Files
local-transcription/config/default_config.yaml
Developer 8c7f4e8008
All checks were successful
Tests / Python Backend Tests (push) Successful in 14s
Tests / Frontend Tests (push) Successful in 8s
Tests / Rust Sidecar Tests (push) Successful in 2m0s
Fix sidecar pipe crash on state changes and show logged-in state in settings
The state callback in main_headless.py wrote events to stdout
synchronously, so an EINVAL on the Tauri sidecar pipe (Windows) bubbled
up through _set_state and tore down engine init and reload_engine. That
turned PUT /api/config into a "Failed to fetch" for the user. The print
is now pipe-safe and api_server isolates the chained callback so a
future misbehaving listener cannot break the engine state machine.

Settings also now persists remote.email on login and shows a "Logged in
as <email>" indicator with a Log out button when an auth_token is
present, instead of leaving the email/password fields blank on reload.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-11 18:55:43 -07:00

89 lines
3.6 KiB
YAML

user:
name: "User"
id: ""
audio:
input_device: "default"
sample_rate: 16000
transcription:
# RealtimeSTT model settings
model: "base.en" # Options: tiny, tiny.en, base, base.en, small, small.en, medium, medium.en, large-v1, large-v2, large-v3
device: "auto" # auto, cuda, cpu
language: "en"
compute_type: "default" # default, int8, float16, float32
# Realtime preview settings (optional faster preview before final transcription)
enable_realtime_transcription: false
realtime_model: "tiny.en" # Faster model for instant preview
realtime_processing_pause: 0.1 # Seconds between preview updates (lower = more responsive, default 0.1)
# VAD (Voice Activity Detection) settings
silero_sensitivity: 0.4 # 0.0-1.0, lower = more sensitive (detects more speech)
silero_use_onnx: true # Use ONNX for 2-3x faster VAD with lower CPU usage
webrtc_sensitivity: 3 # 0-3, lower = more sensitive
# Post-processing settings
post_speech_silence_duration: 0.3 # Seconds of silence before finalizing transcription
min_length_of_recording: 0.5 # Minimum recording length in seconds
min_gap_between_recordings: 0 # Minimum gap between recordings in seconds
pre_recording_buffer_duration: 0.2 # Buffer before speech starts (prevents cut-off words)
# Transcription quality settings
beam_size: 5 # Higher = better quality but slower (1-10)
initial_prompt: "" # Optional prompt to guide transcription style
# Performance settings
no_log_file: true # Disable RealtimeSTT logging
# Fast speaker mode - for speakers who talk quickly without pauses
# Reduces silence detection thresholds for more frequent transcription outputs
continuous_mode: false
server_sync:
enabled: false
url: ""
room: "default"
passphrase: ""
# Font settings are now in the display section (shared for local and server sync)
display:
show_timestamps: true
max_lines: 100
# Font settings (used for both local display and server sync)
font_source: "System Font" # Options: System Font, Web-Safe, Google Font, Custom File
font_family: "Courier" # System font name (local only, won't work with server sync)
websafe_font: "Arial" # Web-safe font name
google_font: "Roboto" # Google Font name
custom_font_file: "" # Path to custom font file (.ttf, .otf, .woff, .woff2)
font_size: 12
theme: "dark"
fade_after_seconds: 10 # Time before transcriptions fade out (0 = never fade)
# Color settings (used for both local display and server sync)
user_color: "#4CAF50" # User's name color (default green)
text_color: "#FFFFFF" # Text/font color (default white)
background_color: "#000000B3" # Background color with alpha (default semi-transparent black)
web_server:
port: 8080
host: "127.0.0.1"
remote:
mode: byok # local | managed | byok
server_url: "https://transcribe.shadowdao.com" # Proxy server URL for managed mode
auth_token: "" # JWT stored after login (managed mode)
email: "" # Email of the logged-in managed-mode account (for UI display)
byok_api_key: "" # Deepgram API key for BYOK mode
deepgram_model: nova-2 # Deepgram model to use
language: en-US # Language code
fallback_to_local: true # Fall back to local Whisper if remote fails
updates:
auto_check: true # Check for updates on startup
gitea_url: "https://repo.anhonesthost.net" # Base URL of Gitea server
owner: "streamer-tools" # Repository owner/organization name
repo: "local-transcription" # Repository name
skipped_versions: [] # List of versions the user chose to skip
last_check: "" # ISO timestamp of last update check
check_interval_hours: 24 # Hours between automatic update checks