Code dictation

Code-grammar mode applies deterministic transformations before LLM cleanup when the active preset is code.

Enablement

cleanup:
  code_grammar:
    enabled: true
    presets: [code]

The code preset is selected for terminals, VS Code, Xcode, JetBrains IDEs, Cursor, Sublime Text, and other bundle IDs in config/app_map.yaml.

Case conversions

Say a case trigger followed by two to five words:

Spoken	Output
snake case user id	user_id
camel case user id	userId
pascal case user id	UserId
kebab case user id	user-id
screaming snake user id	USER_ID

Symbol substitutions

Spoken	Output
triple equals	===
double equals	==
not equals	!=
less or equal / greater or equal	<= / >=
less than / greater than	< / >
open paren / close paren	( / )
open bracket / close bracket	[ / ]
open brace / close brace	{ / }
open angle / close angle	< / >
fat arrow / arrow	=> / ->
double pipe / pipe	|| / |
double ampersand / ampersand	&& / &
at sign, hash, dollar	@, #, $
underscore, dash, hyphen	_, -, -
plus, minus, star, slash, backslash	+, -, *, /, \
percent, caret, tilde, bang, question	%, ^, ~, !, ?
quote, string, single quote, apostrophe, backtick	", ", ', ', `
semicolon, colon, comma, dot, period	;, :, ,, ., .
new line, tab	newline, tab

Examples

Before cleanup:

const snake case user id equals request dot params dot id semicolon

After code grammar:

const user_id = request.params.id;

Before cleanup:

if open paren camel case is active double ampersand count greater than zero close paren open brace new line return true semicolon new line close brace

After code grammar:

if (isActive && count > zero) {
return true;
}