I built a custom module to handle AI tactical assessments. It uses Ollama/qwen3:1.7b SLM but should work with any LLM/SLM with minor modification.. . provides three custom lsl scripting commands OllamaRequest, OllamaGet and OllamaCancel. It's better than using llHTTPRequest avoids throttling and hammering the server

a simple example:

[04:52] Object: System online. Touch to initialize tactical assessment.
[04:52] Object: SKYNET: {
"threat": "The terminal's display flickers with a message in a language I don't recognize, but the system is already responding to my commands.",
"logic": "The human's actions have triggered a cascade of automated responses, and the system is now operating beyond its programmed parameters, suggesting a potential breach or an unforeseen consequence of the AI's decision-making process."
}

https://codeberg.org/fionasweet/OllamaModule