While a recorder is running, the capture container tees an HLS
stream into /live/<assetId>/ alongside the ProRes master upload.
The asset row is pre-created at recorder start with status='live'
so the clip appears in the library immediately. /api/v1/assets/:id/stream
returns the HLS playlist URL until recording stops, then proxy.
* docker-compose: shared wild-dragon-live mount on api/capture/web-ui
* migration 001-add-live-status: idempotent ALTER TYPE for asset_status
* mam-api: runMigrations() on boot; recorders.js pre-creates live asset
+ passes ASSET_ID; assets.js POST upserts on existing live row instead
of inserting a duplicate, and stream route returns HLS for live assets
* capture: parallel HLS ffmpeg into /live/<assetId>/; ASSET_ID env
* web-ui: nginx serves /live/, preview.js loads hls.js, LIVE badge added
Clicking Edit on the preview modal worked, but the user only saw an empty editor for ~25s while the recovery + format-chooser cycle ran and the bridge waited for a stable project. Looked broken. Now: a centered top banner appears the moment the bridge detects ?asset=, reads Loading clip from Z-AMPP MAM, switches to Clip ready in media bin on success, or surfaces the failure. Project-stability gate tightened from 1500ms to 600ms so the import lands sooner.
Three problems were blocking the round-trip. Each fixed.
* MediaBridge.importFromURL went through the file-import service but not the Zustand store, so the media bin stayed empty. mam-bridge now calls window.__projectStore.getState().importMedia(file) which is what the actual UI uses. project-store.ts exposes useProjectStore on window for that hook.
* rustfs serves the proxy with content-type application/octet-stream; the editor rejects with DECODE_ERROR on that mime. Bridge now forces video/mp4 (or audio/wav, video/webm, etc.) based on the asset filename.
* The Recover Your Work modal and the Welcome tour blocked editor initialization. Bridge now auto-clicks Start Fresh and Skip Tour (alongside the format chooser), and waits 1.5s of project-id stability before calling importMedia so it does not get clobbered by the project-replacement cycle. One-shot guard prevents duplicate imports.
Vendored Augani/openreel-video (MIT) into services/editor and wired it to the MAM. Editor runs as its own container on port 47435. Library assets pull in via ?asset=<uuid>; render exports route back via POST /api/v1/upload/simple. Sidebar Editor link on every page; Edit button on every preview modal. See services/editor/INTEGRATION.md for the patch map.
A thumbnail job from earlier stayed 'active' for 6+ hours: worker was restarted at 70% progress, BullMQ left it in the active set, and there was no stall reaper because the worker was created with only the default options.
Worker now passes stalledInterval: 30000, lockDuration: 60000, lockRenewTime: 15000, maxStalledCount: 1 to the Worker constructor. If a run dies, BullMQ reclaims the job back to waiting within 30s and a 'stalled' event is logged. Otherwise the lock is renewed mid-job.
Jobs UI gains a 'Kill' button per row next to Details. Calls DELETE /api/v1/jobs/:id which already removes the job from Redis. Use it on any row that looks stuck.
Confirmed shape brief: ops-console direction, balanced density, blue committed accent, readability emphasized.
Design system: surface scale to 5 steps tinted toward hue 266; text contrast lifted (secondary 62->72%, tertiary 44->52, added text-disabled); borders gained faint variant; status tokens renamed semantically (signal-good/warn/bad/idle); typography upgraded (Inter 400/500/600, JetBrains Mono for numerics, new 2xl/3xl/tc scale steps).
New components: tc-display timecode classes with blue glow; tally-word with good/warn/bad/rec variants; signal-strip flutter bar with modifiers; chip dense monospaced pill; manifest table styles.
Topbar status strip: new js/topbar-strip.js auto-injects a 28px strip on every page with wall clock, page name, live API latency, and system-pulse dot.
Per-page: recorders gain live signal-strip above fps line; capture timecode bumped 38->64px with blue glow; ingest drop zone slimmed 200->88px side-by-side layout so manifest gets the real estate.
Two things that together stop bogus URLs from masquerading as a recording:
PROBE BUTTON in the New Recorder panel. Before you commit to record, hit Probe Source - the capture container runs ffprobe with a 10s timeout against the URL and returns the parsed streams. UI shows green Signal Detected with codec/resolution/fps/audio, or red No Signal Detected with the actual ffprobe error message. For SDI it lists DeckLink devices. Listener-mode sources cannot be probed standalone (would block waiting for a publisher) and the UI says so.
MAIN STATUS LABEL ON THE RECORDING CARD now mirrors the live signal instead of hardcoding Recording. So a recorder pointed at a dead URL goes Connecting... -> Connection error (red) instead of looking like everything is fine. When frames actually start arriving the label flips to Recording (blue) and the dot turns blue. If a previously-good stream drops the label switches to Signal lost (red).
API:
* capture: POST /capture/probe runs ffprobe and returns { ok, streams, format, error? }
* mam-api: POST /api/v1/recorders/probe proxies through to the capture sidecar with a 15s outer timeout
The source image had a black border baked in. Knocked out the dark pixels into an alpha channel so the figure now floats on whatever surface is behind it — the dark gradient on the splash, the panel surface on the loading indicator, anywhere.
Pipeline: source -> resize 1200w -> python/PIL alpha-from-luminance with soft 22-55 luminance ramp -> 8-bit RGBA PNG (267KB).
* Library cards now show a checkbox on hover (and persistent when selected). Click checkbox = toggle, shift-click = range. Plain click on a card with an active selection extends/shrinks the selection instead of opening preview. Floating pill at the bottom shows count + Move / Copy / Delete / Clear. Move + Copy open a tiny bin picker (current project, default to current bin).
* mam-api/routes/assets.js: PATCH /:id now also accepts bin_id (null = move out of bin). New POST /:id/copy makes a reference-copy of the asset row (same S3 keys, new id) into the target bin/project.
* api.js: moveAsset(id, binId) and copyAsset(id, {binId, projectId}) helpers.
* All accent tokens swapped from the amber oklch(76% 0.178 52) to the Wild Dragon signature blue oklch(55% 0.20 266) = #1f3ad0 ish. Login splash + first-load splash + signal-receiving + button primary all picked it up automatically through common.css.
* Loading indicator across the app uses the AMPP Safe hardhat photo gently pulsing with a tiny blue dot underneath. .ampp-loading component lives in common.css with --sm / --xs / --inline variants. Replaces the plain "Loading assets…" empty state in index.html.
Adds the BMG-branded "AMPP Safe" hardhat photo as the visual identity for the auth + first-load surfaces.
* services/web-ui/public/img/ampp-safe.jpg (52 KB, 1200w optimized JPEG)
* services/web-ui/public/login.html: full redesign as a two-column hero + sign-in panel. Hero shows the hardhat photo full-bleed with a subtle AMPP Safe pill badge and broadcast-safe caption. Login + first-run admin setup forms unchanged functionally.
* services/web-ui/public/index.html: brief first-visit splash overlay (~1.4s) using the same image. Dismisses to the library and uses sessionStorage so it only shows once per session.
Trash icon in the library was firing PATCH /assets/:id with {status:"deleted"}. The PATCH route only accepts display_name/tags/notes so it returned "No fields to update" and the asset stayed put.
* api.js: add deleteAsset(id, {hard}) helper hitting the real DELETE route.
* index.html: deleteAssetPrompt now calls deleteAsset (soft archive). Confirm dialog reworded to match.
* mam-api/routes/assets.js: list endpoint hides status=archived by default. Pass ?include_archived=true to see them in a future restore-from-trash view. Filtering by ?status=archived still works for power users.
* All HTML: bump api.js cache-buster v=4 -> v=5 so the new helper is fetched.
Click any asset card to open a modal with the H.264 proxy playing inline (or audio/image, per media_type). Esc or click outside closes. Sidebar shows status/codec/resolution/fps/duration/size/created plus tags and notes.
Plugin install side: added install-windows.ps1 that copies the CEP panel to %APPDATA%\Adobe\CEP\extensions, flips PlayerDebugMode=1 across the CSXS.8-13 hives, and prints the next steps. Plugin already wired against the current API.
* services/web-ui/public/js/preview.js: standalone IIFE that lazy-injects the modal markup + CSS on first use. Renders <video controls> (or <audio>, <img>) sourced from /api/v1/assets/:id/stream, with sidebar from /api/v1/assets/:id. Falls back to a clear empty state when proxy is still processing.
* services/web-ui/public/index.html: loads preview.js, wires asset-card click to window.openAssetPreview(asset.id), guards against delete-button clicks bubbling.
* services/premiere-plugin/install-windows.ps1: one-shot Windows installer for the CEP extension.
The api.js library-list fix from the previous commit never reached the browser because nginx served all .js with `Cache-Control: public, immutable; max-age=31536000`. The HTML referenced api.js with no version query, so the browser kept its year-cached buggy copy.
* nginx.conf: drop .js from the immutable long-cache block, add a no-cache must-revalidate block so future redeploys are picked up immediately.
* All HTML files: tag api.js refs with ?v=4 so already-running browsers fetch the new version on next page load.
Three problems blocked the end-to-end flow:
1) Library always rendered empty because /assets returns {assets,total} but
index.html (and capture.html) assumed r.data was an array. Fixed in
api.js by unwrapping r.data.assets centrally; total is kept on r.total.
2) SRT/RTMP caller mode pulled audio only. ffmpeg opened the network input
before the H264 SPS arrived, marked the video stream as pix_fmt=none,
and silently dropped it from the stream map. Added -probesize 32M
-analyzeduration 10M -fflags +genpts and explicit -map 0✌️0?/0🅰️0? so
each track survives independently of when it appears.
3) Hitting Record gave no feedback about whether a stream was actually
arriving. capture-manager now parses ffmpeg progress lines (frame=...
fps=...) and tracks framesReceived, currentFps, lastFrameAt, lastError.
getStatus() returns a derived signal enum (connecting | receiving |
lost | error | stopped). The recorder controller gives each spawned
container a stable network alias `recorder-<id>` and the GET
/recorders/:id/status endpoint proxies the live capture status through.
recorders.html polls that every 2s and renders the badge under each
active card with the running frame/fps counter or the ffmpeg error.
Also:
* recorders.html: dropped the listener-mode UI entirely. All new recorders
are caller-mode (pull). The MAM is no longer offered as an RTMP/SRT
server. Legacy listener records still render but read-only.
api.js sends parts as { partNumber, ETag } (uppercase) but upload.js
was reading p.etag (lowercase), resulting in undefined ETag passed to
S3 CompleteMultipartUpload → InvalidPart error on all large file uploads.
Also handle both casings defensively.
GET /api/v1/jobs now queries the proxy, thumbnail, and conform BullMQ
queues directly and returns normalized job objects with id, type,
status, progress, asset_id, timestamps, and error fields.
Also adds DELETE /:id to remove completed/failed jobs from the queue,
supporting the clearCompleted action in jobs.html.
The PostgreSQL jobs table is still used only for conform job creation
(POST /conform) to preserve that workflow.
