altbot v2.5 is out!
fixed non-english support! switched to a two model setup: gemma3:1b handles translations while qwen3-vl does the image processing stuff. gemma3:1b is surprisingly good at translations for such a tiny model??
now running 23.7gb out of 24gb vram on my server, but everything stays hot and loaded so no cold start penalty (which would have been very bad for power usage especially)
also pixelfed should work again since thes gdpr update (hopefully?? i don't have an account to test with so if you're on pixelfed lmk if it breaks)
https://github.com/micr0-dev/Altbot
edit: reworded "vision stuff" to "image processing stuff" poor word choice on my part, was thinking about the term VLLM (Vision Large Language Model) when writing that