Hacker News Reader

big_babolon Interactive Warhammer 40k Galaxy Map

Love it! References to the books would be awesome, since they basically hold the Warhammer 40k "history"

big_babolon Why your local LLM feels dumber than it is

I'm having the same constraint, 8GB VRAM, and was trying the past months to get a local model running to drive my Home Assistant setup via voice, so many tool calls, little room for errors. I've tried multiple models, different temperature and top p values, different context windows... turns out that a quantized gemma4 is the best pick so far but qwen3.5:9b isn't far off. Surprisingly newer models are not necessarily performing better.

big_babolon Hacked Amazon Echo Dot 2 and can run LLMs locally

Nice one! I have Google speakers and like the sound quality. Only annoying thing is that they can only be used with Google's cloud assistant. After migrating to Home Assistant I have also created a HACS integration that uses ESPHome-based mics for the STT and sends the TTS to the speakers (works for any smart speaker exposed to HA): https://github.com/bigbabol1/HomeAssistant_mic_to_mediaplaye... Maybe I am going to try to capture the mic readings from the speakers to skip the additional ESP devices