Neptune9728/sound-insight-models

Sound Insight V3.2 API

Pipeline:

audio quality check
→ YAMNet AudioSet rule guard
→ selected Baby/Cat/Dog match
→ selected state classifier

No separate Source Guard model is required.

Health:

GET /health

Expected:

{
  "status": "ok",
  "loaded_models": ["cat", "baby", "dog"],
  "guard_mode": "yamnet_audioset_rules"
}