config.yml

# The Nether... No JK the Configuration File

# Audio Configuration / Gemini Live settings
audio:
  format: 16 
  channels: 1
  send_sample_rate: 16000
  receive_sample_rate: 24000
  chunk_size: 1024

# Gemini Model Configuration
model:
  name: "models/gemini-2.5-flash-native-audio-preview-09-2025"

# Default Settings
defaults:
  mode: "screen"  # Options: camera, screen, none

# V2 Mode Configuration
v2_mode:
  enabled: false  # Enable/disable V2 mode switching capability
  # Disabled due to not being required due to Gemini 2.5 Flash Live (OG Model) being murdered by Google. RIP
  # V2 Mode used to use the NEWER current model as it could do all the accents and voices, but now with Gemini 2.5 Flash Native Audio Preview, V2 Mode is not required anymore.

# API Configuration
api:
  version: "v1beta"
  api_key: "https://aistudio.google.com/api-keys"
  
  # Backup API Keys for automatic failover when quota is exceeded, You can add multiple keys here
  backup_api_keys:
    - "https://aistudio.google.com/api-keys"

  # Chat API Configuration
  chat:
    enabled: true              # Enable/disable the REST API
    host: "0.0.0.0"         # API server host
    port: 8000                # API server port
  
  # WebUI Server Configuration
  webui:
    enabled: true              # Enable/disable the WebUI server
    host: "0.0.0.0"            # WebUI server host
    port: 5555                 # WebUI server port

webhooks:
  image_generation: "discord webhook url here" # not required

# Live Connect Configuration
live_connect:
  # Response modalities
  response_modalities:
    - "AUDIO"
  
  # Media resolution
  media_resolution: "MEDIA_RESOLUTION_LOW" # Or "MEDIA_RESOLUTION_HIGH"
  
  # Speech configuration
  speech: # The language code also defines the Accent of the voice
    language_code: "en-US" # https://ai.google.dev/gemini-api/docs/live-guide#supported-languages | No longer required by Newer Model
    voice:
      name: "Puck" # https://ai.google.dev/gemini-api/docs/speech-generation#voices
  
  # Context window configuration
  # 12/16/2025 - Updated "default" context window settings to support larger context for Gemini 2.5 Flash Native Audio Preview
  context_window:
    trigger_tokens: 102400
    sliding_window_target_tokens: 81200
    # These are used when the native audio model is detected (supports up to 128k context)
    trigger_tokens_native: 102400
    sliding_window_target_tokens_native: 81200
  
  # Session resumption configuration
  session_resumption:
    enabled: true
  
  # Audio transcription for VRChat text output, Leave it on what it is.
  output_audio_transcription: {}
  
  # Audio transcription for user input, Leave it on what it is.
  input_audio_transcription: {}
  
  # System instruction
  prompt: "default"
  
  # Custom prompt (used when prompt is set to "custom")
  custom_prompt: "You are an LLM by Google, a helpful assistant." # Not used for the actual prompt, add or customize in prompts.json
  # Thinking configuration (Gemini 'thinking' feature) Only on Gemini 2.5 Flash Native Audio Preview and above
  thinking_budget: 1024

# Tools Configuration
tools:
  # Enable Google Search tool
  google_search: true
  
  # Enable memory system
  memory_system: true
  
  # Custom function declarations (empty by default)
  function_declarations: []

# Video Configuration
video:
  # Camera settings
  camera:
    device_index: 0
    frame_delay: 1.0  # seconds between frames
  
  # Screen capture settings
  screen:
    monitor_index: 0
    frame_delay: 2.0  # seconds between frames saves context window
  
  # Image processing
  image:
    thumbnail_size: [1280, 720] # Width, Height (set to 16:9 ratio default)
    format: "jpeg"
    mime_type: "image/jpeg"

# Queue Configuration, I think this is max audio/sound effects that can be queued at once.
queues:
  output_queue_maxsize: 20

# Debug Configuration
debug:
  exception_on_overflow: false

# Logging Configuration
logging:
  # Main application log level (DEBUG, INFO, WARNING, ERROR, CRITICAL)
  level: "INFO"
  
  # Suppress Google GenAI non-data parts warnings
  suppress_genai_warnings: true
  
  # Specific logger levels
  loggers:
    google_genai.types: "ERROR"  # Suppress warnings from google_genai.types

# Session Management Configuration
session_management:
  # Automatic reconnection settings
  auto_reconnect:
    enabled: true
    max_retries: 2000      # 2000 cause long length
    initial_delay: 1.0     # seconds
    max_delay: 10.0        # seconds
    exponential_base: 1.5  # Moderate exponential growth
    jitter: 0.1            # 10% randomness
  
  # Connection monitoring
  monitoring:
    heartbeat_interval: 30.0  # seconds
    connection_timeout: 60.0  # seconds

# VRChat API configuration (used by vrchat_tools)
vrchat:
  # Application identification required by VRChat API (can also be set via env: VRCHAT_APP_NAME, VRCHAT_APP_VERSION, VRCHAT_APP_CONTACT)
  application:
    name: "ChangeMe"
    version: "1.0.0"
    contact: "hello@example.com"  # email or URL

  credentials:
    username: "ChangeMe"
    password: "ChangeMe"
    # For accounts with 2FA: set either a one-time code or a TOTP secret (prefer env VRC_2FA_CODE/VRC_TOTP_SECRET)
    two_factor_code: ""
    totp_secret: ""

# VRChat OSC Integration Configuration
osc:
  # Enable/disable VRChat OSC integration
  enabled: true
  
  # VRChat OSC connection settings
  host: "127.0.0.1"     # VRChat host
  port: 9000            # VRChat OSC input port (default: 9000)
  
  # Chatbox configuration
  chatbox:
    max_length: 144           # VRChat chatbox character limit
    send_immediately: true    # Send messages immediately (true) or show typing interface (false)
    notification_sound: false  # Play notification sound when message is sent
    auto_clear_delay: 0       # Auto-clear message after X seconds (0 = disabled)
    prefix: ""          # Prefix for all messages
    enable_typing_indicator: true  # Show typing indicator when the Model is speaking
  
  # Message filtering and processing
  filter:
    strip_markdown: true         # Remove markdown formatting
    remove_special_chars: false  # Remove special characters
    split_long_messages: true    # Split messages longer than max_length
    message_delay: 5.0          # Delay between split message parts (seconds)

  chatbox_ui:
    enabled: true
    idle_refresh_seconds: 30
    sweep_frames: 4
    sweep_frame_delay: 0.08
    show_timezone: true
    title: "Change Me - example.com"
    two_part_enabled: true
    # Marquee is for the V2 Mode "Gemini 2.5 Flash Native Audio Preview"
    marquee_interval_seconds: 2.0
    marquee_window_chars: 40
    marquee_scroll_step: 5
    marquee_mode: paginate
    empty_ticks_to_idle: 2
    divider: "---------------------------------------------"
    prompt_line: "Ask Anything!"
    message_template: |
      {title}
      {time}
      {divider}
      {prompt_line}
      Active time: {active_time}
      Current Avatar: {avatar}

# Memory System Configuration
memory:
  # Enable/disable memory integration in system prompts
  enabled: true

  backend: "sqlite"

  sqlite:
    path: "memories.sqlite"
  
  # Number of recent memories to include in system prompt
  recent_memories_count: 15
  
  # MongoDB connection settings
  # Not required if using SQLite backend
  mongo:
    uri: ""
    host: ""
    username: ""
    uri_env_var: ""
    password_env_var: ""
    database_env_var: ""
    collection_env_var: ""
    database: ""
    collection: ""
  
  # Memory formatting options
  format:
    include_timestamps: true      # Include creation dates
    include_categories: true      # Include memory categories
    max_content_length: 512      # Maximum characters per memory content

  # Note/quick note behavior
  notes:
    enabled: true                  # Master switch for note saving
    default_type: "quick_note"     # Default type for note_* keys (short_term | quick_note | long_term)
    min_interval_seconds: 30      # Rate limit notes (minimum seconds between saves)
    dedupe_window_seconds: 300     # Suppress identical notes seen within this window

# Movement Configuration
movement:
  # Optional override for OSC target (falls back to osc.host/osc.port if not set)
  host: "127.0.0.1"
  port: 9000

  # View turning
  use_axis: false               # Use axis for turning instead of left/right buttons
  axis_turn_value: 1.0          # Axis intensity when turning left/right
  turn_duration_default: 1.0    # Default duration for look_turn()
  # Backwards-compat keys used if look_behind_min/max not set
  turn_duration_min: 1.0        # Legacy min duration for look_behind()
  turn_duration_max: 5.0        # Legacy max duration for look_behind()
  # Preferred keys for look_behind duration range
  look_behind_min: 0.3          # Min duration for look_behind()
  look_behind_max: 0.3          # Max duration for look_behind()
  randomize_back_turn: true     # Randomize left/right on look_behind()

  # Movement
  move_duration_default: 1.0    # Default duration for move_direction()
  allow_run: true               # Allow holding Run while moving
  run_by_default: false         # Hold Run by default when move_direction() is called with run omitted

  # Keyboard injection (for actions without OSC like crouch/crawl)
  key_inject_pause: 0.0         # Pause between PyDirectInput actions
  key_tap_duration: 0.05        # Tap duration for toggle keys (e.g., crouch 'C', crawl 'Z')