kurwov v4.0.0

API

Generating a dataset.

import { MarkovData } from 'kurwov';
const sentences = ['i love hamburgers', 'i love cats'];
const chain = new MarkovChain(sentences);

Using your dataset to generate a sentence.

chain.generate(); // i love hamburgers or i love cats

Adding an sentence to the dataset.

chain.add('i love dogs');

Completing a sentence.

chain.complete({ data, start: 'i love' }); // i love dogs, i love hamburgers, or i love cats

Comparison

If you use markov chains to generate sentences, kurwov is far better than other libraries. It's the fastest and smallest package in the comparison, and has typings. * While markov-generator is only 7.9KB, it doesn't include ESM support and doesn't include typings. It also doesn't support higher statesizes which results in less code.

Speed

Benchmarks ran on a 6c/12t AMD Ryzen 5 5600 @ 4.45GHz with 32gb ddr4 3200mhz RAM on Ubuntu through WSL2 with Node.js v22.5.1 Using data from amazon q/a dataset, using the first n (10k or 100k) answers from the electronics dataset.

kurwov speed compared to other markov packages

kurwov speed compared to other markov packages with higher statesize

Benchmark | kurwov | markov-typescript | markov-strings Generating a dataset with 10000 sentences. | 182.08ms | 702.25ms | 1349.19ms Generating a dataset with 100000 sentences. | 3481.44ms | 10154.96ms | 77600.49ms