The Metal tensor backend is host-Vec based: every op copied its inputs up
and its result back, which put the DINOv3 backbone at 2.1 s a frame on an
M3 Max. try_vit_backbone_resident_f32 runs a whole pre-norm ViT stack
(LayerNorm, q/k/v, rotate-half rope from tables, flash attention, out,
LayerNorm, SwiGLU, down; residuals in place) inside one command buffer
against cached bf16 weights, with a new kernel_makepad_rope_half_tables_f32
kernel for the rope. The precompiled metallib now carries the bf16 GEMM
kernels the runtime source compile already enabled on bfloat devices.
The common backend exposes it as gpu_vit_backbone_resident (CUDA declines;
its per-op path is already resident) and the body backbone tries it first:
233 ms a frame on the M3 Max, same oracle parity (kp3d 1.5 mm).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The FP8 backbone mode added each linear's bias by downloading, tiling and
re-uploading it per call, which cost more than the FP8 GEMM saved. A
gpu_add_cols_broadcast op (CUDA kernel; host loop on the Metal tensor
backend) adds a cols-wide bias to every row on the device, and each linear
keeps its bias resident after the first upload.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
BodyModel::set_backbone_fp8 quantises every backbone weight to E4M3 with
a per-tensor absmax scale and runs the tensor-core FP8 GEMM (bias
broadcast after). A backend without FP8 turns it off per layer on the
first refusal, so Metal keeps bf16. The oracle test reports its
accuracy and timing next to bf16; the default stays bf16 until the
numbers say otherwise.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The backbone's linears now take the cuBLASLt bias-epilogue / bf16 mm
paths (bf16 operands, f32 accumulation, bf16 output — the reference's
precision) with the f32-accumulating GEMM as the fallback. The rig's
pose correctives run on the final refinement step only by default: the
intermediate steps only feed keypoints back into the decoder, and the
oracle shows the difference is 0.4 mm (2.1 vs 1.7 mm against the
reference) for half the loop's rig time; `correctives_every_step` keeps
the exact mode.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The backbone, conditioning and decoder now take any square crop whose
side is a multiple of 16 (512 is the trained size); BodyModel::set_crop_size
selects it and caches the dense positional grid per size. Measured on the
oracle image against the reference: 512 gives 1.7 mm mean keypoint error,
384 and 256 about 2 cm mean (6 to 7 cm worst joint), 192 falls apart —
the knob is a real accuracy trade, not free speed. The decoder loop
reports its split (layer chain, heads, rig+camera, refinement) so the
next optimisation is chosen on numbers.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Apps ask the hub for a recognizer or a voice and get one; where it runs is
the hub's decision. AiHub::start_stt / start_tts return poll-driven
sessions shaped like the chat session. The Auto ladder is Whisper/Kokoro in
this process (weights present, machine election), on the machine node over
loopback, on a LAN node, else the OS engine; SpeechReach::Local is the
"don't reach out" knob. Audio always comes back as PCM: the app owns the
device.
Three layers:
- makepad-ai-speech is the whole speech model family, engines only.
libs/voice (Whisper + Silero VAD) folds in as the `whisper` and `vad`
modules next to kokoro and indextts, each a cargo feature; the Apple
bridges and the Speaker/VoiceTranscriber selection leave it.
- makepad-system-speech (new) is the OS speech services as blocking fns:
Apple SpeechAnalyzer/AVSpeechSynthesizer via Swift, Windows.Media.Speech*
on the vendored bindings, Android SpeechRecognizer/TextToSpeech through
MakepadSpeech.java (API 26 floor), espeak-ng on Linux. It models the two
STT shapes honestly: PCM in (Whisper, Apple) versus an engine that owns
the microphone (Android, Windows), with capabilities the caller reads.
- the hub grows speech sessions, in-process Whisper/Kokoro workers with the
residency election, a `whisper` wire backend (stt domain, registry entry
pinned to ggerganov/whisper.cpp) so a Mac can serve a Quest, and a
`language` field on the generate request.
Consumers: the Window voice input runs on an STT session and switches to
engine-mic mode when the recognizer owns the microphone; converse's
SpeechOutput is a lazily started TTS session plus a pump thread; route
drops its private speech copy for converse; vj's lyrics fallback and the
alignment bakes call the engines directly.
Verified here: speech-roundtrip through the real sessions (Apple voice in,
in-process Whisper on Metal out, 4.3% WER); system-speech-test TTS->STT
verbatim; hub/converse/system-speech unit tests; msvc, aarch64-android and
linux-gnu cross-checks; Java against android-34. Windows, Android and Linux
bridges are compile-checked only.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
libs/windows/windows-rs/src/Windows/mod.rs was a hand-pruned snapshot with
no recipe behind it; adding a WinRT namespace meant hand-copying vtables.
Now tools/windows_bindgen (standalone, windows-bindgen 0.62.1) regenerates
it from filter.txt: package mode with the COM `_Impl` traits, Win32 imports
linked through windows_core, upstream's 800-column rustfmt (the repo-root
rustfmt.toml disables formatting and has to be overridden), the filter
closed over every dependency the generator would otherwise skip a member
for, the namespace tree folded into one file with every feature gate kept,
and the two spots where the published 0.62.1 generator predates the vendored
core 0.62.2 normalized (Error::from_thread, imp::array_proxy).
The filter is the old file's item list plus Windows.Globalization.Language,
Windows.Media.SpeechRecognition and Windows.Media.SpeechSynthesis for the
OS speech engines. The generated `deprecated` gate is declared as a feature.
Checked on x86_64-pc-windows-msvc: platform, video, network, mpterm,
system-speech.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The backbone (20 heads) and the decoder (8 heads) both have 64-wide
heads, which the stack's FA2 kernel covers: f16 operands with f32
softmax and accumulation, the reference's own precision class, instead
of the composite path that materialises the 1029 x 1029 scores per head.
The composite path stays as the fallback where a backend lacks the
kernel. Oracle parity unchanged.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The ray-conditioning conv used to download the 1024 x 1280 embedding,
concatenate the 99 ray features on the host and upload the result; it is
now two resident linears over the two column blocks with the no-mask
term folded into the bias, plus an add (32 ms -> a few on the 4090).
The decoder loop downloaded the whole normalised token block every
layer for its one pose row; it now slices that row and fetches the
block once at the end (or per layer under a trace). Same numbers.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
On the 4090 the frame was 885 ms of which 840 were host scalar loops in
the decoder loop: the three 70-row refinement FFNs (about 240M
multiply-adds a step), the two heads, and the rig's 55k x 3000
corrective output layer plus its identity blendshape sum. Those are now
GPU-resident linears (heads.rs GpuStepHeads, MhrRig::prepare_gpu), the
skinning computes one transform per joint instead of one per influence,
and the results are identical (all 29 oracle tests unchanged). The rig
still runs entirely on the host when no GPU side was prepared.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The recut in a6341981d applied its patch against the wrong directory
and restored the six shared files to the previous tree without the
body-native hunks. This commit adds exactly those: the `body-native`
feature and optional dependency, the pinned `sam3dbody` registry entry
and its test, the backend arms and the module declaration. Working
tree untouched.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
66e5e2f11 committed the working tree of the shared hub files and with it
another lane's uncommitted edits (a new domain, request fields, a
backend arm, manifest lines). This commit restores those files to the
previous tree plus only the body-native hunks. The working tree is
untouched: the peer's edits stay on disk as their uncommitted work,
exactly as before.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
model.rs closes the loop: crop -> backbone -> ray-conditioned context ->
six decoder steps, each through the pose head, the rig, the camera and
the projection, then the packet the sandbox already reads (kp3d/kp2d in
camera axes, the 204 rig parameters, global rotation, camera translation,
joint positions). Against the reference on the oracle image, on Metal:
3D keypoints within 1.7 mm, 2D within 0.4 px, rig parameters, camera
and rotation within 2e-3. packet.rs writes the JSON by hand with the
reference worker's rounding and field order.
The hub gains the `body-native` feature (default on): registry entry
`sam3dbody` pinned to the Comfy-Org repack by revision, size and sha,
body_native_backend.rs beside the subprocess reference backend with the
same live_step contract, the `body` capability advertised when the
feature is compiled, and a stubbed test double for the CPU-only tests.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
decoder.rs builds the 145-row token set (pose, previous, prompt, two
hand-box rows, 70 keypoint rows, 70 3D-keypoint rows), runs the six
layers with the SAM-style repeated positional encoding (none on the
first layer's self-attention), cross-attends against the ray-conditioned
context, and after each step hands the normalised pose token to the
heads and updates the keypoint rows from the caller's feedback: the
2D-keypoint positional FFN and the bilinearly sampled context features
on the valid rows, the pelvis-centred 3D positional FFN on the rest.
heads.rs holds the host-side ReLU FFN heads, the refinement FFNs, the
hand-box MLP and the hand classifier. Against the oracle on Metal, every
layer's residual stream is within 6e-3, every step's pose head within
8e-4 and camera head within 2e-5; token assembly and the host heads
match to 1e-4.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
dino.rs is the 32-block ViT-H+/16 on the gpu_* surface after the TRELLIS
conditioner: bf16-resident linears with f32 accumulation, layer scale
folded into the output projections, rotate-half rope, SwiGLU, the
model's own final norm. preprocess.rs is the crop (box -> 1.25 pad ->
3:4 -> square -> 512 bilinear warp, ImageNet normalisation), the CLIFF
condition vector and the patch rays; condition.rs is the dense
positional encoding and the ray-conditioned decoder context. Against the
reference oracle on Metal: backbone 1.1% mean relative (bf16 noise),
ray-conditioned context 8e-5, crop within one u8 rounding step, rays and
dense PE 1e-7. One reference detail the paper does not state: its shrink
of the ray field is an antialiased filter whose taps clip at the image
edge, so the two edge patches sample inside their block centre (9.03 and
501.97 rather than 7.5 and 503.5); block centres left 0.1 of error.
The Metal tensor backend was missing rope_half (it aliased the
interleaved layout) and layer_norm_mul_add; both now exist with the
CUDA contract, which is what lets this backbone run on Apple silicon.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
mhr.rs is the Momentum Human Rig on the CPU: blendshapes, the 889x249
parameter transform, parents-first similarity kinematics (x,y,z,w
quaternions, Rz Ry Rx eulers, 2^s scales), the sparse-then-dense pose
corrective MLP over joints 2.., linear blend skinning and the 308-row
keypoint regression. pose.rs decodes the 519-wide head output: 6d
rotations, the 23 ball / 58 hinge / 6 translation body layout, the
mixed-dof hand layout, scale and hand component bases, the camera
translation and the perspective projection. Against the reference
oracle: vertices within 1e-4 cm with correctives on, keypoints within
1e-6 m, rig parameters within 1e-7. Two conventions the spec could not
settle on paper are now settled by the fixture: the head's global
rotation triple arrives Z,Y,X-ordered, and the corrective features start
at joint 2 (750 wide). Fixture tests skip cleanly without the oracle
directory or the weights.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
makepad-ai-body joins the AI model workspace: the constants of the
architecture (DINOv3 ViT-H+/16 at 512, the 6-layer promptable decoder,
the 519-wide pose head, the 127-joint / 18439-vertex MHR rig) and the
single-file safetensors reader for the Comfy-Org repack, which fails
closed on a Meta checkpoint-style header and checks the shapes the port
is written for at load. The backbone, decoder and rig modules follow
in their own lanes against the spec under local/agent_state/sam3dbody.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The doctrine catches up with the engine: landforms and tunnels are one
call and digging works anywhere; railways and roads are edited by their
path points (crossings generate, styles switch, any model drives);
freshly modeled aliases are live immediately — never park a display
substitute.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The voxel layer's volume gate is gone — any edit materializes chunks
anywhere on the map (volumes remain as styling), with the lazy-chunk
economics proven by test: zero storage until the first op. The composed
surface seam (GameWorld::surface_height_at: heightfield where untouched,
voxel surface where a chunk owns it) is what ground queries, senses and
the drape corridor read, so a mountain raised mid-game and a dug crater
are the same truth the vehicles drive on. landform.rs adds the macro
shapes (mountain/hill/ridge/valley/crater/plateau, fbm detail, slope
rock + snow recolor) and the tunnel op rides the capsule-filtered
materialization.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
searchdb imported std::os::unix::fs::FileExt for read_exact_at, which
does not exist on Windows (reported from a Windows checkout of work).
A small ReadExactAt trait now wraps unix pread and Windows seek_read,
both cursor-free, so the shared reader keeps serving lookups from
several threads without a lock. Checked on x86_64-pc-windows-msvc.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A realtime feed session parked in its wait-for-a-frame loop forever when
the client died without sending stop (a sandbox quit left job-2 live on
.123 for five minutes holding the GPU slot). The wait loop now returns
to the top of the session loop once no socket is left, where the idle
timeout counts a socketless session down. Test covers it.
SkinnedModel gains node_parent/node_count, joint_skinned_centroid (the
direction a leaf limb actually runs, from the flesh it skins) and
from_nodes (a mesh-less rig for hierarchy-maths tests) — what the
sandbox's webcam mocap retarget needs.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B626urtY1Xo4hQdLzvSK6F
EmitterAnchor::EntityLocal carries an offset that yaws with the entity —
a locomotive's chimney keeps its smoke over the funnel through every
corner. The step lookup resolves (position, yaw) instead of position.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Same orphaning as the sandbox: context::assemble had no caller post-P8,
so gen/vj sessions ran brief-less. The executor's capability doc leads
with the assembled profile layer.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The verbs existed with the right doc strings, but the model reads the
brief, not the verb table — and the brief's ONE CALL list never said
railway, so it hand-placed track pieces, horrendously. Now it says it,
with the two-line essential shape.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Source Library wall's pixel engine, extracted for anyone's pictures:
libs/image_tiles carries the NV12 slot pyramid, the 32x32 shard atlas
geometry and the one-HEVC-intra-frame-per-file tape codec (tape.rs), a
small SQLite index (db.rs), a priority decode pool (store.rs), and the
TileGrid widget — instanced tiles batched per atlas page, anchor-locked
wheel zoom with log-space glide, continuous per-shard LOD with
crossfades, full-resolution promotion under byte budgets with LRU
eviction, and uniform-only re-presents inside the pad of the last build.
image-tiles-bake downloads a manifest of URLs (fetch pool free to be
wide, encode pool hard-capped — concurrent VT session churn has panicked
the encoder's IOMMU) and bakes a library; examples/image_tiles views one.
No JPEGs on disk; decode and encode are the hardware's.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Post-P8 the game session still advertised assets.query/assets.schema but
the in-process executor answered both Unavailable — the model's main
alias-search tool was dead, and every "find me X" turn gave up on the
store. Both now execute over the server's bounded /v1/assets/query route
(schema via sqlite_master plus the usage notes). The game brief gains the
new-game-vs-edit rule, interior design law, generate-after-search, and
explicit y semantics so player_pos().y stops being pasted into
ground-relative fields.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
One DrawSceneScreen serves both the in-world video screen and every
sprite billboard, and the sprite loop left its last pose, size and atlas
behind — so the next frame's "is there a screen" gate (zero screen_size
draws nothing) read the leftover and drew the whole spritemap as one
opaque quad: under the last-drawn unit in C&C, and hanging in the sky of
any level that draws no sprites at all. Snapshot the host's pos/size/
texture before the lane and restore after.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New `body` backend (sam3dbody-ref): a persistent length-prefixed-PNG /
JSON-lines worker subprocess seam with ready handshake, per-frame timeout,
bounded restarts. LiveFrameOut grows aux_json — structured per-frame JSON
sent to the client before the frame — and output_encoding "none" makes a
session pose-only (refused with loop_mode feedback, also on control flips,
which upgraded apply_control to Result). Worker code+model stay
box-provisioned via MAKEPAD_SAM3DBODY_WORKER; the repo carries only the
MIT seam. Codex lane + Fable review (ready handshake, spawn timeout).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0165w1ZL1f1TruX5u2qC7mSX
1.98's sharper macro-use tracking flags the live-id macro globs five script
modules kept without using; windows-strings' strlen extern takes *const
c_char with a cast at the two call sites instead of tripping
suspicious_runtime_symbol_definitions (PCSTR is transparent over *const u8,
so the ABI never changed).
the vision and ocr backends' ContentBackend machinery lives behind the
llm feature; the imports and the lane-refill helpers now do too, so a
client crate compiling the hub with a leaner feature set (vj, chat-ui)
builds without warnings. pipeline_submit_collect stays feature-free for
its tests and says so.
Squashed from work:
- docs: button shader annotations — widgets/button.rs complete + splash demo buttons
- tweaker: the Shader tab shows the pinned widget's ANIMATOR STATES as little posed swatches under the well — one per tr
- widgets: the glass stops reading draw_pass.time, so glass apps idle again
- render, sim, platform: what a strategy round needs underneath
- draw, platform: overlays now composite above content that uses draw_depth
- asset: mp4 sample index for range-streaming, chat tools, import profiles
- platform: native file and save dialogs, in-house on all three desktops
- sim: a per-unit decision hook — one commandable unit can think for itself, before the kit steers
- makepad_ai is deleted — every backend is a hub pipe, the agent seam lives with its consumers (aicore §14, decided 2026
- libs: the zero-warning sweep — stitch casts say what they mean, xatlas keeps upstream's surface quietly
- zero-warning sweep, round two — the first full-workspace pass
- zero-warning sweep, round three — the model lanes and the deep examples
- zero-warning sweep, round four — the last stragglers
Squashed from work:
- pdf_parse: indirect /Length, and recover the glyph names a symbol font uses
- mpwm: make the Linux desktop integration usable (#1204)
- mpfiles: the size map goes all the way down, and remembers what it found
- mpfiles: the treemap walk streams every directory at every depth — no more opaque growing blocks over deep heavy subtr
- mpfiles: local qwen chat sidepanel over the treemap selection
- mpfiles: a camera over the treemap — scroll zooms at the cursor with real re-layout detail, drag pans, double-click fi
- mpfiles: a 2.5D treemap — the map extrudes by nesting depth under a 3d toolbar mode
- mpfiles: a real orbit camera over the block view — left-drag orbits, right-drag pans, click stays a click; the project
- mpfiles: the wheel glides, the filter stops paying per frame and per keystroke — measured weights cached against the t
- mpfiles: the 3d view wears no name tags — labels only on hover, the tooltip carries the numbers
- mpfiles: a camera gesture never re-flows the map — mid-motion the picture rides one rigid remap of the last layout, an
- mpfiles: a small zoom keeps the map still — the layout is only remade when the camera truly leaves it (a 1.3x detail b
- mpfiles: a camera reveal is not an arrival — cells the cull swings into view appear instantly at full strength, cells
- mpfiles: the treemap packs in canon space — the arrangement is zoom-invariant by construction; zoom only refines the t
- mpfiles: the map stops breathing — insets are fractions of map space, names reserve no room and float on scrims with a
- mpfiles: a camera settle is calm — zoom detail materialises in place, no arrival fade; the ceremony stays with filter
- mpfiles: camera motion animates nothing but the camera — flat labels freeze per relayout and ride the remap; a calm se
- mpfiles: a fast zoom-out never outruns the map — cull escape relayouts the same frame, the glide's destination is laid
- mpfiles: wheel zoom in the raised modes anchors on the tile top under the cursor, not the ground hiding behind it
- mpfiles: first consumer of the ai-hub — the ask panel rides AiHub::start_local_chat, its private engine is gone (aicor
- ai-hub: chats run the machine election — route to a serving holder, wait on a loading one, claim and publish when open
- zero-warning sweep, round two — the first full-workspace pass
- zero-warning sweep, round three — the model lanes and the deep examples
Co-authored-by: Alex <blackanger.z@gmail.com>
Squashed from work:
- vj decks: sync is a held lock — pinned master, per-pump rate servo
- vj: the deck explorer opens on music
- frametween: the VJ's in-betweener becomes a library — the whole mode set, one definition
- frametween: the gate says which tier moved the picture, from pixels
- frametween: BGRA words go in as they are, and a host may own the clock
- vj: archive.org as a content source
- platform: native file and save dialogs, in-house on all three desktops
- vj: rounder slider caps, and the cap body reads as a blob
- ai-hub: makepad-asset-ai becomes makepad-ai-hub at libs/ai/hub, the chat pane becomes makepad-chat-ui, the service bin
- vj: DREAM runs execute in the app — pipelines.rs becomes the run it used to watch (aicore §9 / F1)
- vj: plain generations execute in the app too — the store's job queue loses its last vj client (aicore §9 / F2)
- asset-creator: the runner — generate one thing and put it in the catalog, one implementation for every surface (aicore
- importer + asset-server host: the coordination era ends (aicore P7)
- client + chat dispatcher: the dead wire comes out (aicore P7/P8)
- zero-warning sweep, round five — vj and chat-ui
- zero-warning sweep, round six — three cascades
- zero-warning sweep, round seven — the last two
Squashed from work:
- rtsmap: one seeded generator for tiled strategy maps
- map: bake a runnable Amsterdam test map from inside the route app
- route: the first-run test map starts itself, and its buttons are guarded
- map: the test map bakes its road faces too, from the same shared pass
- map_tiles: mkmap-extract — the weave is reversible, world-cells reconstructible from world.mkmap
- route: the local dispatcher rides the hub — its copied engine is deleted, limits preserved (aicore P1)
- route + converse: off makepad_ai — the Agent seam moves to converse, route's cloud dispatcher rides the hub's Claude p
- ai-hub: chats run the machine election — route to a serving holder, wait on a loading one, claim and publish when open
- libs: the zero-warning sweep — stitch casts say what they mean, xatlas keeps upstream's surface quietly
Squashed from work:
- score: a headless music engraving, playback and notation engine
- score: the notation app — pianist mode, editing, playback
- piano_model: it was a plucked string by construction, and 20 voicings
- score: one document you can pan, zoom and navigate
- piano_model: the body tap was a click, and the objective was rewarding noise
- score: add the sound panel and library modules
- piano_model: a second engine, and the attack that finally sounded right
- score: two instruments, reverb and brightness — and the rest of the panel gone
- score model: a note remembers how it was struck, and the score remembers the pedal
- score import: keep the velocities and the pedal the file was carrying
- score playback: play the performance, not a flattened copy of it
- score ui: the music list moves to the sidebar, and the view stops fighting itself
- score: the application ships its font and eight performances
- piano_model: a limiter that rides the music, so the knee stops shaping chords
- piano_model: the forte bell was the treble's dynamic slope, and the bass was dying at its own prompt rate
- piano_model: the bridge decides each partial's decay, and a fixed multiplier cannot say that
- piano_model: each partial gets its own two coupled modes, from the eigen algebra
- piano_model: a median that fell between the peaks made every bass partial a drain
- score-ai: LocalBroker — the seam's in-process implementation over the session engine (aicore P8)
- client + chat dispatcher: the dead wire comes out (aicore P7/P8)
- score_pdf: the score model grew a pedal map — the pdf importer initialises it
- libs: the zero-warning sweep — stitch casts say what they mean, xatlas keeps upstream's surface quietly
- score app: the shipped-piece test speaks the PERFORMANCES table
- zero-warning sweep, round three — the model lanes and the deep examples
Squashed from work:
- asset-ai: FastH3 4-step fast video backend; clip keyframes on the wire
- asset-ui: loop video chains — text→image→video that ends where it began
- h3: safetensors -> pruned-Q4_K GGUF quantizer for the 24GB DiT tiers
- h3_quant_gguf verify: row-error gates calibrated to the measured Q4_K floor
- asset-ai realtime: the feedback loop — the source anchors, the drifted frame inits
- asset-ai realtime: a feedback loop survives a resize and travels by default
- asset-ai realtime: the feedback loop frees itself from the feed handshake and pauses for its listener
- asset-ai realtime: the outbound encode leaves the loop's critical path
- asset-ai ocr: the ocr domain — Chandra 2 at page resolution, and the tower goes planner-owned
- llm slots: a lane can hold an image span — embedding prefill and a rope cursor of its own
- vision tower on CUDA: the encode leg gets its two missing kernels
- llm/ocr: one M-RoPE grid encoder for both image paths, and a livelock made an error
- vision tower on CUDA: the f16 GEMM keeps the precision it was throwing away
- live: a feed that moves box takes its trip with it — one seed image
- vision tower on CUDA: the tiled attention becomes bit-exact, and tensor cores go
- llm prefill on CUDA: the MMA attention kernel gets the tile a 4-to-1 model needs
- asset-ai ocr: the CUDA encode lane joins the integration — vision-parity sits beside run's three arms, and the kernels
- Merge branch 'ocr-perf-integration' into work
- asset-ai: the live anchor can follow the trip, and text leaves the 5090
- asset-ai: the camera moves the world, and the world starts still
- asset-import: the EA strategy classics, in the one 2D contract
- rtsmap: one seeded generator for tiled strategy maps
- asset-ui: one card for the strategy classics, with a pack dropdown
- asset-ai: music3 reference-audio path, ocr/h3 backends, registry
- asset: mp4 sample index for range-streaming, chat tools, import profiles
- cnc: tiberium is twelve growth frames, not twelve empty variants
- platform: native file and save dialogs, in-house on all three desktops
- chat: the scan holds out for a lane home
- chat: a full home queues you — take the free lane
- chat: the preload has a percentage, and the boundless cap stops showing
- llm cuda: the 32x2 attention tile — even GQA ratios stay on MMA
- sa3 gets a bake path: the sfx model's tables precomputed by a diffusion-side bin
- sqlite_query: anti-join regression test
- td import: HARV's second frame block is its harvesting cycle, not a turret
- asset-ui: sprite enhancement runs on the 32B dev DiT — distillation, not the prompt, was the ceiling
- ai-hub: makepad-asset-ai becomes makepad-ai-hub at libs/ai/hub, the chat pane becomes makepad-chat-ui, the service bin
- asset-ui: test health fixtures grow the realtime field they were born without
- ai-hub: one home at ~/.makepad — weights/ run/ cache/ logs/, the service cache migrates from ai_content by a single re
- ai-hub: subprocess workers die with the node — process groups everywhere, PDEATHSIG on linux, one KILL_ON_JOB_CLOSE Jo
- ai-hub: the hub object — AiHub::in_process, pipes vocabulary, and the local LLM engine generalized out of mpfiles (aic
- strict-json: the dependency-free JSON module gets its own crate; asset-client re-exports it so nothing downstream move
- ai-hub: the machine layer — node entries, the 0600 machine token, and the residency election that IS the lock (aicore
- ai-hub: MPHUB1 — the fabric beacon only dedicated nodes can send (aicore §4)
- ai-hub: job leases — work lives only while it is renewed (aicore §8)
- asset-creator: the pipeline library is born — specs, the deps gate, and the derived-state law (aicore §9)
- ai-hub: RAM residency facts — the CPU-side twin of residency.rs (aicore §3)
- ai-hub: ETA placement primitives — relative GPU throughput, the four-term estimate, and an observable breakdown (aicor
- ai-hub: leases go live on the wire — origin fields on submit, /job/<id>/keepalive, /bye, and the reaper that cancels w
- ai-hub: the chat providers move in — fleet qwen, openai, grok, claude/codex/grok CLIs, the responses driver, and the w
- asset-creator: the engine — one pipeline run against the hub, deps-gated, spliced, cancellable, resumable-by-construct
- ai-hub: the machine node mode — --machine binds loopback, registers in ~/.makepad/run, and exits on its own once idle
- asset-creator: makepad-creator-run — the detached client for runs that must outlive a window (aicore §9)
- ai-hub: a native Claude Messages-API provider — API-key or Claude Code OAuth, bounded SSE streaming, injected tools (a
- route + converse: off makepad_ai — the Agent seam moves to converse, route's cloud dispatcher rides the hub's Claude p
- asset-creator: the preset tables move in — fifteen chain-policy constants shared by every creator app (aicore §9 / P6)
- makepad_ai is deleted — every backend is a hub pipe, the agent seam lives with its consumers (aicore §14, decided 2026
- ai-hub: loads hold the machine residency election — set_model_state claims on Loaded and publishes the service port (a
- ai-hub: chats run the machine election — route to a serving holder, wait on a loading one, claim and publish when open
- ai-hub: pick_for_domain_eta — ETA-ranked placement over the shared hard-filter core (aicore §6 / P4)
- asset-creator: the engine picks a provider per stage at dispatch time — a chain's later stages see fresh fleet state (
- ai-hub: the fabric secret gates the service HTTP surface — bearer on everything but /health and the ticketed peer path
- vj: DREAM runs execute in the app — pipelines.rs becomes the run it used to watch (aicore §9 / F1)
- asset-creator: the runner — generate one thing and put it in the catalog, one implementation for every surface (aicore
- chat-ui: the session runs in the app — no broker anywhere on the chat path (aicore P8 / F5)
- asset-store: assets.query is a first-class query endpoint — the bounded SQL surface outlives the broker (aicore P8 / F
- asset-creator: CreatorTools — the chat tool pack for a store that only stores (aicore §9 / P8)
- asset-store: the shrink — the store stores (aicore P7)
- importer + asset-server host: the coordination era ends (aicore P7)
- store config purge + asset-ui goes fleet-direct; the derive protocol gets its route proof (aicore P7)
- client + chat dispatcher: the dead wire comes out (aicore P7/P8)
- ai-hub: 0.3.0 — the health version says which era a node runs
- ai-hub: the default fleet is 'gen' — apps hear the LAN without env plumbing
- ai-hub: the preload note percents the prefill, not the job bar
- ai-hub: conversations keep their KV — the wire mirror, the lane identity, the in-turn dynamic context (aicore §7)
- ai-hub: an open-think model is thinking from its first token
- libs: the zero-warning sweep — stitch casts say what they mean, xatlas keeps upstream's surface quietly
- zero-warning sweep, round two — the first full-workspace pass
- zero-warning sweep, round three — the model lanes and the deep examples
- zero-warning sweep, round four — the last stragglers
- zero-warning sweep, round five — vj and chat-ui
- zero-warning sweep, round six — three cascades
Co-authored-by: Claude <info@makepad.nl>
Squashed from work:
- video: hardware first-frame decode straight from RAM — no temp files
- video: the encoder transform is reported once, not once per encoder
- video: a single still does not need a whole AVAssetWriter
Squashed from work; the fine-grained history is under tag archive/work-2026-08-29:
- mp* wave: mpwm window manager + the mp app family, WM API, theme bridge, PDF engine fix
- mpwm polish wave: terminal key focus, focus-history close order, pop-back-to-origin, occupied-workspace cycling, demo
- work: land the sources the last commits reference
- route: the assistant icon the committed UI references
- fast_inflate: the benches name their dev-deps
- gif/weezl: drop the vendored benches nobody can run
- weezl: the decode tests generate their own LZW fixture
Squashed from work; the fine-grained history is under tag archive/work-2026-08-29:
- mp* wave: mpwm window manager + the mp app family, WM API, theme bridge, PDF engine fix
- mpwm polish wave: terminal key focus, focus-history close order, pop-back-to-origin, occupied-workspace cycling, demo
- mpwm: ghost-launch/menu-flash root cause, scroll z-gate, browser binds, path-free status, viewers delisted, Fab opens
- mpwm: Hyprland-exact close animation over a frozen snapshot
- mpwm: warm-instance pool, flat-luminance opens, flicker-free CEF resize
- work: land the sources the last commits reference
- mpwm: quick-look gap fixes; image cache eviction on preview unload
- tweaker: material thumbnails + vibecode popup + ctrl-space notes, undo/redo over the edit ledger, capture-semantics pi
- tweaker: tabbed side panel (Props/Shader/Tree) - shader tab with checkerboard material well + prompt, complete widget-
- mpwm+mpterm: text quick-look — pty teardown deadlock fix, in-place retarget verified, debug probes stripped
Squashed from work; the fine-grained history is under tag archive/work-2026-08-29:
- mp* wave: mpwm window manager + the mp app family, WM API, theme bridge, PDF engine fix
- work: land the sources the last commits reference
- kenney: catalogue all 50 free 3D kits; Modal dismissed() never fired
- asset store: central vision-annotation queue; Kenney donate prompt
- asset-ai: vision domain — image + prompt -> text on every fleet node
- annotation runs the fleet's vision services through the normal job queue
- vj: responsive DJ mixer + Windows drag-and-drop, cherry-picked from PR #1199 (vjroger)
- store search: WordNet synonym expansion, query-side; per-term seeks
- video scheduling: a cold model pin never downloads past a warm one, and a dedicated box only serves its role
- an expansion can never lose a run — and `expand: true` finally means something
- fleet panel: a slow box is not a missing box
- h3: first+last keyframe conditioning (the weights were always FL2VA)
- registry: the three H3 FL2VA tiers name their real conditioning
- fleet scheduling: spread before stacking, and a stuck job moves
- a download never steals a job from a box that has the weights
- the faster GPU takes the tie: 6000 > 5090 > 4090
- an evicted flux model gives the card its VRAM back
- a job now records what each of its stages was handed
- open a stage in RUNS and read what it sent
- every expanded song was an instrumental: the lyrics had nowhere to go
- the writer never overwrites words the person wrote
- store: a dependent job's body is spliced from its deps' results at claim
- a fleet box's job row says what that job was asked for
- the store runs a whole pipeline, and one record says how far it got
- a client can declare a whole run, watch it, and stop it
- 100% now means published, and the expander is a job you can queue
- one card says what a spawned task is doing, everywhere
- a run that ends says so, instead of being noticed later
- when every box holding the model is busy, buy another copy
Squashed from work; the fine-grained history is under tag archive/work-2026-08-29:
- mp* wave: mpwm window manager + the mp app family, WM API, theme bridge, PDF engine fix
- mpwm polish wave: terminal key focus, focus-history close order, pop-back-to-origin, occupied-workspace cycling, demo
- mpwm: warm-instance pool, flat-luminance opens, flicker-free CEF resize
- work: land the sources the last commits reference
- kenney: catalogue all 50 free 3D kits; Modal dismissed() never fired
- platform: windows check green again — SetWindowTextW binding
- map: exact warp-aware inverse projections — pointer ops work folded
- mpwm: quick-look gap fixes; image cache eviction on preview unload
- tweaker: material thumbnails + vibecode popup + ctrl-space notes, undo/redo over the edit ledger, capture-semantics pi
- tweaker: vibe popup card chrome + dispatch order, ctrl-space notes verified, sploded design v2 chapter
- tweaker: tabbed side panel (Props/Shader/Tree) - shader tab with checkerboard material well + prompt, complete widget-
- sploded v2: nesting-depth z, hairline scope frames, body pass
- sploded: pin the depth convention with a test, kill the draw_depth residue
- sploded: real body-pass split (scene-pass capture, panel flat) + y-convention source of truth with anti-flip gate test
- sploded: hollow outlines, flat-band input, ray-pick unprojection
- tweaker: shader tab defaults to the selection's first draw layer, stale hint trimmed
- sploded: outlines become clipped, antialiased strips; tighter deck
- sploded: merge the lane's v2 (nesting-depth z, clipped AA strip outlines, flat-band input, unproject, SplodedStack bod
- sploded: the exploded view is a LIVE view — pointer events route through the inverse explode transform (ray -> plane -
- tweaker: tabs are real widgets (uid, tree node under the dock, own plane in 3D) and pickable; navigation-class clicks
- sploded: pinned/hover outlines render on the widget's own plane in 3D — per-widget nesting depth lives on the platform
- tweaker: the material well renders the pinned widget's actual shader — the swatch byte-copies the widget's live draw c
- tweaker: the Shader tab shows the shader as written — the layer's pixel/vertex fn source (nearest definition up the co
- sploded: I = true isometric preset (yaw 45°, pitch atan(1/√2))
- tweaker: eyedropper — the colour popover's pick button arms a pixel probe; the next press in the app samples that devi
- tweaker: the shader loop closes — /tweak/apply resolves the pinned widget by uid (anonymous path segments never round-
- vj: responsive DJ mixer + Windows drag-and-drop, cherry-picked from PR #1199 (vjroger)
- tweaker: the material well is a magnifier — the mirrored instance draws at the widget's native size in the well's own
- tweaker: per-layer material thumbnails — the Widget derive emits WidgetNode::layer_areas() (every #[live] Draw… field
- tweaker: the shader source view is the real CodeView (syntax highlighting, selection, editing) when the app registers
- tweaker: Ctrl+Enter sends on every platform (TextInput treated only Cmd as primary on macOS, so Ctrl+Enter inserted a
- Modal claims no layout slot: the DJ page fills its window again
- tweaker: every fn apply recompiles (eval_chunk ran every chunk under ONE synthetic callsite, so the script body — and
- tweaker: an apply whose draw shader fails to compile is rejected — the layer goes back (last live fns / the fn as writ
- tweaker: the Shader tab's source view owns its scrolling (the ScrollYView around the CodeView double-scrolled the care
- widgets: set_visible belongs to every widget, not just View (#1194)
- script: a dead heap's resource handles must not outlive it (#1195)
- Resources: search the executable's directory, not only the working directory (#1196)
- Windows: fit a restored window to the displays that are actually attached (#1197)
- d3d11: a failing GPU call reports the loss instead of killing the process (#1198)
Co-authored-by: Kevin Boos <1139460+kevinaboos@users.noreply.github.com>
Squashed from work; the fine-grained history is under tag archive/work-2026-08-26:
- vj: the thumbnail pipeline becomes one honest machine, and effects go livecodable
- widgets: boxed labels center on their ink, not on the font's line box
- metal: a fresh texture forgets nothing it never had — reallocated vec textures upload whole
- fab: a 3D creation shell and the viewer built on it
- texcomp: the block codec and the container every texture will travel in
- mixer: a live LR-Mix surface for the XR18 — strips paired the way the desk is run, auto-connect, EQ that bends its own curve, and a sweep paced so the console drops nothing
Squashed from work; the fine-grained history is under tag archive/work-2026-08-26:
- vj: thumbnails become mp4 — hardware-coded sheets at measured-4K cells, and the bake stops racing the GPU
- vj: the console grows real transports, and the deck stops lying about reverse
- vj: reverse earns a memory, and the effects stop aging
- vj: video goes NV12 end to end, and the GPU does the unpacking
- vj: the GPU learns to see motion — realtime frame tweening on every deck
- vj: the tweener learns — RIFE runs on the Mac and feeds the warp
- vj: the classical tweener grows up, and every deck gets a tween chip
- vj: the tween clock tells presented time, not producer time
- vj: the transport becomes a platter — velocity in, position out, one map
- vj: the tween presenter reads the platter — one clock per deck, cued once at the frame on screen
- vj: the OFF tier joins the platter — a resident clip's picture is cache[nearest(pos)]
- vj: the media thread loses its second clock — resident clips park the decoder
- vj: the producer gets a contract — keyed ladders, deadlines, and a capacity law
- vj: two decks, one law — identical inputs are bit-identical, and the warp agrees to the byte
- vj: the presenter switch lands without its scaffolding
- vj: the next pair's fields are fetched ahead of the change under the capacity law — a pair change costs an ordinary beat; macos: the layer's own display link paces the frame when the system offers it, the old path stays as fallback
- vj: AI3 subdivides adaptively — one, three or seven neural frames per pair, chosen from measured synth time against the pair's own period, with classical flow between them and a 7-3-1-FL fallback; the deck shows the depth
- video_flow: the flow debug bins, declared behind the convert feature so --no-default-features skips them instead of failing
Squashed from work; the fine-grained history is under tag archive/work-2026-08-26:
- fab: a 3D creation shell and the viewer built on it
- raytrace: the traced pane starts coarse and doubles to native, with the raster underneath
- fab: a colour picker, a material's textures, and dials that move the scene while they drag
- texcomp: the block codec and the container every texture will travel in
- fab: FAB_PROBE_MAT — per-material triangle counts, texture presence and uv spread in the roof probe
