Skip to main content

Troubleshooting

IsLLMReady() returns false​

  • Confirm a model is downloaded (Tools → Elys Mind → Download LLM Models) and LLM Model Path in Project Settings → Elys Mind (LLM) points to a .gguf.
  • Confirm LLM Backend is llama.cpp (Local) and Enable LLM is on.
  • Model loading takes several seconds; check the log for Backend initialized.

Out of memory / crash on load​

  • Use a smaller model or a more aggressive quantization (Q4_K_M is a good default).
  • Lower Max Context Length.
  • If GPU offload fails, turn off Enable GPU Acceleration to run CPU-only.

Responses are slow​

  • Smaller models and shorter Max Response Tokens respond faster.
  • Enable GPU acceleration where available.
  • Generation runs off the game thread; the first call after load is slowest (warm-up).

Responses are rambling or off-character​

  • Tighten the System Prompt ("Answer in one or two sentences.").
  • Lower Temperature (e.g. 0.5) for more deterministic replies.
  • Add Stop Sequences to cut generation at natural boundaries.

Packaged build can't find the model​

  • Ensure Content/ElysMind/Models/ is packaged (Packaging → Additional Non-Asset Directories to Package) or ship the .gguf and set an absolute path.

Still stuck? Ask on the Discord or via FAB Q&A.