It's a twist in the classic gamebook mechanics where instead of navigating an explicit menu of options, you are a given a console in which you are free to type any command, after each section, using clues from the text. The engine then tries to match your input with one of the predefined options, yielding a gameplay more akin to interactive fiction.

Try it!

The gamebook data is entirely contained in a JSON file with a simple structure. This means that the engine operates client-side only, with no need for any server interaction. The main object contains a sections object with a key corresponding to each section:

{
    "sections": {
        "1": {
            "text": "Captain D’Val and his guards...",
            "options": [{
                "section": "273",
                "text": "If you wish to draw your weapon and attack your unknown assailant,\nturn to 273.",
                "words": ["weapon", "attack", "assailant", "combat"]
            }, {
                "section": "160",
                "text": "If you wish to try to pull free of his grasp, turn to 160.",
                "words": ["pull", "free", "grasp", "struggle", "assailant"]
            }]
        },
    ...
    }
}

The engine will tokenize the player's input and try to match it against each option's list of words, with a certain tolerance (using the Levenshtein distance). The option with the greatest number of matches is considered the chosen one (a confirmation is asked though).

Of course this mechanism is so simple that it could hardly be defined as a command parser, but in the context of a gamebook, where there are only a few options per section anyway, it makes sense. To increase the illusion that it's a "real" (or deeper) parser, it has a few more features:

• the engine uses a dictionary of synonyms (defined in the data file) to expand the game vocabulary

• an ambiguous command (resulting in two options having the same number of matches, for instance) will often gently push the player into making more complex entries; e.g. "assailant" alone would be ambiguous in the example above, whereas "attack assailant" wouldn't be

• there is a syntax for compound commands, for which each word must be matched, for example:

"words": [["don't", "want"], "attack"]

would match "don't want" (or even "don't really want"), but not "want" alone.

The console environment was created using the jQuery Terminal plugin.