StreamBot/config.yaml at main · Kisahime/StreamBot · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
server:
  host: 127.0.0.1
  port: 8080
  # The booth has no auth. Stay on localhost unless you set this true on purpose.
  allow_remote: false

ollama:
  host: http://127.0.0.1:11434
  # snappy profile keeps this 3B model. quality profile switches a 3B default to the 7B below.
  model: qwen2.5:3b
  timeout_seconds: 120
  num_ctx: 4096
  num_predict_chat: 80
  num_predict_voice: 160
  keep_alive: 30m
  # Bounds chat replies, alert replies, and the recap vibe paragraph.
  chat_timeout_seconds: 20

# snappy: Whisper on CUDA + qwen2.5:3b (fits an 8 GB GPU with Whisper).
# quality: forces Whisper onto CPU and, if the model is still the 3B default, uses the 7B.
vram:
  profile: snappy
  warn_mb: 7000
  quality_model: qwen2.5:7b-instruct-q4_K_M

whisper:
  model_size: small
  # auto | cuda | cpu  — auto uses CUDA when ctranslate2 sees an NVIDIA GPU
  device: auto
  language: en
  compute_type_cuda: float16
  compute_type_cpu: int8

tts:
  enabled: true
  mute: false
  # Cute US female (Amy). Other females: en_US-kristin-medium, en_US-hfc_female-medium
  voice: en_US-amy-medium
  length_scale: 1.05
  piper_dir: models/piper
  max_sentence_chars: 280
  voices:
    - en_US-amy-medium
    - en_US-kristin-medium
    - en_US-hfc_female-medium

audio:
  sample_rate: 16000
  mic_device: null          # null = system default; override with a device index
  mic_device_name: null     # re-resolved at startup if the index moves
  speaker_device: null
  speaker_device_name: null
  silence_ms: 800
  min_utterance_ms: 350
  energy_threshold: 0.012   # RMS; raise if it false-triggers
  max_utterance_seconds: 20
  # Optional per-app session ducking. Off by default. On Windows, pycaw lowers matching
  # app sessions while Pixel talks. Pixel's own process is never ducked. An empty
  # duck_apps list ducks every other app except duck_skip. Without pycaw, use an OBS sidechain.
  duck_enabled: false
  duck_gain: 0.25
  duck_apps: []
  duck_skip:
    - obs64.exe
    - Discord.exe
  # Talk-over stops Pixel. Echo gate ignores speaker bleed unless PTT is held or the level is much louder.
  barge_in: true
  echo_gate: true
  echo_gate_ratio: 2.5
  # Local wake word via openwakeword. Off by default, and never the only way to talk.
  wake_word: false
  wake_models:
    - hey_jarvis
  wake_threshold: 0.5

chat:
  buffer_size: 40
  cooldown_seconds: 12
  max_replies_per_minute: 8
  scan_interval_seconds: 3
  mention_boost: true
  post_voice_replies_to_chat: true
  chat_max_chars: 200
  per_user_cooldown_seconds: 20
  auto_reconnect: true
  min_send_gap_seconds: 1.2
  ambient_context_lines: 6
  commands:
    - "!so"
    - "!pixel"
    - "!ask"
  # !so is also a mod command: viewers are dropped with skip reason "not a mod".
  mod_commands:
    - "!so"
  shoutout_cooldown_seconds: 120
  # Pending viewer lines older than this are dropped (streamer lines and commands stay).
  max_age_seconds: 45
  # First-time chatters get a short typed welcome instead of a spoken reply.
  welcome_first_time: true
  # Gap between mod !pixel mute/unmute/chill/hype/stop commands.
  control_cooldown_seconds: 4

memory:
  chat_context_lines: 20
  events: 80
  viewers: 48
  thread_turns: 12
  catchphrase_seconds: 600
  retention_days: 14

events:
  enabled: true
  announce_follows: true
  follow_batch_seconds: 8
  speak: true
  post_to_chat: true
  # Each of these is off until you turn it on. Missing scopes skip that feed only.
  bits: false
  channel_points: false
  hype_train: false
  polls: false
  predictions: false
  channel_point_rewards:
    hydrate: speak
    # A reward can also be a mapping. {user} and {input} are filled in, and input is fenced.
    # "roast me":
    #   action: speak
    #   prompt: "Playfully roast {user} in one line, nothing mean about looks. They wrote: {input}"
    #   cooldown: 120

overlay:
  alert_hold_seconds: 6

safety:
  # Pixel strips links from chat posts unless this is true.
  allow_links: false

# Named booth presets. POST /api/profile applies one and stores the name in data/booth.json.
profiles:
  "Just chatting":
    mood: hype
    cooldown_seconds: 8
    per_user_cooldown_seconds: 12
    events.speak: true
    post_voice_replies_to_chat: true
    max_replies_per_minute: 12
  "Focus game":
    mood: chill
    cooldown_seconds: 30
    per_user_cooldown_seconds: 45
    events.speak: false
    post_voice_replies_to_chat: false
    max_replies_per_minute: 3