Voice Communication With ARC And A Framework

Hello, I use to communicate with ARC, SARAH framework (NodeJs). Through this framework I use the grammar of an XML file which uses Windows Speech recognition and for more specific words such as research on wikipedia I use api speech Google. I use this way of keywords, for exemple: Victor turn left, Victor can you turn left ?, Victor I want you turn on the left now. The keywords are "turn" and "left." Framework and ARC communicate via HTTP. But I would go further by retrieving or audio or xml, txt or other the voice control to use such recognition for exemple with Pandorabots or script. Is it possible? When we pronounce a sentence with ARC, are there a particular file that retrieves his voice command. Thank you