node package manager


generates a markov chain of words based on input files


markovchain generates a markov chain based on text passed into it.

  • npm install markovchain
var MarkovChain = require('markovchain')
  , fs = require('fs')
  , quotes = new MarkovChain(fs.readFileSync('./quotes.txt', 'utf8'))

This will read a file, "quotes.txt", generate a word chain, then attempt to generate sentences starting with the word

"The" that are 5 words long, and then output to console.

content: the content that you want passed into the markov chain normalizeFn: the function to apply to every word (generally to clean it up, e.g. removing commas and other non-letter words)

The start method can take in either a string str, in which case it will look to use that word to start the sentence.

If you instead pass a function func with one parameter, wordList, you will be given the entire list of word chains in which you can decide what words to use to start a sentence. For example, you can generate sentences based on the number of times a word occurs, or if the word starts with a capital letter.


var useUpperCase = function(wordList) {
  var tmpList = Object.keys(wordList).filter(function(word) {
    return word[0] >= 'A' && word[0] <= 'Z'
  return tmpList[~~(Math.random()*tmpList.length)]

The end method can take a String, Integer, or Function

  • If you pass a String, str, the markov chain will generate words until the word matches str or the generator can no longer find words to chain.
  • If you pass an Integer, int, the markov chain will generate words until the sentence length matches int or the generator can no longer find words to chain.
  • If you pass a Function, func, the markov chain will generate words until function func returns true. func will be passed one parameter, sentence that returns the generated sentence so far


// same as passing value, 5 to end function 
var stopAfterFiveWords = function(sentence) {
  return sentence.split(" ").length >= 5
  • If you pass nothing in end, the markov chain will generate words until it can no longer find words to chain.

The parse function adds more content to the markov chain. This allows you to content later on, rather than needing all the next when you instantiate the markov chain.

If you pass a second parameter, parseBy, it specifies how the content will be parsed into sentences (which are then further parsed down into words). By default parseBy will parse into newlines, periods, and question marks.


var m = new MarkovChain('some text here')
m.parse('add additional text')
console.log(m.parse('more and more text').end(5).process())


  • Deprecate older version. Still accessible with this library the same exact way it was called before: var MarkovChain = require('markovchain').MarkovChain
  • This library now will be focused on the markov chain processing portion rather than file processing. Thus all file processing related functions were removed.


  • Fix when passing a string to end()


  • Added default startFn/endFn functions
  • use() now actually handles array of strings
  • Use async.parallel instead of async.series


  • Fix undefined sentence if start was passed a function


  • Passing a Function into end() has changed a little bit, before the markov chain would continue until the Function passed returned false, now the Function being passed into end() should only return true when you want the markov chain generator to stop generating the sentence.
  • start can now accept a Function instead of just a String
  • The logic to split sentences has changed from just a newline, to both a newline and a period.
  • Also previous versions changed all the words to strip them of any non-letters/numbers and also lowercased them. This version now doesn't modify the string other than to delete a period at the end of a word.


  • Small change to how words are stripped


  • Initial Release
  • MIT