Fully local conversational AI: Whisper + Hermes 8B + Kokoro, zero cloud, running inside a plush toy

A conversational AI plush toy where nothing leaves my network. Everything runs on my LAN:

• STT: Whisper

• LLM: Hermes 3 Llama 3.1 8B (Q8_0)

• TTS: Kokoro

• Memory: per-person (LangGraph). It knows who it's talking to and remembers past conversations

• Vision: face identity (dlib) + emotion recognition (FER+), so it knows who's talking and adapts to their mood

The plush itself is a dumb terminal: Pi Zero WH with mic, camera, speaker and a head servo, streaming to a laptop over WebSocket, which connects to a local AI inference API. All the intelligence is local models end to end. No cloud, no API keys.

Repo (MIT): github.com/msalsas/philosopher Full write-up: https://www.hackster.io/msalsas/the-philosopher-plush-a-diy-ai-toy-that-runs-100-local-d0d268