changed prompts

added skill and changed model back.

tie recording UI limit to engine guard

document local speech setup

wire speech recorder ui

add speech input event helpers

test speech input app events

harden speech transcription against unreadable audio and empty transcripts

add local speech transcription boundary

tighten speech boundary tests

test speech transcription boundary

spec local speech input

change tune

tune character behavior

updated picture

try to align chracters better

implement trust gate architecture

refine trust scoring architecture

add trust gate architecture spec

updated zerogpu llama version

updated llama-cpp version

added correct space

first commit, testing mvp

initial commit