fix(studio): make static-seek adapter honor keepPlaying option

[calcarazgre646]

Summary

createStaticSeekPlaybackAdapter.seek now accepts the { keepPlaying } option declared by the PlaybackAdapter contract (playbackTypes.ts:10) and aligns its default-pause semantics with wrapTimeline.seek (recently hardened in 3e7b464b, "ensure GSAP timeline stays paused after seek"). Same call sites, same wrapTimeline default — now the static-seek path matches.

Background

This is the follow-up I called out as "Out of scope" in #863 and that @jrusso1020 invited in review:

"createStaticSeekPlaybackAdapter type drift (out of scope, called out in PR). Worth tightening the type now — it's a one-line change and the body already preserves playback by accident. Leaving the type signature mismatched is a footgun for the next person who reads the adapter and thinks keepPlaying is unhandled. Happy to follow up in a separate PR if you'd rather keep this one tight." — #863 review

What was wrong

The adapter's seek signature was (time) => void, narrower than the PlaybackAdapter.seek contract. The body also had an asymmetric default vs wrapTimeline.seek:

• wrapTimeline.seek (line 137, post-3e7b464b): default seek tl.pause() → tl.seek() → tl.pause(). keepPlaying: true skips both pauses.
• createStaticSeekPlaybackAdapter.seek (before this PR): default seek kept playing = true, so the RAF ticker continued calling renderSeek in the iframe even when the public useTimelinePlayer.seek wrapper had already called stopRAFLoop() + setIsPlaying(false).

User-observable on non-GSAP compositions (the path where this adapter is used — useTimelinePlayer.ts:160 fallback when win.__timelines is absent): scrubbing the timeline during playback would leave the iframe content silently advancing while the UI reported "paused".

Fix

packages/studio/src/player/lib/playbackAdapter.ts — 7 lines:

seek: (time, options) => {
  renderSeek(time);
  if (options?.keepPlaying) {
    if (playing) {
      playStartTime = currentTime;
      playStartNow = clock.now();
    }
    return;
  }
  // Default seek aligns with wrapTimeline: stop the RAF ticker so the
  // adapter's `playing` flag matches the public seek contract instead of
  // silently driving renderSeek in the background.
  playing = false;
  stopTicker();
},

keepPlaying: true keeps the existing "rebase playStartTime/playStartNow" behavior so the next tick renders from the new position. Default and explicit keepPlaying: false now clear playing + cancel the RAF.

Internal callers checked

All adapter.seek() call sites in packages/studio/src/player/hooks/useTimelinePlayer.ts are safe under the new default:

• useTimelinePlayer.ts:199 (forward loop wrap-around): immediately calls adapter.play() after the seek, so a pause-on-default seek is replayed in the same tick.
• useTimelinePlayer.ts:259 (replay from end inside play()): same — followed by adapter.play().
• useTimelinePlayer.ts:286 (playBackward init): preceded by an explicit adapter.pause(), so default-pause is a no-op.
• useTimelinePlayer.ts:307 (backward loop end, no-loop terminal): immediately followed by setIsPlaying(false); default-pause is the desired terminal state.
• useTimelinePlayer.ts:316 (backward tick mid-traversal): the reverse path drives playback through its own RAF (reverseRafRef), not the static-seek internal tick, so cancelling the static-seek tick on every reverse step is harmless.
• useTimelinePlayer.ts:362 (public seek(time, options)): forwards options straight through. keepPlaying: true from A/E shortcuts now reaches the static-seek adapter end-to-end.
• useTimelineSyncCallbacks.ts:161 (initial adapter setup): adapter is paused at construction; default-pause is consistent.

Tests

packages/studio/src/player/lib/playbackAdapter.test.ts only covered wrapTimeline (no coverage for createStaticSeekPlaybackAdapter). Adds 7 tests under a new describe("createStaticSeekPlaybackAdapter seek keepPlaying option") block, driving the adapter with a deterministic fake StaticSeekPlaybackClock:

• default seek stops the RAF ticker (isPlaying() === false, cancelAnimationFrame observed)
• default seek prevents queued frames from advancing renderSeek beyond the seek target
• keepPlaying: true preserves playback and rebases the ticker so the next frame renders from the new position
• explicit keepPlaying: false matches the default
• keepPlaying: true does not force playback when the adapter is already paused (intent is preserve, not force play)
• back-compat: calling seek(t) without options matches the contract (still pauses)
• default seek clamps overshoot to duration and still pauses

Validation

Check	Result
bun run --cwd packages/studio test	612/612 ✅ (+7)
bun run --cwd packages/studio typecheck	clean
bunx oxlint (touched files)	0/0
bunx oxfmt --check (touched files)	clean
Lefthook (filesize + lint + format + typecheck + fallow + commitlint)	✅
Fallow audit	✓ No issues in 2 changed files

Cross-PR check

Branched from upstream/main at 3cd6cd6a. Of the 30 open PRs, only #1085 (@miguel-heygen) touches the seek path — it refactors useTimelinePlayer.ts to extract shouldResumeForwardPlaybackAfterSeek / shouldStopAfterSeek helpers but does not touch playbackAdapter.ts. Zero file overlap. #1085's new public-seek logic still forwards options to adapter.seek(nextTime, options), so the contract this PR establishes is what that refactor relies on.

Out of scope (intentionally)

The remaining nits from @jrusso1020's #863 review (sibling play→pause→play churn in seekMasterAndSiblingTimelinesDeterministically, missing onDeterministicPause emit on seek(), setIsPlaying(true) asymmetry in the keepPlaying branch of runtime/player.ts) are independent of this adapter and not addressed here.

[miguel-heygen]

Approved. I reviewed the live head fd2a97e, ran the focused Studio playback tests/typecheck/lint locally, and smoke-tested the keep-playing seek path in Studio with agent-browser. No blockers from my side.\n\nNote before merge: the commit on this PR is currently unverified by GitHub (verification reason: unknown_key). The PR author needs to verify/sign the commit(s) so the branch protection checks can allow the PR to merge.

[calcarazgre646]

Thanks @miguel-heygen — signing key added; fd2a97ea now verified by GitHub. Should be unblocked for merge.