A convenient API for creating dragonfly grammars with automatic CCR (continuous command recognition).
- Very quick start-up
- Command activity can be controlled either using dragonfly contexts or using "enable" and "disable" commands.
- All commands which match the current context may be chained together in any order in the same utterance.
pip install dfly-breathe
- If you are creating a command set from scratch, start by cloning the Breathe skeleton project, which will give you a file structure to start with.
from dragonfly import *
from breathe import Breathe, CommandContext
Breathe.add_commands(
# Commands will be active either when we are editing a python file
# or after we say "enable python". pass None for the commands to be global.
context = AppContext(title=".py") | CommandContext("python"),
mapping = {
"for each" : Text("for in :") + Key("left:5"),
"for loop" : Text("for i in range():") + Key("left:2"),
"from import" : Text("from import ") + Key("home, right:5"),
"function" : Text("def ():") + Key("left:3"),
"(iffae | iffy)" : Text("if :") + Key("left"),
"iffae not" : Text("if not :") + Key("left"),
"import" : Text("import "),
"lambda" : Text("lambda :") + Key("left"),
"while loop" : Text("while :") + Key("left"),
"shell iffae" : Text("elif :") + Key("left"),
"shells" : Text("else:"),
"return" : Text("return "),
# ------------------------------------------------
"method <snaketext>" : Text("def %(snaketext)s(self):") + Key("left:2"),
"function [<snaketext>]": Text("def %(snaketext)s():") + Key("left:2"),
"selfie [<snaketext>]" : Text("self.%(snaketext)s"),
"pointer [<snaketext>]" : Text(".%(snaketext)s"),
"classy [<classtext>]" : Text("class %(classtext)s:") + Key("left"),
},
extras = [
Dictation("snaketext", default="").lower().replace(" ", "_"),
Dictation("classtext", default="").title().replace(" ", ""),
]
)
For full details of the available contexts, actions and extras you can use, see the dragonfly documentation.
Breathe provides the command "rebuild everything" for reloading all of your commands,
allowing you to modify commands without restarting the engine. In order for this to work,
your command files need to be loaded by giving your directory structure to
Breathe.load_modules()
.
For example, given a directory set up like this:
| _main.py
| __init__.py
+---my_commands
| | __init__.py
| +---apps
| | chrome.py
| | notepad.py
| | __init__.py
| +---core
| | alphabet.py
| | keys.py
| | __init__.py
| +---language
| | c.py
| | python.py
| | __init__.py
Inside _main.py
, the file which will be loaded by the engine, we load all of our command
files by passing a dictionary with keys representing folder names and values being either a
single module to import, a list of modules to import, or another dictionary. Like so:
from breathe import Breathe
Breathe.load_modules(
{
"my_commands": {
"apps": ["chrome", "notepad"],
"language": ["python", "c"],
"core": ["keys", "alphabet"],
}
}
)
Given this setup, calling the "rebuild everything" command will reload all of your command files, making any changes available.
Advanced feature, if you are just getting started you should ignore this.
Top level commands allow you to embed sequences of breathe CCR commands inside other commands. This gives finer control over the way in which commands are recognised and executed.
Top level commands should be added in a separate add_commands
call
with the top_level
option set to True
. A couple of new elements -
Exec
and CommandsRef
- are required to control them.
For example in the following, the first command implements "greedy" dictation by creating a top level command which recognises between zero and twelve of the commands which are active in the current context, followed by a dictation command which will consume the rest of the utterance. The second allows an arbitrary sequence of commands to be repeated a given number of times.
from dragonfly import *
from breathe import Breathe, CommandsRef, Exec
Breathe.add_commands(
None,
{
"[<sequence_of_commands>] dictate <text>":
Exec("sequence_of_commands") + Text("%(text)s"),
"<sequence_of_commands> and repeat that <n> times":
Exec("sequence_of_commands") * Repeat("n"),
},
[
Dictation("text"),
IntegerRef("n", 1, 100),
CommandsRef("sequence_of_commands", 12)
],
top_level=True
)