kurwov

A fast, dependency-free library for creating Markov Chains.

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

Feature	kurwov	markov-typescript	markov-generator	markov-strings	markov-chains	mrkv
Dependency-free	✔️	❌	✔️	❌	❌	✔️
Typings	✔️	❌	❌	❌ (incorrect)	❌	✔️
Generating sentences	✔️	✔️	✔️	✔️	❌ (errors)	✔️
Completing sentences	✔️	❌	❌	❌	❌	✔️
Higher statesize support	✔️	✔️	❌	✔️	❌	❌
Adding stuff other than strings	❌	✔️	❌	❌	✔️	❌
Total package size	9.8KB	1.38MB	7.9KB*	636.9KB	49.6KB	15.43KB

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 7600X @ 5.3GHz with 32gb ddr5 6000MHZ RAM on CachyOS with Node.js v26.1.0 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

Benchmark	kurwov	markov-typescript	markov-generator	markov-strings	mrkv	markov-chains, foswig
Generating a dataset with 10000 sentences.	46.31ms	690.60ms	782.95ms	996.24ms	243.54ms	N/A (errored)
Generating a dataset with 100000 sentences.	692.35ms	8332.901ms	85632.84ms	39130.50ms	2845.63ms	N/A (errored)
Creating a sentence from a 10k dataset	6026.28ns	121402.97ns	9288.93ns	420396460.80ns	11297.58ns	N/A (errored)