Skip to content

Home

Jump to bottom
Wiki Sync Script edited this page Jul 10, 2026 · 7 revisions

RimSynapse - Core

Table of Contents

About RimSynapse - Core

RimSynapse - Core

The foundational LM Studio and Ollama Connector for RimWorld.

RimSynapse Core is a powerful, asynchronous background engine that allows RimWorld mods to communicate with Large Language Models (LLMs). On its own, this mod does not change gameplay, but it serves as the essential bridge required by all RimSynapse companion mods to bring true narrative intelligence to the Rim.

Features

  • Universal Backend Support: Connects seamlessly to LM Studio, Ollama, or any OpenAI-compatible local API.
  • Background Threading: Never drops a frame. All LLM requests are queued, processed, and sanitized asynchronously on separate threads.
  • Dynamic Priority Queue: Intelligently manages hundreds of requests with an advanced Linux-style top-monitor queue system, prioritizing critical events and starving obsolete prompts.
  • Auto-Discovery: Automatically detects your active models and manages keep-alive pings to ensure your LLMs don't sleep when you need them.
  • GPU Monitoring: Built-in framework to keep track of local VRAM and processing overhead.
  • Native Event Interception: Automatically captures major vanilla game events (marriages, legendary crafts, pawn deaths) via Harmony patches and routes them to companion mods for AI processing.
  • Faction Perception Tracking: Each faction maintains subjective "Perceived Wealth" and "Perceived Strength" values for your colony, updated only through direct contact (visitors, traders, caravans).
  • Memory Infrastructure: Provides the per-pawn memory bank, weighted tag system, and decay engine used by companion mods to give colonists persistent, evolving memories.
  • Colony Snapshots: When events occur, the framework captures a snapshot of the colony's wealth, nutrition, and pawn status, providing rich context for AI-generated responses.
  • Opportunistic Task Scheduling: A background task manager that feeds low-priority LLM requests (backstory generation, memory creation) into idle GPU cycles without impacting gameplay performance.

A Personal Note

AI is incredibly cool, and having a framework to utilize powerful, no-cost local models really appealed to me. RimWorld is relatively light on GPU overhead, making it the absolute perfect candidate to max out those graphics cards by running a local LLM in the background. My goal with RimSynapse is to push the boundaries of emergent storytelling using the hardware you already own.

Roadmap and Future Goals

While this framework was built with local, free models in mind (via LM Studio), I want to make sure everyone can enjoy narrative AI. My future goals for the core framework include:

  • Built-in API configurations for ChatGPT (OpenAI)
  • Built-in API configurations for Gemini (Google)
  • Built-in API configurations for Claude (Anthropic)

For Developers and Open Source

RimSynapse is an open-source framework designed to be easily hooked into by other modders. I highly encourage you to read the framework documentation and build your own LLM-driven mods!

Requirements

  • Harmony is required.
  • You must be running a local LLM server (like LM Studio or Ollama). We recommend 8B+ parameter models for optimal reasoning.

Support the Project

If you enjoy this framework and want to support the development of genuine AI-driven gameplay, consider buying me a coffee! ☕ Buy me a coffee

Clone this wiki locally