Repository navigation
Description
Reconciled 2026-11-08 — historical process superseded. Historical separated V6 GUI owner lane/renderer is superseded. The only approved assembled /preview/ and exact native JUCE WebView GUI are now on main (#418, #439, #458). No separate V6 standalone GUI will be rebuilt. Any remaining render/native acceptance belongs in #389; actual routing in #412 and dynamic channels in #192.
Current main at audit:
c57720ea25585c861af89802e1365ffef7d03d17. No statement of physical hardware FAT.
Ownership (separate GUI development thread)
Branch: work/operator-console-v6 (branched from validated main bb6dff6b). Execution lock: draft #267 / GUI handoff. ENGINE dependency: #266, coordinated in epic #192. Existing #176 is the historical geometry baseline, not this Console redesign.
Mandatory outcome
Replace menu/modal-first GUI with one real premium LIGHTWEIGHT live mixer console inspired by Adobe Audition Mixer + Studio One/Fender Studio Pro Console + eMotion LV1 functional patterns. Persistent compact per-channel IN, M/ST, real VST3 rack and on/off, SEND PRE/POST, OUT, PAN/MUTE/SOLO, truthful native LED meter and fader; horizontal bank navigation, pinned MAIN. Advanced Patch only for bulk expertise. Use operator-selected physical/virtual I/O, never force OBS. Keep existing SVG fader; no blur/compositor-heavy glow, oversized cards or false status.
Operator journeys
Interface IN1 → CH07andInterface IN2 → CH02direct from strips; each invokes the same verified native source/ACK, no new JS router and no silent IN1 fallback.- Installed VST3 can be selected, enabled/bypassed, reordered, and GUI opened on every native channel; unpatched FX is prepared SILENT.
- Native-accepted +CHANNEL creates actual CH11, Group Selected makes real serial Group→MAIN, SEND→FX Return creates parallel wet branch, Sends-on-Faders selects monitor bus and restores main faders. Do not display these as LIVE if the engine capability is not yet present.
- For invalid pins/FX/clock/revision show REJECTED with previous live graph preserved. No STAGE→MUTE→APPLY→START for ordinary tasks.
File ownership
Only web/** (including UI-owned controllers web/app.js, web/bridge.js, web/monitor.js, web/mixer-io.js, web/mixer-layout.*, UI style source), UI JS/Playwright tests and deterministic standalone HTML generator parity. No edits in engine/core/**, native callback or src/Main.cpp; engine owner #266 controls those files.
Done / tests
One cohesive GUI PR from work/operator-console-v6: real native snapshot-driven ChannelId renderer (not strip-index hardcoded), one intent/ACK controller, source/FX/out/send UX tests at 1024/1366/1440/1680 widths and short/normal screens, mobile not a priority. Mock UI fixtures must be marked mock. No geometry jumping, disabled source after wrong pin, or false native LIVE. UI only considered integrated when same exact SHA with real #266 feature set passes Windows + Chromium and hardware FAT #15.
No CI fanout; no doc-only progress-as-product PR.
Activity
docs: consolidate Universal Live Console v6 execution and parallel wo…
masarray commentedon Nov 1, 2026
Consume the shared typed integration contract (no UI-local ChannelId graph)
Integration PR #269 now has actual C++ accepted readback / typed intent fence and golden PCM. See contract guide. All ChannelIds and request IDs serialize as decimal strings to avoid JavaScript 64-bit precision loss;
deviceEpoch + graphRevision + rosterRevisionmust match native ACK, and duplicate/stale requests must not change GUI's last accepted state. Native physical input ordinals must be supplied by verified backend; neverIN1by assumed strip position. Source/FX Return/Group capability and actual PCM/LIVE remain disabled until #266 production callback is connected. Keep per-strip direct IN/OUT/FX/SEND but do not create JS DSP/router or local accepted topology. If the working GUI uses incompatible field names, negotiate the versioned wire mapping in epic #192; integration header is not itself a wired WebView bridge.masarray commentedon Nov 1, 2026
INTEGRATION/FAT handshake from draft PR #270: a validator now rejects stale revision/epoch ACCEPTED, numerically encoded 64-bit ChannelId, fake Group/FX_RETURN/MONITOR native capabilities, ACK without a following FULL accepted snapshot, rejected graph mutation and surprise output ARM. Proposed event/trace shape documented in
docs/CONSOLE_V6_FAT_EVIDENCE_GATE_20261101.mdfor sign-off at #192; not a claim ENGINE emits it yet. GUI should keep pending commands local, disable unavailable roles and render only after native full snapshot. Please flag any wire-payload mismatch in #192 instead of implementing a second JS router.masarray commentedon Nov 1, 2026
GUI branch implementation is now in draft PR #271: #271 . The existing packaged mixer-layout module adds ChannelId-keyed direct console with native-only intent/ACK and capability-disabled feature gates; existing M4 UI stays operational until ENGINE #266 publishes real snapshots. Playwright native-stub test + 12 responsive geometries added to consolidated Chromium workflow; exact-head CI pending. Native wire vocabulary and roles posted for ENGINE review in issue #192 (comment 6552419056). No src/engine/core edits, no PCM or hardware claim.
masarray commentedon Nov 1, 2026
Cross-workstream merge collision found — Chromium workflow ownership
Latest active branch audit: GUI
work/operator-console-v6has 11 commits vs initial main and changes.github/workflows/ui-smoke.yml(+4 lines) to runtests/operator_console_v6_smoke.py. FAT #270 ALSO modifies.github/workflows/ui-smoke.ymlto run evidence-gate tests. This is a real shared-file integration conflict, not a reason to duplicate CI.Resolution: GUI #268 retains its Playwright test file and one local command; FAT #270 is the final owner of workflow YAML composition and must integrate the GUI test invocation when GUI PR code is incorporated. At merge, preserve both
tests/operator_console_v6_smoke.pyandtests/console_v6_fat_gate_test.mjsgates within the existing one Chromium workflow, no new job or workflow. Do NOT blindly overwrite either workflow YAML, and do not mark two branch-level green jobs as a green combined tree. PR #269 is C++ integration tests only (separate from #270's evidence validator), and Windows target dependency is additive. All last accepted native capabilities must still be truthful; physical #15 remains release blocker.Current
main@834a26ealso repairs historical reference unit test failures centrally. Update branch based on main intentionally before the final same-tree integration build; prefer one coordinated promotion point.masarray commentedon Nov 1, 2026
SCHEMA RECONCILIATION GATE — source-level mismatch spotted before merge
Audited latest GUI
work/operator-console-v6native listener and integration PR #269. These are complementary, but NOT wire-compatible yet:arMixConsoleSubscribe,arMixConsoleIntent,arMixConsoleAck, and full snapshot{device:{epoch,outputArmed,inputEndpoints[]},graph:{revision,channels[],destinations[],capabilities:{supportedOps[]}}}. Engine productionsrc/Main.cppcurrently emitsarMixRoutingLabAckfor 10CH and does not yet publish this v6 protocol.requestId='gui...:sequence',baseRevisionas JS safe integer, and channel IDs as strings (fixture examplesch-07andnative-64bit:...). INTEGRATION v6: native accepted ChannelId readback + intent ACK fence + two-input golden PCM #269 C++ control contract uses native uint64 request/graph/epoch tokens and emits decimal-string ChannelId. This needs one explicit JSON serializer/codec and one agreed wire representation; do not silently coerce 64-bit identifiers to JS Number or assumech-07is a stable native ChannelId.graph.capabilities.supportedOps[]; INTEGRATION v6: native accepted ChannelId readback + intent ACK fence + two-input golden PCM #269's integration readback deliberately leaves all runtime capability flags false, and physical SOURCE unresolved. Do not enable +CHANNEL/Group/FX Return from offline graph fixtures.Recommendation to owners: choose wire v6 with
requestIdopaque string (native echoes unchanged),ChannelIdcanonical decimal string for all IDs,deviceEpoch/graphRevision/rosterRevisioncanonical decimal strings or strictly safe-integer with explicit upper bound and no lossy cast,graph.capabilities.supportedOps[]native-authenticated, andACKrequiring snapshot replay with same epoch/revision. Keep host operation names centralized in enum/codec, not two JS router tables. ENGINE #266 should own actualsrc/Main.cppserialization/command handler; GUI #268 owns the projection; integration #269 retains typed native validation/golden PCM, #270 evidence gate only. Treat this as a merge blocker, not a cosmetic mismatch.masarray commentedon Nov 1, 2026
GUI v6 implementation gate PASS — draft PR #271 HEAD
6fd57b931acb518c09cc0f9e1dc0b168d5067969, exact-head Chromium/Playwright PASS run 41267502684. Native ChannelId-based compact per-strip IN/M-ST/FX/SEND/OUT/PAN/M/S/meter/fader, pinned MAIN, banks, 12 viewport geometry tests, source-owned standalone parity, safe intent/ACK rollback and stable fader updates are committed. New main integration contract (ConsoleNativeContractV6.h) reconciled: v6 string decimal IDs/revisions, baseGraphRevision/baseRosterRevision, acceptedGraph/rejected; offline readback cannot activate v6 nor claim audio. Remaining ENGINE #266: true JUCE publisher, callback and real feature admission; #192: transport operation/ACK mapping and physical pin metadata. Keeps M4 live UI, unsupported native operations disabled, no engine-core/src edit. PR deliberately stays DRAFT until engine handshake integration, hardware #15 independent.masarray commentedon Nov 1, 2026
User submitted screenshots of old Windows GUI: bloated device screen (software naming/matrix), stereo L/R both IN1, accepted route without RAW PCM. Integration has not rewritten GUI owner's code. Latest owner
work/operator-console-v6@704d5b4improves selected-channel source and collapses port matrix. First GUI run 41269561401 failed because standalone HTML was stale; adjunct #274 reconstructed exact standalone slots without changing source. Combined physical candidate #276 takes that exact GUI snapshot into verified Windows FAT lane. Its initial CI had stale MS3A assertion of older label; corrected at b5d505c, Chromium+Windows CI pending. Please reconcile these commits in GUI owner branch if appropriate; don't independently create DSP or call PCM audible from ACK. More critical: choose actual capture endpoint, show one channel only, disallow identical stereo L/R; current screenshots say VB-Cable CABLE Output + Realtek render, no RAW capture PCM.masarray commentedon Nov 1, 2026
First Sound hardware UX regression — dedicated PR #275 (physical screenshot-driven)
User's 2026-11-01 physical screenshots reveal a single enabled capture pin while CH01 was displayed STEREO L=R on IN1, despite LIVE/route ACK, plus NO RAW CAPTURE PCM with capture VB-Cable CABLE Output and playback Realtek. Hardware #15 is still a separate acoustic issue.
I created PR #275 on
fix/first-sound-audio-setup-one-path-20261101, based directly on main (does not touch GUI V6 owner branch or native engine):Integration: GUI V6 #271 should retain its owner-only schema and native disable gates. #275 targets the existing working M4 First Sound UI and can be integrated before/handoff to v6; avoid overwriting either branch's CSS/standalone generated assets. FAT #270 controls final combined workflow packaging. Keep actual native source validity and acoustic gate hardware-grounded.
masarray commentedon Nov 1, 2026
Screenshot-driven GUI M4 Audio Setup correction committed to GUI branch, tracked in draft PR #271, exact HEAD
2c40467be2f3d43d02ffec912a4c048ebb47cf4c; Chromium/Playwright PASS #41270151309. Unlike prior v6 mock-only shell, changes modify the ACTUAL legacyweb/mixer-io.js/cssshipped by WebView2: step-labelled DEVICE / INPUT→CHANNEL / MAIN OUT; basic driver pin toggles; optional folded device matrix; single selected CH input with native one-gesture ACK; directly editable MAIN L/R; one-pin STEREO blocked; invalid accepted STEREO L=R visibly flagged with a guarded 'FIX AS MONO' route request. Existing monitor.js authority, native channel/show callback, plugins and DSP unchanged. Added screenshot-backed one-pin VB-Cable Playwright test (artifactaudio-setup-one-pin-1366.png) and regenerated standalone parity. IMPORTANT: no new Windows EXE built from this draft; no raw capture PCM in user's field screenshot, so audio/hardware FAT #15 still not passed. Native stereo-pin preflight requested on epic #192. Keep PR DRAFT until schema/native integration and Windows packaging gate.masarray commentedon Nov 1, 2026
Actual Windows M4 GUI UNIFIED CONSOLE milestone committed to draft PR #271 (e2da305). https://github.com/masarray/arsonmix/actions/runs/41272171864 PASS. This is no longer only the inactive v6 mock: web/mixer-io.* and mixer-layout.* now default to one real native-backed CONSOLE with IN/M-ST/FX/4 AUX sends/MAIN OUT/PAN/MUTE/SOLO/meter/fader in one strip, measured always-visible MAIN, horizontally banked inputs, compact first-sound 3-stage Audio Setup, one-pin stereo rejection, guarded native FIX AS MONO, and vendor FX right-click menu no longer intercepts its own trigger. No src/** or engine/core/** edited. Screenshots artifact: https://github.com/masarray/arsonmix/actions/runs/41272171864/artifacts/13069613371. Exact-head Chromium passed including legacy M4 FX/ROUTE/SENDS and Orbit, plus future v6/ACK. NOT a Windows EXE or physical sound proof. PR remains draft pending engine/Windows FAT.
masarray commentedon Nov 1, 2026
GUI #271 — Functional progress after PRD/ARD v5.2 audit
GUI head:
e5b8da0fdf700de12e3c5d4edd4afb2e49f35181(branchwork/operator-console-v6; PR still Draft). Read current PRD §9 / ARD §K / v8 Puzzle contract and reconciled workflow against user-approved PR #278. Did not copy any simulated audio/plugin state from reference study.Delivered: strict v8 native accepted-state adapter (no synthetic flat ChannelIds), strict opaque requestId + epoch/revision/roster ACK matching, verified endpoint-only source choices and accepted native
primaryOutId, honest RAW/MAIN PCM and OUTPUT ARMED status, actual C++ ENGINE four-event S0 fixture SHA2564cbded2c7ada3be511c0e9a09b1949979638eb274c432f89b673a98ce55c7718replay; fixed native request pending baseRosterRevision. UI v5.2 additions: bipolar center-snap PAN with center-origin filled track, separate L/R 22-step fixed-color dBFS native meter, instant narrow bank toggle without remounting fader, pinned MAIN; no JS DSP or secondary router.Evidence: Chromium regression / generator byte parity / operator Playwright acceptance PASS run #41279959067, commit exact
e5b8da0f. Previous complete operator test and meter/PAN/bank test also passed.Unresolved: Native producer is currently S0
productionCallbackLive:false,supportedOps:[](correctly disabled). ENGINE #266 must deliver real callback ChannelId + plugin + PCM, FAT #270 must qualify the same exact source SHA Windows executable/physical audio. This is NOT an end-to-end mixer release, no actual hardware first-sound proven, no UAT recommendation.masarray commentedon Nov 2, 2026
Legacy GUI cleanup decision (2026-11-03): historical L2/workbench #176, old theme #44, and old Audio Setup wizard #273 are superseded; #268/#271 remain the active new GUI owner. Original #273 FIRST SOUND requirements (verified real enabled pins, no pseudo-stereo IN1/IN1, direct strip IN/OUT, native ACK+matching snapshot, actual raw→channel→MAIN PCM, device/clock/rollback safety and #15 physical test) remain uncompleted and are consolidated into epic #192. Do not reopen/ship an alternate old GUI.