NODE.md

LMM Node.js Documentation 🟩

The @wiseaidev/lmm package offers native Node.js bindings for the lmm Rust engine via napi-rs. A fully embedded Tokio runtime means all calls are synchronous; No Promises or callbacks needed for one-off calls.

📦 Installation

npm install @wiseaidev/lmm

Or build locally from source:

npm install -g @napi-rs/cli
npm run build        # builds and places the .node file in the project root

Note

Requires Node.js 22+ and a pre-built .node binary for your platform. Binaries ship for Linux (x64 / arm64), macOS (x64 / Apple Silicon), and Windows (x64).

🛠 Quick Start

node

// npm install: const lmm = require("@wiseaidev/lmm");
const lmm = require(".");

// Tensor arithmetic
const t = new lmm.Tensor([2, 3], [1, 2, 3, 4, 5, 6]);
console.log(t.shape); // [2, 3]
console.log(t.norm());

// Symbolic expression
const expr = lmm.Expression.parse("(sin(x) * 2)");
console.log(expr.evaluate({ x: Math.PI })); // ≈ 0
console.log(expr.diff("x").simplify().toString()); // (cos(x) * 2)

// Causal graph
const g = new lmm.CausalGraph();
g.addNode("x", 3.0);
g.addNode("y", null);
g.addEdge("x", "y", 2.0);
g.forwardPass();
console.log(g.counterfactual("x", 10.0, "y")); // 20.0

// Physics simulation
const osc = new lmm.HarmonicOscillator(1.0, 1.0, 0.0);
const sim = new lmm.Simulator(0.01);
const s = sim.rk4StepOsc(osc, new lmm.Tensor([2], osc.state()));
console.log(s.data);

// Text encode / decode (lossless)
const enc = lmm.encodeText("The Pharaohs encoded reality.", 80, 4);
const dec = lmm.decodeMessage(enc.expression, enc.length, enc.residuals);
console.log(dec); // The Pharaohs encoded reality.

// Symbolic text continuation
const predictor = new lmm.TextPredictor(20, 30, 3);
const result = predictor.predict("Wise AI built the first LMM", 80);
console.log(result.continuation);

// Symbolic regression
const sr = new lmm.SymbolicRegression(3, 50, 50);
const eq = sr.fit([[0.5], [1.0], [1.5]], [2.0, 3.0, 4.0]);
console.log(eq);

// Consciousness (perceive → act)
const brain = new lmm.Consciousness(4, 5, 0.01);
const state = brain.tick(Buffer.from("Hello, LMM!"));
console.log(state);

// Spectral image generation
const path = lmm.renderImage(
  "ancient egypt",
  512,
  512,
  "warm",
  "plasma",
  8,
  "out.ppm",
);
console.log("Saved:", path);

📖 Full API Reference

Classes

Export	Constructor	Key Methods
Tensor	new Tensor(shape, data)	.shape, .data, .norm(), .scale(s), .add(t), .dot(t)
Expression	Expression.parse(s)	.evaluate(obj), .diff(var), .simplify(), .toString()
CausalGraph	new CausalGraph()	.addNode(name, val), .addEdge(src, dst, w), .forwardPass(), .intervene(var, val), .getValue(name), .counterfactual(var, val, target), .topologicalOrder()
HarmonicOscillator	new HarmonicOscillator(omega, x0, v0)	.omega, .energy(), .state()
Simulator	new Simulator(stepSize)	.eulerStepOsc(model, state), .rk4StepOsc(model, state)
SymbolicRegression	new SymbolicRegression(maxDepth, iterations, popSize?)	.fit(inputs, targets) → string
TextPredictor	new TextPredictor(windowSize?, iterations?, depth?)	.predict(text, predictLength?) → PredictionResult
StochasticEnhancer	new StochasticEnhancer(probability)	.enhance(text) → string
SentenceGenerator	new SentenceGenerator(iterations?, depth?)	.generate(seed) → string
ParagraphGenerator	new ParagraphGenerator(sentenceCount?, iterations?, depth?)	.generate(seed) → string
TextSummarizer	new TextSummarizer(sentenceCount?, iterations?, depth?)	.summarize(text) → string
Consciousness	new Consciousness(stateLen, lookahead?, stepSize?)	.tick(Buffer) → number[]

Free Functions

Function	Returns	Description
encodeText(text, iterations?, depth?)	object	{expression, length, residuals}
decodeMessage(expression, length, residuals)	string	Reconstructs original text
mdlScore(expr, inputs, targets)	number	MDL fitness score
computeMse(expr, inputs, targets)	number	Mean squared error
rSquared(expr, inputs, targets)	number	R² coefficient
aicScore(nParams, logLikelihood)	number	Akaike information criterion
bicScore(nParams, nSamples, logLikelihood)	number	Bayesian information criterion
renderImage(prompt, width?, height?, palette?, style?, components?, output?)	string	Spectral field synthesis → PPM path

TextPredictor.predict() Return Object

{
  continuation:       string,  // seed + generated continuation
  trajectoryEquation: string,  // GP equation driving word tone
  rhythmEquation:     string,  // GP equation driving word length
  windowUsed:         number,  // context window size actually used
}

encodeText() Return Object

{
  expression: string,    // symbolic equation string
  length:     number,    // character count of original text
  residuals:  number[],  // integer correction residuals (lossless)
}

📄 License

Licensed under the MIT License.