Repository navigation
The REVIEW page - #177
The REVIEW page#177
Conversation
Quest-Claim: e3d36e2c89eab2a2
… recordings REVIEW replaces the Replay panel and the recording half of the Output panel, as quest/m0/ux/mock/review.html draws it. The last 30 seconds of the program or a source show as a strip of keyframes with the sound under them, decoded on a thread of their own while the page is shown, from the groups the tracks already keep. Two amber handles pick any stretch of it for Replay and Save clip, and the picture under the scrub head shows large. Beside it, this show's recordings and clips as cards with their length and size, one repaired after a crash marked so, and the folder. The title bar's record button shows its clock, and the recorded picture, sound, and each source too move to Settings, Recording. Clips are saved into the recordings folder, beside the recordings. 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>
The REVIEW page's quest is done: pictures on finished cards wait on moq-mux's regular MP4 import (m3, as moq's import is), and the human feel check is the maintainer's try-review quest. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- An aged-out stretch's end is taken by a press on its side, so it can always be dragged free. - Strip picture ids stay unique across the strip's workers, so a texture the page kept never shows an older picture. - The shelf runs one listing at a time. - With more than one replay source, which one Replay plays in is picked again, as before this page. 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>
|
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 23 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.
Important
The Replay and Save clip hotkeys keep reusing a stretch picked between two moments: they replay that old moment, and stop working entirely 30 s after the pick. The README also still says REVIEW holds the recording's picks in two places this PR didn't touch.
Reviewed changes
This review covers the whole PR at de36545: the new REVIEW page, the strip worker, the folder listing, the stretch model for replays and clips, the Settings Recording tab, and the quest files.
replay::Stretchreplaces fixed lengths: two handles pick either the last N seconds (Last) or a fixed span between two moments (Between).takeshifts each target's end back by the pickedago.past.rsstrip worker: while REVIEW is shown, a thread re-takes the 30 s window about once a second. It decodes each group's keyframe at 480 px on the CPU, caches them by timestamp, and reads a sound peak per packet.shelf.rsfolder listing: the recordings folder is listed in the background every 2 s. File names are grouped into recording, repair and clip cards, lengths come frommvhdwithmdatskipped by seeking, and free space is shown.- REVIEW page rewrite: large picture under the scrub head, a "replay what" column, the strip with handles and ruler, speed, an "Into" picker only when there is more than one replay source, and Replay/Save clip buttons that say why they can't act. Cards for the recording under way and for finished files sit on the right.
- Clips move to the recordings folder, the Settings Recording tab takes over the recording picks, and the title bar's record button shows elapsed time.
- Quests: deletes
quest/m0/ux/review.mdand adds thetry-review,card-picturesandpin-mp4-importfollow-ups.
I could not run quest context (no quest or nix in the review shell), as REVIEW.md asks me to note. The UI holds controls.replay and then locks engine.state; Controls::replay takes them in the same order, and nothing takes them the other way round, so there is no new deadlock risk.
ℹ️ README still says REVIEW holds the recording's picks
This PR moves the recording's picture and sound picks to Settings' Recording tab. Two README passages outside the diff still point at REVIEW, which goes against AGENTS.md's rule that the README changes in the same PR.
Technical details
# Stale README references to REVIEW's recording picks
## Affected sites
- README.md:416 — "A recording holds the audio scene on program unless REVIEW picks another": the audio scene a recording holds is now picked in Settings' Recording tab (`Panels::recorded_audio` in `rs/moq-studio/src/ui/settings.rs`).
- README.md:1201 (Hotkeys) — "Replay, Save clip, and a recording use REVIEW's picks": a recording now uses Settings' Recording picks; only Replay and Save clip use REVIEW's.
## Required outcome
- Both passages name Settings' Recording tab for the recording's picks, matching the already updated Recording section.claude-opus-5-5 | 𝕏
A stretch fixed between two moments on REVIEW kept a hotkey replaying that old moment, and refusing once it aged out of the window. A key pressed from another page means what just happened. The README no longer says REVIEW holds the recording's picks, and names the replay picker. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
This review covers the two commits since the prior Pullfrog review at de36545.
- Hotkeys and the clip button take the last stretch:
Controls::takenow callsStretch::last()whenpressed, so aStretch::BetweenbecomesLast(length)(capped atWINDOW) and no longer ages out. REVIEW's own buttons still honorBetween. - Test for the new behavior:
a_hotkey_takes_the_last_stretch_as_longcovers the conversion, its span long after the pick, andLastpassing through unchanged. - README and tooltip: the Hotkeys paragraph and the shell clip button's tooltip now describe the behavior, and the two stale "REVIEW picks" lines about recordings point at Settings' Recording tab.
Both concerns from the earlier review are addressed.
claude-sonnet-5-5 | 𝕏
# Conflicts: # README.md # rs/moq-studio/src/controls.rs # rs/moq-studio/src/kit/icon.rs # rs/moq-studio/src/shots.rs # rs/moq-studio/src/ui.rs # rs/moq-studio/src/ui/settings.rs # rs/moq-studio/src/ui/shell.rs
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>
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
This review covers what changed since the last Pullfrog review at 55d98ab: a merge of main, a pair of commits that hosted the views and then removed them, and a wording change to a quest file.
- Merged
main(0ee8deb): I checked the conflict hunks (git show --cc 6fa235b) and each one keeps both sides.- In the README, the page table and Settings tabs take main's wording plus REVIEW's row and the Recording tab.
- The Hotkeys paragraph drops the old sentence about
Cut/Transition, since main now offers them as TAKE and AUTO. Icon::ALLnow has 53 icons: main's 50 plusFolder,SaveandRepaired.Setting::ALLbecomesShow, Recording, Hotkeys, Effects, Advanced.- In
shell.rs,Page::Reviewgoes toself.review, and main's MIX keepsaudio_scenes. shots.rskeeps theReviewviews beside main'sCuedandOffAir.
- Hosted and dropped the shots: the views were committed as WebP and deleted in the next commit, as AGENTS.md asks.
quest/m3/card-pictures.md: now records that the maintainer kept the recommended approach.
claude-opus-5-5 | 𝕏
Brings in BUILD's layers and inspector (#174), Studio mode (#176), REVIEW and Settings' Recording tab (#177), the moq pin and Flatpak camera fix (#173), and the quest cleanup (#185). The copied source tracks take main's group starts and skips while still stopping at END STREAM; BUILD's picture frames the program by whether it is on air and a scene open off program as main does; Settings has both Stream and Recording; and the hotkeys add Go live and End stream to TAKE, AUTO, and Studio mode. Records the maintainer's choice for program-encoder-priority.md: release the sources' replay encoders when the program's cannot open. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Claims, implements, and deletes quest/m0/ux/review.md: the REVIEW page from
mock/review.html.Summary
REVIEW is now two halves, as in the mock-up:
Also part of this PR:
<show>-clip-<n>.mp4, so they show up on the shelf. That folder still defaults to the one beside the scene file.settings-recordingframe.rs/moq-studio/src/past.rs(the strip's decode worker) andrs/moq-studio/src/shelf.rs(the folder listing).replay::Stretchreplaces the fixed lengths.The quest file is deleted: the two items left on it became follow-up quests (see Decisions and Follow-ups).
Views
reviewreview_picked(a stretch picked, no replay source yet, so the refusal shows)review_into(a show with two replay sources: which one Replay plays in is picked under the speed, in place of its note)recording(Settings, beside main's Show, Hotkeys, Effects, Advanced)kit(new icons: Folder, Save, Repaired)Public API and wire impact
None. moq-studio is a binary; nothing in moq-scene or the moq submodule changes, and no wire format changes. The user-visible behavior change is that clips are saved into the recordings folder instead of beside the scene file. Both are the same folder unless
--recordingsis set. README covers it.Choices made
Stretch::Last); otherwise it is fixed to two instants (Stretch::Between) and is refused once it ages out of the 30 s window.xdg-open,open,explorer). There is no in-app playback of finished files.settings-recordingframe. The recording options (picture, sound, each source) moved there from the old recording panel on REVIEW, since the record button works from any page.LAST_IDinpast.rs), and the shelf runs one listing at a time, both from review round 1.grip_at).just shots' settle skips an idle replay source's track, which starts only with its first replay (Feed::publishin moq-scene runs fromreplay).mvhdviamp4-atom(already a dev-dependency, now a dependency), skippingmdatby seeking, so listing stays cheap on large files.statvfs(rustix) on unix andGetDiskFreeSpaceExWon Windows.Validation
Merged
main(BUILD with layers and the inspector, moq-dev/studio#174; Studio mode and editing off air, moq-dev/studio#176; panic hides the top layer, moq-dev/studio#179). REVIEW replaces main's replay-and-record panels as before; the Recording tab joins Settings' Show, Hotkeys, Effects, Advanced; theReplayhotkeys sit beside the newStudioModeaction; README keeps main's page rows and wording (layers, always on top) with REVIEW's row. On the merged headjust cipasses (347 tests) andjust shotsrenders all 32 views; the four views above are re-rendered from it.just cipassed before the merge (340 tests), including clippy with thecapture,pipewire,vaapi,openh264features.the_program_reaches_a_relay(lavapipe, openh264) now also checks the strip: its pictures decode from the live program's keyframes, are the canvas red, 480 wide, in order, and its sound peaks are in range and in order.New unit tests:
a_press_takes_the_handle_on_its_side,a_hotkey_takes_the_last_stretch_as_long,stretch_tests(spans, clamping, aging out, empty stretches),shelf::names_read_backandthe_folder_lists_cards(name parsing, repairs, clips, mvhd lengths), and the review sizes, clock and cover helpers.just shotsrendersreview,review_picked,review_into,recordingandkitstill (each renders twice and matches).UI frame CPU in a release build over 120 kittest steps: LIVE 68.87 µs, REVIEW 92.63 µs per frame. I used a throwaway test for this and did not commit it.
UI frame time under headless sway (sway 1.12, 1440x900@60, Mesa 26.3 RADV on the host's AMD Radeon 610M; no NVIDIA GPU, so the program is encoded with openh264 in software). Both builds are release with
openh264, opened on REVIEW, publishing to a localmoq-relay(--connect https://localhost:4443/anon), onshots/show.tomlplus a loopingclip.mp4in Main and areplaysource. On this branch the strip fills with the program's keyframes, and Replay and Save clip are enabled. An uncommitted patch logged the frame interval and the UI thread's on-CPU time (/proc/thread-self/schedstat) per 300 frames, dropping the first two windows; each thread's on-CPU ticks (100 Hz) were read 55 s in. Four alternating 60 s rounds, host load average 28 to 45 from other agents.main(99d8c3b) REVIEW: interval / UI thread per framereview-pastBoth hold the display's 60 fps. The new page costs the UI thread about 0.3 ms more per frame (1.39 ms against 1.06 ms, 8% of a 16.7 ms frame), and the strip's worker about 2 to 2.5% of one core while REVIEW is shown. The software encoder dominates both builds (1050 to 1350 ticks).
That sway run is also the real app against a relay; scrubbing and picking by hand are left to try-review, since an agent cannot judge feel.
Decisions
Taken by the agent on the maintainer's behalf while they were away, as recommended; each is for review:
mp4-atom, since AGENTS.md puts demuxing in moq; dropping the pictures from the mock-up). That import is not on moq'smain, so this PR keeps the placeholder tile and the pictures become follow-ups. They sit in m3 beside moq's import, which the maintainer placed in moq's m3 on 2026-11-05; a quest cannot require a later milestone, so moving them up means moving the import up too.Maintainer:item, liketry-live.md(rejected: holding this PR a draft for it).Follow-ups
moqpin once moq-mux's regular MP4 import lands on moq'smain. Waits on moq-dev/moq. Kept by the maintainer (2026-11-15).mp4-atom).Maintainer:item. Kept by the maintainer (2026-11-15).quest/m2/markers.mdalready owns markers on the strip.Models
Size: M
Implementer: Claude Code, claude-opus-5-5, unknown effort (arm A)
Implementer: Claude Code, claude-opus-5-5, unknown effort (iterate, M, arm A)
Implementer: Claude Code, claude-opus-5-5, unknown effort (iterate, S, arm A)
(Written by Claude Opus 5.5)
🤖 Generated with Claude Code
Review
Rewritten by
quest reviewafter each round.Head: e6e2fb7
Base: origin/main at 0ee8deb
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:
01a1e0d3-581f-79d2-b198-b0e27f92e441, round 5Score: 10/10,
The merge preserves the reviewed REVIEW behavior, and code inspection found no new issues in the branch’s incremental changes; builds and tests were not run under the review constraints.Rounds: 5; first round 4 (high 1, medium 3); findings 5 (high 1, medium 4): fixed 5, refuted 0, open 0
Findings: none
Reviewer: Codex, gpt-6.1-sol, medium effort, shadow
Session:
01a1e0d3-5821-7a22-be10-458e2ef7e8efRounds: 2; first round 3 (high 1, medium 2); findings 3 (high 1, medium 2): fixed 3, refuted 0, open 0