config.yml reference

Last modified: 25 July 2026

The file

The plugin writes plugins/VoiceSniffer/config.yml on first start. This page is the reference for it, and it matches what the web panel generates key for key — if you would rather click than type, generate the file there and paste it in.

Unknown keys are rejected at startup rather than ignored, so a typo is reported instead of silently doing nothing.

Every key

Key	Default	Required
`engine.mode`	`none` as shipped; the panel writes `sidecar`	Yes — nothing is moderated until it is `sidecar` or `embedded`
`embedded.tier`	`standard`	Embedded only
`embedded.profile`	`safe`	Embedded only
`processor.url`	`http://127.0.0.1:8090/v1/moderate`	Sidecar only. Pinned by your subscription on cloud
`processor.token`	empty	Sidecar only. Requests are rejected with 401 without it
`processor.language`	`auto`	No
`moderation.fail-closed`	`false`	No
`actions.severity-1.type`	`warn`	No
`actions.severity-2.type`	`mute`	No
`actions.severity-2.duration-seconds`	`300`	Required for `mute`, rejected for `warn`
`actions.severity-3.type`	`mute`	No
`actions.severity-3.duration-seconds`	`3600`	Required for `mute`, rejected for `warn`
`consent.opt-out-mode`	`disable-voice`	No
`consent.message`	See consent and privacy	Yes — this is your disclosure
`capture.vad.rms-threshold`	`450`	No
`capture.vad.hangover-ms`	`600`	No
`capture.vad.minimum-utterance-ms`	`250`	No
`capture.vad.maximum-utterance-ms`	`8000`	No
`license.key`	empty	Self-host and embedded only. Cloud uses the processor token as its entitlement.

Mute durations must be between 1 and 86400 seconds. The engine, embedded and processor settings are read at startup; the embedded tier, profile and model directory need a full server restart, not a reload.

A complete file

A sidecar install with the shipped defaults — the same shape the panel emits:

engine:
  mode: sidecar
processor:
  url: http://127.0.0.1:8090/v1/moderate
  token: 'your-processor-token'
  language: auto
moderation:
  fail-closed: false
actions:
  severity-1:
    type: warn
  severity-2:
    type: mute
    duration-seconds: 300
  severity-3:
    type: mute
    duration-seconds: 3600
consent:
  opt-out-mode: disable-voice
  message: Voice chat may be transcribed for moderation. Audio is not stored.
capture:
  vad:
    rms-threshold: 450
    hangover-ms: 600
    minimum-utterance-ms: 250
    maximum-utterance-ms: 8000
license:
  key: 'YOUR-LICENCE-KEY'

For embedded mode, replace the processor block with the embedded one from engine modes.

Beyond this reference

The generated file also carries advanced timing and queue settings that the panel does not offer and that almost nobody should change — connection timeouts, queue capacities, the embedded model directory. They are commented in the file itself. This page documents the settings a server owner is expected to set; if a setting is not here, its shipped default is the answer.