Repository navigation
feat(studio): markers on REVIEW, with a clip around each - #209
Conversation
Quest-Claim: aab052a34eb3b900
… each A key or the flag beside the panic key drops a marker. REVIEW lists the session's markers to name, clip around, or remove, and shows those in the last 30 seconds on its strip. They are kept beside the recordings in <show> <stamp>.markers.toml. Chapters wait on moq-mux; the quest now holds that work. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…in the background A name typed into a marker's row is kept when another row's name takes the focus, and a field keeps its marker when a new one is dropped above it. The markers file is written on a thread of its own, the latest list each time, so a slow disk never holds up the window or a hotkey; REVIEW says why a write failed until one succeeds. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 56 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. |
There was a problem hiding this comment.
ℹ️ No critical issues — one rough edge inline.
Approach: sound. Markers are one row in the action table, so Hotkeys, the palette, and the portal pick them up for free. A clip around a marker goes through the same Controls::take copy that Save clip uses, and the file is written as a whole snapshot on a background thread. Splitting chapters out to moq-mux follows AGENTS.md's Upstream rule.
Reviewed changes — the full PR at 22f880f: the new action, the markers store and writer, the stretch around a marker, the REVIEW list and strip pennants, the shell flag, the window tests, and the quest and README updates.
Action::Marker: "Drop a marker" sits in the Review group. It firesMarkers::markwith the clock's now and is refused once 255 markers are held.markers.rs: numbers are never reused. Each change queues a whole-file TOML write, and only the latest waiting one is kept. The writer thread uses a temp file and a rename, removes the file when the list is empty, and keeps showing the last failure until a write succeeds.Dropwaits for the final write.Stretch::around: a clip runs from 10 s before the marker to 5 s after it, stopping at now. A marker older than the 30 s window is refused.takenow takes the stretch directly.- REVIEW: marker rows show newest first, each with a name field whose ID and draft are keyed by marker number. A draft is saved when its field loses focus. Numbered pennants mark the strip.
- Shell: a flag button sits beside the panic key's scissors in both layouts.
claude-opus-5-5 | 𝕏
Save clip already refused off air, when the program keeps no past; the scissors on a marker's row checked only its age, so they failed with a toast on click instead of saying why. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Save clip, Replay, and a marker's scissors refused whenever REVIEW's strip could not decode the pick's pictures, though a replay or a clip copies the past without decoding it there. The pictures now carry their own error. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
✅ No new issues found. The marker scissors now refuse exactly when Save clip does.
Reviewed changes: the commits since 22f880f. They fix the open scissors finding, merge main, and move the strip's decode failure off held.
- Scissors follow the held past:
refusals()now keepsheldonRefusals, andreplays()returns it.clip_refusalputs that reason ahead of the marker's age, so marker scissors are disabled off air with Save clip's message. - Decode failure no longer means "nothing held":
past::View::picturesis now aResult, soDecoding::readrecords a picture-decode error there and still reportsheldasOk. Replay, Save clip, and the scissors copy the take without decoding it, so they are no longer refused. The large picture and the strip show the decode error in place of "No pictures held". - Tests: added the unit tests
a_marker_is_clipped_only_while_its_past_is_heldandpictures_it_cannot_decode_leave_the_past_held. The window testa_marker_is_named_where_it_is_typednow waits off air until every row's scissors turn off. Theengine.rstest is adapted to theResult.
claude-opus-5-5 | 𝕏
|
Maintainer decision (/quest-complete): /quest-merge. Keep quest/m2/pin-mp4-chapters.md, quest/m3/marker-clips.md and quest/m2/try-markers.md as planned, and the moq-mux chapters follow-up for /quest-plan in moq-dev/moq. (Written by Claude Opus 5.5) |
Markers (#209) added Action::Marker; a remote keeps to scene pads, mutes, TAKE and AUTO, so it joins Replay, Save clip and Record as desktop-only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Works on quest/m2/markers.md: markers and a clip around each land here; chapters wait on moq-mux, and the quest now holds that work.
Problem
An operator has no way to mark a moment as the show runs and come back to it: Save clip takes only what just happened, as long as the stretch REVIEW picked.
Approach
controls.rs): one variant and one row, so it shows in Hotkeys and the palette. The flag beside the panic key's scissors fires it too.markers.rsholds the session's markers: a number counted from 1 and never reused, the moment on the clock that stamps the program and every source's past, the wall time, and an optional name. Each change rewrites<show> <stamp>.markers.tomlin the recordings folder on a thread of its own, the latest list each time, through a temp file and a rename, so a slow disk never holds up the window or a hotkey; the file goes once no marker is left. A session holds at most 255 markers, names at most 60 characters.Stretch::around(replay.rs): from 10 s before to 5 s after, cut at now, fed to Save clip's own copy (Controls::take), so nothing is re-encoded and the same keyframe limits apply.Views (
just shots), hosted at 486ede3; no mock-up frame covers markers, so the page follows REVIEW's mock-up for type and spacing:review_markers: three markers, two on the strip, one named, the oldest too old to clipreview: none yet, saying how to drop onelive: the flag beside the scissorslive_portrait: the same in the tall arrangementsettings_hotkeys: Drop a marker under ReviewImpact
Marker.<show> <stamp>.markers.toml, named for when the studio opened the show..unfinishedfile is now called a sidecar, so "marker" means one thing.Alternatives
moovwith mp4-atom: the studio cannot tell where the movie starts on the program clock, which only moq-mux's export knows, and on moq'smainFinishis opaque, so the edit would break at the next bump. Rejected for a moq-mux change, asAGENTS.mdasks.Decisions
Taken while the maintainer was away; each is easy to change in review:
origin/mainis merged (#194 keep out of the way, #200 landscape and portrait at once), as the iterate brief asked; it merged without conflicts.past::View::picturescarries its own error, likesound, andheldfails only when the take does, since a replay or a clip copies the past without decoding it there. Found by the local review once the scissors followedheld; it was already true of Save clip and Replay onmain. At that review's design checkpoint (its third round with findings), the approach was kept: the scissors refuse exactly when Save clip does, and the one cause that should not refuse either is fixed at its source.Caused by
Follow-ups
/quest-planin moq-dev/moq: moq-mux's regular MP4 finish (Export::finish,repair) takes chapters, each a moment on the broadcast's timeline and a title, written into themoov(mp4-atom'schpl).Validation
just ci: passed on the merged head (492 tests).just shots, every view and window test on the merged head: passed (88);live,live_portrait, andreview_markerschecked by eye after the merge.just shotsfor the views above and the window testsa_click_takes_and_the_panic_key_cutsandthe_palette_finds_and_runs_anything: passed.shots::a_marker_is_clipped_around(nightly, withjust shots, about 20 s): goes live to a local relay, clicks the flag, waits, clicks the scissors on REVIEW, and reads the clip. Checked independently with ffprobe on a copy of that run's clip: video 16.200 s (h264, from the keyframe 1.19 s before the span), sound 15.02 s (opus) against the 15 s span; the markers file heldnumber = 1and the UTC moment.shots::a_marker_is_named_where_it_is_typed(nightly): a name typed into one row survives clicking another row's name, and a marker dropped by the hotkey path mid-typing leaves the typing on its marker; it fails with the field's stable id removed. Off air, it also waits for all three rows' scissors to turn off; with the held-past refusal removed fromclip_refusal, it times out and fails, as does the unit testa_marker_is_clipped_only_while_its_past_is_held.past::tests::pictures_it_cannot_decode_leave_the_past_held: a take whose avc1 description is no avcC stays held, its pictures an error; it fails with the old early return put back.engine::tests::the_program_reaches_a_relay,past::tests, and thereview,review_picked, andreview_markersviews: passed after that change, andjust ciagain (493 tests).Models
Size: M
Implementer: Claude Code, claude-opus-5-5, high effort (arm B)
Implementer: Claude Code, claude-opus-5-5, high effort (iterate, S, arm B)
Review
Rewritten by
quest reviewafter each round.Head: 4d998bf
Base: origin/main at 5ae83b0
Risk: medium,
the base has no RISK.md to name what is always high or customer-facing, so every change is mediumClassified: medium by default
Reviewer: Codex, gpt-6.1-sol, high effort
Session:
01a1f22d-da60-7540-8aa8-c3afebee9da8, round 4Score: 10/10,
The preview-failure finding is fixed, and the incremental code review found no new issues; builds and tests were not run as instructed.Rounds: 4; first round 3 (high 1, medium 2); findings 4 (high 1, medium 3): fixed 4, refuted 0, open 0
Findings: none
The last round's findings, at c02b8a0:
rs/moq-studio/src/ui/review.rs:877:Separate preview failures from clip availability;In rs/moq-studio/src/past.rs, Decoding::read now records preview failures in pictures while preserving held as successful, so they no longer disable marker clips; REVIEW displays those preview errors separately.Reviewer: Codex, gpt-6.1-sol, medium effort, shadow
Session:
01a1f22d-da66-7732-8c71-cec3ef562108Rounds: 2; first round 2 (high 1, medium 1); findings 2 (high 1, medium 1): fixed 2, refuted 0, open 0