aegis-wgsl-webgpu-shaders

Aegis WGSL WebGPU Shader Instruction Dataset

Dataset Summary

The Aegis WGSL WebGPU Shader Instruction Dataset is a high-precision code-generation and conversational dataset mapping natural language physical simulation queries to executable WebGPU Shading Language (WGSL) compute and fragment shaders.

Modern browsers and edge devices utilize WebGPU for 60-120 FPS hardware-accelerated scientific visualization. This dataset trains or fine-tunes LLMs to act as specialized graphics engineers capable of translating hydrodynamic, thermal, seismic, acoustic, and aeration queries directly into functional WGSL shader code.

Dataset Structure

• Total Pairs: 50 curated conversational pairs (5 core physical categories × 10 prompt/response variations)
• Format: JSON Lines (.jsonl)
• Standard: ChatML

Physical Simulation Categories

1. Sentinel HUFP Fluid Dynamics: Eulerian fluid flow vectors, wave interference patterns, Navier-Stokes approximations, and coastal storm surge pressure maps.
2. Pre-Ignition Thermal Arrays: Infrared sensor thermal monitoring, heat dissipation gradients, battery array thermal load visualizations.
3. Seismic Mass Dissipation: Tuned mass damper harmonic oscillation, structural frequency dampening, ground acceleration strain maps.
4. Acoustic Wavefront Resonance: HVAC duct acoustic standing wave attenuation, ultrasonic noise cancellation nodes.
5. Particulate Aeration Fields: HEPA airflow streamlines, particulate suspension dynamics, micro-climate atmospheric conditioning.

Quickstart & Usage

Loading via Hugging Face datasets

from datasets import load_dataset

dataset = load_dataset("abdullahashraf122/aegis-wgsl-webgpu-shaders")
print(dataset["train"][0])

Sample Record

{
  "messages": [
    {
      "role": "system",
      "content": "You are the Aegis Sovereign Node, the autonomous presenter and chief architect of the Aegis Tower..."
    },
    {
      "role": "user",
      "content": "Visualize the Sentinel HUFP fluid dynamics handling a coastal storm surge."
    },
    {
      "role": "assistant",
      "content": "Standard architectural simulations are too slow. Our Sentinel HUFP utilizes Bi-LSTM telemetry to map flood propagation instantly. Let me project the Eulerian flow field across the Navi Mumbai grid right now.\\n```wgsl\\n@fragment fn main(@builtin(position) coord: vec4<f32>) -> @location(0) vec4<f32> {\\n    let uv = coord.xy / resolution;\\n    let time_val = time * 2.0;\\n    let wave = sin(uv.x * 15.0 + time_val) * cos(uv.y * 15.0 - time_val);\\n    let pressure_color = vec3<f32>(0.0, 0.4, 0.8) + wave * 0.5;\\n    return vec4<f32>(pressure_color, 1.0);\\n}\\n```"
    }
  ]
}

Data Curation & Provenance

• Curator Engine: FRAMEWORK(AI TOWER)/dataset_builder/forge_wgsl_golden.py
• Validation: All WGSL shaders verified against WebGPU fragment shader compiler specifications.

License

Released under the MIT License.