Refactor wake phrase handling: introduce REQUIRE_WAKE_WORD configuration, update documentation, and adjust assistant behavior for improved user experience.

2026-10-25 22:48:37 -06:00
parent ca65670638
commit a8c7db9520
7 changed files with 84 additions and 54 deletions
@@ -1,30 +0,0 @@
-# Bolt Assistant. Copy to .env and fill in the first three.
-BOLT_SERVER_URL=https://your-bolt-server.example.com
-DESK_API_KEY=
-# The voice replies are spoken in (an ElevenLabs voice id the server can use)
-ELEVENLABS_VOICE_ID=
-
-# Microphone / speaker: blank = system default, or an index / name from --devices
-#MIC_DEVICE=
-#SPEAKER_DEVICE=
-
-# Wake phrases, comma-separated (matched on the transcript, at the start of what's said)
-#WAKE_WORDS=hey bolt,bolt,thunderbolt,hey thunderbolt,ok bolt,okay bolt,hey assistant
-# Words that count anywhere in what's said; the whole sentence becomes the request
-#ANYWHERE_WAKE_WORDS=assistant
-
-# What Bolt may do on this machine when he relays a command:
-#   off | read (default: list and read files) | all (shell commands and file changes too)
-#ASSISTANT_COMMANDS=read
-
-# Local speech detection: nothing is sent to the server until someone seems to be talking.
-# Raise the threshold if background noise keeps getting transcribed.
-#VAD_RMS_THRESHOLD=500
-#VAD_SILENCE_END_SEC=0.8
-#MIN_UTTERANCE_SEC=0.35
-#MAX_UTTERANCE_SEC=12
-#COMMAND_WAIT_SEC=8
-#FOLLOW_UP_SEC=8
-#CHIMES=true
-#STREAMING_REPLIES=true
-#DESK_SESSION_ID=