config.example.yaml

# talk-rs Configuration File
# 
# This file contains all required settings for talk-rs operation.
# Replace all <PLACEHOLDER> values with your actual configuration.
# 
# For more information, see: https://github.com/talk-rs/talk-rs

# Required: Output directory for recordings, screenshots, and clipboard saves.
# The `record` command saves recordings here by default (memo-YYYY-MM-DD-HH-MM-SS.ogg).
# Must be an absolute path that exists and is writable.
output_dir: <PLACEHOLDER>

# Required: Transcription Providers Configuration
# Configure at least one provider (Mistral or OpenAI)
providers:
  # Mistral API Configuration
  mistral:
    # Your Mistral API key for authentication
    # Get this from: https://console.mistral.ai/
    api_key: <PLACEHOLDER>

    # Base URL for the Mistral API (optional, defaults to "https://api.mistral.ai")
    # Override to point at a self-hosted or API-compatible endpoint.
    # Only the base URL — path segments are appended automatically.
    # url: https://api.mistral.ai

    # Model name for transcription (optional, defaults to "voxtral-mini-2507")
    # Available models: voxtral-mini-2507, voxtral-mini-2602, voxtral-mini-latest
    # V2 models (2602+) support speaker diarization via --diarize
    # model: voxtral-mini-2507

    # Context bias: comma-separated words/phrases to improve transcription accuracy
    # Up to 100 words. Useful for proper nouns, technical terms, domain vocabulary.
    # Optimized for English; experimental for other languages.
    # context_bias: "Kalysto,talk-rs,Voxtral,cpal,tokio"

  # OpenAI API Configuration (optional)
  # openai:
  #   api_key: <PLACEHOLDER>
  #   url: https://api.openai.com               # base URL (optional, for compatible endpoints)
  #   model: whisper-1                          # batch model (whisper-1, gpt-4o-transcribe)
  #   realtime_model: gpt-4o-mini-transcribe    # realtime WebSocket model

# Optional: Default provider when --provider is not specified
# transcription:
#   default_provider: mistral

# Optional: Visual and Audio Indicators
# Configure feedback mechanisms during recording
indicators:
  # Interval between boop sounds in milliseconds
  # Set to 0 to disable audio feedback
  boop_interval_ms: 5000
  
  # Show visual indicator overlay during recording
  # Set to false to disable visual feedback
  visual_overlay: true

  # In-badge audio visualizer during recording (disabled by default).
  # Available modes: waterfall, amplitude, spectrum
  # Also settable via --viz <MODE> or TALK_RS_INDICATORS_VIZ env var.
  # viz: waterfall

  # Monochrome visualizer (theme-aware: white on dark, black on light).
  # Also settable via --mono or TALK_RS_INDICATORS_MONO env var.
  # mono: false

# Optional: Paste Behaviour
# Configure how transcribed text is pasted into the target application
# paste:
#   # Maximum characters per clipboard paste chunk (default: 150).
#   # Text longer than this is split on word boundaries into consecutive
#   # Ctrl+Shift+V keystrokes.  Keeping chunks small avoids terminal
#   # paste-summary behaviour.  Set to 0 to disable chunking (paste in
#   # one shot).  Also overridden by --no-chunk-paste (which forces 0).
#   chunk_chars: 150