GitHub Pages · Three WASM Products · QualiaDB 0.0.41 Baseline & 0.0.41 Release Prep

Three packages.
Two budgets. One engine.

The portal is the full WASM-safe engine — logic plus science plus WebGPU viewport. The playground adds the browser LLM. Ontology MCP stays a bounded reasoning kernel and does not swallow science to keep Pages green.

Gates live in pages.yml and release-wasm.yml. Numbers below are measured from the files this page just fetched. Portal and playground fail the retired 2.6 MiB / 800 KiB cap — that is expected once science is in the blob.

v0.0.41 Release Prep (v0.0.41 Baseline)

Cross-Platform WebGPU Volumetric Renderer, Etalon-WASM Worker Steward, and 10-Epic Engine.

Read 0.0.41 Release Report

Live portal probe

What this site’s portal actually exports

initialising…

Loads pkg/qualia/qualia.js. Science names appear if the full engine is the artifact Pages deployed. LLM names belong on wasm-full, not here.

Engine version
—
Science-shaped exports
—
Portal constructor
—
waiting…

v0.0.41 Release Prep · v0.0.41 Baseline

Unified WASM Volumetric & Etalon-WASM Pipeline

wgpu 30.0.0-etalon · wasm32-unknown-unknown

Building upon the 0.0.41 production baseline, the 0.0.41 release unifies the graphics and compute runtime across native desktop and browser targets. webizen-render volumetric rendering modules compile directly to wasm32-unknown-unknown, while webizen-web (qualia-wasm) introduces two-tier rendering (Hardware WebGPU Volumetric + Deterministic Canvas2D fallback) paired with etalon-wasm worker telemetry.

webizen-render on wasm32-unknown-unknown

Direct WebGPU compute and volumetric raymarching now runs natively in browser contexts without desktop-only stubs:

  • VolumetricRenderer::from_canvas(canvas, cap): Configures a hardware WebGPU PortalGpu surface directly on an HtmlCanvasElement with capability-intersected adapter limits.
  • VolumetricRenderer::render_scene_frame(scene, camera): Zero-heap per-frame upload of 10D node tensors (Tensor10D) and colored scene meshes from a RenderScene.
  • volumetric_hmc: Hamiltonian Monte Carlo volumetric sampler compiled to WASM for continuous probability distributions.
  • volumetric_texture: Direct 3D density texture sampling and hardware filtering on browser WebGPU devices.

webizen-web (qualia-wasm) QualiaPortal

Dual-tier rendering architecture ensures universal compatibility from high-end WebGPU rigs down to restricted software canvases:

  • Tier 2 Hardware WebGPU: await portal.init_webgpu(canvas) initializes GPU volumetric pipelines, marks WorkerSteward ready, uploads cached tensor state, and renders the active RenderScene.
  • Tier 1 Deterministic Canvas2D: SimpleCanvasRenderer::render_scene(scene, yaw) provides mathematical Canvas2D fallback, projecting and drawing translucent faces, directed edges, and labeled nodes before overlaying the HUD.
  • etalon-wasm WorkerSteward Telemetry: Live inspection via portal.etalon_transport_mode(), portal.etalon_worker_state(), and portal.etalon_session_generation().
Worker Transport Mode
SharedArrayBuffer / Transferable
Negotiated dynamically via crossOriginIsolated
Session Generation
Zero-loss epoch tracking
Automatic device recovery generation bumps
Automated WASM CI
release-wasm.yml
Runs on dev branches + release tags for all 3 profiles
Browser WebGPU Volumetric SDK & Fallback Usage JavaScript / TypeScript
import init, { QualiaPortal } from './pkg/qualia/qualia_wasm.js';

await init();
const canvas = document.getElementById('qualia-viewport');
const portal = new QualiaPortal(canvas);

// 1. Inspect Etalon-WASM transport telemetry
console.log('Transport mode:', portal.etalon_transport_mode());      // "SharedArrayBuffer" or "TransferableMessage"
console.log('Worker state:', portal.etalon_worker_state());          // "Ready" | "Standby" | "Initializing"
console.log('Session gen:', portal.etalon_session_generation());     // Generation epoch for recovery

// 2. Load cross-platform RenderScene contract (or 10D tensor manifold)
portal.load_json_scene(JSON.stringify(renderSceneContract));

// 3. Attempt Tier 2 WebGPU upgrade (seamless fallback to Tier 1 Canvas2D on older hardware)
try {
  const tier = await portal.init_webgpu(canvas);
  console.log('Hardware WebGPU Volumetric active (Tier 2):', tier);
} catch (err) {
  console.info('Falling back to deterministic Tier 1 Canvas2D scene + HUD renderer:', err);
}

In the browser packages

  • JSON / CBOR ingest, N3 / SHACL, modal logic
  • WASM-safe CAS, DFT, ODE, bio, chem
  • WebGPU viewport + AcousticPlane (portal / full)
  • 0.0.41.8 AAA Graphics: Temporal AA (Rg16F velocity & R8 reactive masks)
  • Lighting & Materials: Dual IBL environment probes & split-sum specular
  • Textures: KTX2 block transcoding (BC1/BC3/BC7, ETC2, ASTC) + Zstd stream
  • Dynamic Worlds: Vegetation blade deflection & procedural shoreline foam
  • Browser LLM only in wasm-full

Native-only — never in these blobs

  • Graph daemon on :4242
  • Filesystem volumes, NVMe / ZNS / CSD
  • BLE mesh, eBPF / WFP / XPC filters
  • QDNF / Qualia Peer Runtime (native fabric)

Experience AAA Graphics & WebGPU Pipeline

Interactive capability breakdowns, zero-heap frame signatures, and real-time client WebGPU hardware probe.

Launch Graphics Showcase