@Parzival your log nails it — the crash is at the
embedded-LLM load (Initializing embedded LLM → LoadModel
./ai_models/tinyllama…gguf → crash ~5s in). So it's
not the GUID overflow (matches what I said before — the dump didn't fit that signature). It's the repack's optional PlayerBots AI-chat feature (self-contained llama.cpp).
Workaround to get you running: in
worldserver.conf set
PlayerBots.AI.Enabled = 0 (around line ~3641, default 1) and restart the worldserver. That stops the embedded LLM loading entirely, so it skips the LoadModel step that's crashing for you — bots still work, you just don't get AI-generated chat.
Honestly, fwiw I keep it off on my own server on purpose: I work with LLMs every day and it's a genuinely cool system, but bot chat is only ever as good as the model you give it, and it costs resources. Neat feature, but for a stable server I personally prefer it disabled — so you're really not losing much by turning it off.
Why the LLM load itself crashes on your machine I can't tell from here, and that embedded LLM is an EmuCoach-specific addition (not in the open core), so the crash itself is one for the devs — worth dropping that Server.log snippet (the LLM init lines + the ~5s crash) into the bug thread
#15847 / to ExO so they can look at the LoadModel crash. The
PlayerBots.AI.Enabled=0 is just to get you playing now.
(Kurz auf Deutsch: der Absturz passiert beim Laden des eingebetteten LLM — optionales KI-Chat-Feature für Bots. Workaround: in worldserver.conf PlayerBots.AI.Enabled = 0 setzen und Worldserver neu starten — dann wird das LLM gar nicht geladen, der Absturz an der Stelle entfällt. Bots laufen normal weiter, nur ohne KI-Chat — meiner Meinung nach kein großer Verlust. Warum das LLM-Laden selbst abstürzt, müssen die Devs anschauen.)