Prime Radiant Advanced Wasm

prime-radiant-advanced-wasm

Advanced mathematical AI interpretability framework compiled to WebAssembly, providing Category Theory functors, Sheaf Cohomology, Spectral Analysis, Homotopy Type Theory, Causal Inference, and Quantum Topology for understanding and auditing neural network behavior.

Quick Reference

Installation

Install: npx prime-radiant-advanced-wasm@latest See Installation Guide for the full ecosystem.

Key API

PrimeRadiant

Main interpretability analysis engine.

Node.js:

import { PrimeRadiant } from 'prime-radiant-advanced-wasm';

const pr = new PrimeRadiant({
  modules: ['spectral', 'causal', 'sheaf'],
  precision: 'f64',
});

Browser:

import init, { PrimeRadiant } from 'prime-radiant-advanced-wasm';

await init(); // Initialize WASM module

const pr = new PrimeRadiant({
  modules: ['spectral', 'causal', 'sheaf'],
  precision: 'f64',
});

Constructor Options:

Methods:

SpectralAnalysis

WASM-accelerated spectral decomposition for weight matrices.

import init, { SpectralAnalysis } from 'prime-radiant-advanced-wasm';
await init();

const spectral = new SpectralAnalysis({ precision: 'f64' });
const eigenvalues = spectral.eigenvalues(weightMatrix);
const svd = spectral.svd(weightMatrix);

Constructor Options:

Methods:

SheafCohomology

Sheaf cohomology for topological analysis of neural network structure.

import init, { SheafCohomology } from 'prime-radiant-advanced-wasm';
await init();

const sheaf = new SheafCohomology();
const result = sheaf.compute(simplicialComplex, coefficients);

Methods:

CausalInference

Causal analysis for understanding model decision pathways.

import init, { CausalInference } from 'prime-radiant-advanced-wasm';
await init();

const causal = new CausalInference({ method: 'do-calculus' });
const effect = causal.interventionalEffect(graph, intervention, outcome);

Constructor Options:

Methods:

Common Patterns

Neural Network Weight Audit

import init, { PrimeRadiant } from 'prime-radiant-advanced-wasm';

await init();
const pr = new PrimeRadiant({ modules: ['spectral', 'sheaf'] });

const report = await pr.analyze(modelWeights);
console.log(`Spectral condition: ${report.spectral.conditionNumber}`);
console.log(`Betti numbers: ${report.topology.bettiNumbers}`);
console.log(`Risk score: ${report.riskScore}`);

Browser Model Inspector

import init, { SpectralAnalysis } from 'prime-radiant-advanced-wasm';

await init();
const spectral = new SpectralAnalysis();

// Analyze each layer's weight matrix
for (const layer of model.layers) {
  const eigenvalues = spectral.eigenvalues(layer.weights);
  const gap = spectral.spectralGap(layer.weights);
  renderLayerSpectrum(layer.name, eigenvalues, gap);
}

Causal Decision Explanation

import init, { CausalInference } from 'prime-radiant-advanced-wasm';

await init();
const causal = new CausalInference({ method: 'do-calculus' });

const graph = buildDecisionGraph(modelArchitecture);
const effect = causal.interventionalEffect(graph, { feature: 'age' }, 'prediction');
console.log(`Causal effect of age on prediction: ${effect.ate}`);

RAN DDD Context

Bounded Context: Coherence/Interpretability

References

• API reference: See references/commands.md
• Full README
• npm