straight into the screenshot; the shot now waits two frames first.
exhibit=, tour= and clip= now write from store state while that play runs, however it started, and drop when it ends. A hash navigation that leaves out the running key exits it; one that names it does not restart it. The registry clip's id lives on camera.clipId, written by watchClipSaga.
takeover.active is now the one record of what runs: TakeoverSource gains
{ kind: 'clip', id }, watchTakeoverSaga takes startClip as its third start
request, and exitTakeover is the one exit verb. Each kind brings its own
bracket: tours and exhibits run inside withSceneSnapshotSaga (scene
snapshot/restore + FOV pin), and a clip's body (clipBodySaga, formerly
watchClipSaga's worker) freezes and restores the clock.
Deleted: camera.clipId + clipIdChanged, selectPlayingClipId, stopClip,
watchClipSaga, selectTakeoverActive, tour.tourId (tourStarted is
payload-less; selectActiveTour reads the takeover's id). The three hash
rows write from takeover.active; applyNavigation has one keep/exit test.
The HUD still hides only for a tour or exhibit.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…a stale test comment Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Applies the safe-now bin from the branch deletion audit, plus the comment fix it flags separately: - S1: drop branch-history narration in hashParamSources.ts (main never had takeover rows, so there was nothing to narrate becoming one) - S2/S3/S4: drop three tests already covered elsewhere — navigateSaga's "focus cut commits once" (arrivalSaga's mars + late-catalog cases), watchHashReadSaga's pose-before-focus ordering (linkIntentFrom's merge + navigateSaga's focus+pose outcome), linkIntentFrom's "parses each key" (each row's read is pinned in hashParamSources.test.ts) - S5: drop withSceneSnapshotSaga's claim that a body merges its own lens after the snapshot (grep finds no fovDeg/camera: override anywhere) - S6: drop captureScene's exhibitId+focusId guard and the SceneShot doc sentence it backed — the only builder (capture.ts shotFor) sets them from a discriminated union, so both can never be defined at once - S7: drop clipBodySaga's fork-time warning, carried over from the deleted watchClipSaga — the body is only ever invoked on demand, with no fork-time alternative to warn against - ArrivalVeilContainer.tsx: fix a stale comment — since a2c270b a running takeover's key stays in the hash for the run; only canonicalization, a failed play, or an ended one drops it Skipped every needs-ruling item (R1-R6) per instruction. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
DoD audit READY: suite 14973 green, tsc clean, 22/22 boxes, smoke attested by the user, no new TODOs, deletion audit applied (safe-now a99ad93, R2 fold 4309e0e; R1/R3/R4 declined). Backlogged: afterTwoFrames -> renderScheduler.nextFrame, npm run shot. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…pen-on-subject # Conflicts: # docs/BACKLOG.md # src/services/engine/phases/startLoop.ts
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Spec:
docs/superpowers/specs/2026-10-01-deep-link-arrival-design.mdEvery deep link opens with the camera already on its subject (no fly-in), behind a veil until the subject resolves; adds read-only
#exhibit=,#tour=,#clip=links;readynow means "arrived", so capture tools drop their timeouts and pose re-application.Draft — spec under review; plan + implementation to follow on this branch (prep commits first).
🤖 Generated with Claude Code