Aura AI Plugin

AI/tensor Nexus integration for shape-aware model and inference checks in supported paths.

Scope

AI/tensor Nexus integration for shape-aware model and inference checks in supported paths.

• The repository contains aura-plugin-ai, aura-ai-opt and ONNX Runtime bridge/example assets.
• Tensor/model shape information can participate in plugin/verifier logic where the implementation supports it.
• This is not a claim that every ONNX operator, model or native runtime path is formally verified.

Enable

plugins = [
  { name = "aura-ai", trusted = true },
]

Marking a plugin as trusted places its trusted behavior inside the project trust boundary; it does not automatically prove external hardware or native libraries.

Example

import aura::tensor

cell main() ->:
    val model: Model = ai.load_model("model.onnx")
    val input: Tensor<u32, [2, 2, 3]> = tensor::new<u32>(12)
    val out = model.infer(input)

Editor feedback

Where verification/plugin diagnostics are produced, aura-lsp can expose plugin-attributed information to Aura-aware clients such as Sentinel. Exact diagnostics depend on the current plugin implementation and configured project model.