Tag
An experimental open-source framework that enables LLM agents to perceive and recognize non-speech audio events using local models (CLAP, Whisper, Silero VAD) and a growing concept memory. The system uses event-gated recognition, fingerprinting, and symbol-based reasoning, with no formal benchmarks yet.
A proof-of-concept open-source tool that uses CLAP embeddings to classify car faults from audio, providing calibrated triage for fault detection, location, and part identification.