Inference and training on raw WebGPU. Pure JS + WGSL.

Try the live demo | npm | docs

import { doppler } from '@simulatte/doppler';

// Stream tokens
const model = await doppler.load('gemma3-270m');
for await (const token of model.generate('Describe WebGPU briefly')) {
  process.stdout.write(token);
}

// One-shot
const text = await model.generateText('Explain WebGPU in one sentence');

// LoRA hot-swap
await model.loadLoRA('https://oneshift-twoshift-redshift-blueshift.com/manifest.json');

Registry IDs resolve to hosted RDRR artifacts from Clocksmith/rdrr by default. Tokens stream from a native AsyncGenerator. See the canonical Root API guide.

JS → WGSL → WebGPU. Direct JavaScript orchestration into native WebGPU kernels, avoiding ONNX runtimes, WASM blobs, and bridge layers.

for await streaming. Generation uses a native AsyncGenerator that fits normal app control flow.

LoRA hot-swap. Swap adapters at runtime without reloading the base model.

Independent model instances. Run multiple models concurrently. Each owns its pipeline, buffers, and KV cache.

All models below are verified with deterministic greedy decoding on WebGPU hardware. Registry IDs resolve to hosted RDRR artifacts automatically.

Additional model families (Llama 3, DeepSeek, Gemma 4 MoE, Mixtral, and others) have conversion configs ready but are not yet cataloged. See the full model support matrix for details.

• Sharded weight loading via OPFS moves multi-GB weights into VRAM without blocking the main thread.
• Quantized inference (Q4K, F16) runs practical model sizes on consumer GPUs.
• Kernel hot-swap between prefill and decode paths with zero graph recompilation.
• Config-driven runtime with explicit profiles, kernel-path selection, and sampling.

• Docs index (canonical navigation): docs/INDEX.md
• First-run workflow: docs/getting-started.md
• Runtime config contract: docs/config.md
• Architecture: docs/architecture.md
• Model support matrix: docs/model-support-matrix.md

• WebGPU is required.
• Supported runtimes: WebGPU-capable browsers, or Node with a WebGPU provider.
• Chrome / Edge 113+ supported.
• Firefox support varies (typically behind a flag).
• Safari support is evolving.

Apache License 2.0 (Apache-2.0). See LICENSE and NOTICE.