You have a choice of ways to develop your visually impaired interface for the web.
One is to insist that every website can be accessed text only.
A second is to develop a browser that presents itself to websites as some browser they recognise, tracking that browser's changes but goes straight to voice. You'd also need to develop voice-based equivalents to all the other software that might be needed.
A third is to use an existing browser and only develop a screen to voice layer to sit over it, at the same time gaining access to the massive inventory of other screen-based S/W.
Which gives you more bang per buck?
BTW searching the Debian repository for Braille brings up 45 packages of which 6 are installed by default.