The GitHub repo URL returns a 404.

https://github.com/FlashLabs-AI-Corp/FlashLabs-Chroma

The end-to-end speech-to-speech claim is interesting, especially avoiding the ASR→LLM→TTS pipeline, which is where most latency and error compounding happens.