Igor Barshteyn has published the source code for Athena on GitHub, a fully offline voice assistant that runs entirely on local hardware without cloud dependencies or API keys.
Athena utilizes a four-process pipeline combining Qwen3.5-397B, Orpheus 3B TTS, Whisper-small.en, and SNAC neural audio codec, implemented in C++ with zero Python at runtime. The system features natural emotional speech, real-time voice affect recognition, evolving long-term memory across sessions, and mid-sentence interruptibility.
The assistant is designed to run on a single consumer GPU and system RAM, allowing users to tune its personality and parameters via system prompts embedded in the code.