Argon Echo Bot

A voice bot for Argon that answers calls, records the caller's speech, and plays it back — like a voice echo. After the echo it loops background music until the caller hangs up.

Built on the Argon Bot API and AudioWS (WebSocket audio gateway).

How it works

Each audio variant (girl_echo_*.opus) is a pre-recorded script with timestamp markers (.json):

Phase	Description
Greeting	Bot speaks an intro line
Recording	Bot listens while the caller talks
Transition	Short bridging line
Playback	Bot plays the caller's recorded audio back
Outro	Wrapping up
Background	Loops background.opus until hangup

The bot uses a single duplex WebSocket to both publish and subscribe to audio.

Voice selection

When a call comes in, the CallIncoming event carries the caller's language as fromLocale (a BCP-47 tag like ru, en, ja — Bot API ≥ 1.10.0). The bot uses it to pick a voice:

1. The locale is reduced to its primary language subtag (en-US → en).
2. If that language has a voice bucket, the bot picks from it. Otherwise it uses the fallback_language bucket (default en). If even that is empty, it falls back to any available voice so the call is always answered.
3. Within the chosen bucket, a voice is chosen by a rarity-weighted random pick, so rare presets surface only occasionally.

Configure voices per language under voices. Each preset can declare a rarity tier — common (default), uncommon, rare, legendary — or an explicit numeric weight that overrides the tier. Relative tier weights: common 100, uncommon 30, rare 8, legendary 1 (higher = picked more often).

Preset files live under <voices_dir>/<language>/, where voices_dir defaults to voices. Each preset name resolves to two files in that language folder: <name>.opus (the audio) and <name>.json (its phase markers). The same name in two languages is two different recordings.

voices/
  ru/  girl_echo_00.opus  girl_echo_00.json  …
  en/  girl_echo_00.opus  girl_echo_00.json  …

{
  "fallback_language": "en",
  "voices_dir": "voices",
  "voices": {
    "ru": [
      { "name": "girl_echo_00", "rarity": "rare" },
      { "name": "girl_echo_01" },
      { "name": "girl_echo_02", "rarity": "legendary" }
    ],
    "en": [
      { "name": "girl_echo_00" },
      { "name": "girl_echo_06", "rarity": "legendary" }
    ]
  }
}

Quick start

cp config.json config.json   # edit token, api_base, ingress_url
export BOT_TOKEN="your-bot-token"
go run .

Configuration

Edit config.json or use environment variables:

Config key	Env var	Default	Description
token	BOT_TOKEN	—	Bot API token (required)
api_base	API_BASE	https://gateway.argon.zone	Argon API base URL
ingress_url	INGRESS_URL	ws://localhost:12880	AudioWS ingress URL
log_level	LOG_LEVEL	info	debug, info, warn, error
fallback_language	—	en	Language bucket used for unknown/missing caller locales
voices_dir	—	voices	Base directory holding per-language preset folders
voices	—	see config	Per-language voice presets (see Voice selection)
background_file	—	background.opus	Background music file

Environment variables take priority over config file values.

Legacy variants: a flat "variants": ["girl_echo_00", …] list is still accepted. When voices is omitted, those names are mapped into the fallback_language bucket with common rarity.

Docker

docker build -t argon-echo-bot .
docker run -e BOT_TOKEN="your-token" argon-echo-bot

Audio files

• voices/<lang>/girl_echo_*.opus — pre-recorded voice presets (OGG/Opus, 48 kHz mono)
• voices/<lang>/girl_echo_*.json — timestamp markers for each preset
• background.opus — background music loop

You can replace these with your own recordings — just match the JSON marker format and update the voices map in config.json.

The bot only ships .opus. If you keep lossless .flac masters in the voices/ tree they are git-ignored; transcode them to 48 kHz mono Opus, e.g. ffmpeg -i in.flac -c:a libopus -b:a 48k -ar 48000 -ac 1 -frame_duration 20 out.opus.

License

Source code is licensed under the MIT License.

Voice recordings (girl_echo_*) are proprietary assets owned by Argon Inc. LLC and included for demo purposes only. background.opus is "Classical Music Uplifting Symphony Loop" by Sonican (royalty-free). See LICENSE for details.