RimAI: Colony Advisor

RimAI is a RimWorld mod that uses Google Gemini 3.0 Flash to analyze your colony and provide intelligent, narrative updates and recommendations. It scrapes data from your save state (Colonists, Resources, Events) and sends it to the AI for a situation report.

Features

• Intelligent Colony Analysis: Reads your colonists' moods, health, and skills.
• Narrative Reports: Generates a story-based overview of your current situation.
• Strategic Recommendations: Provides 3 actionable bullet points based on your current threats and shortages.
• Gemini 3.0 Flash: Uses the latest, fastest model for near-instant responses.

Installation

1. Download the latest release.
2. Unzip into your RimWorld/Mods folder.
3. Activate in the Mod Manager.
4. Go to Mod Options > RimAI and enter your Google Gemini API Key.
   • Get a key here: Google AI Studio

For Developers: The "Zero-Touch" Networking Architecture

This mod includes a custom networking architecture designed to solve the "Mono SIGKILL" crash on macOS.

The Problem

On macOS (Apple Silicon/Intel), the Mono runtime used by RimWorld crashes (SIGKILL/Abort) if you attempt to use UnityWebRequest with large JSON payloads on the main thread.

• Cause: Large memory allocation/marshalling conflicts between Managed Mono and Unity's Native networking stack.
• Symptoms: Instant crash to desktop when sending the POST request.

The Solution

We implemented a Synchronous Background HttpClient pattern that completely bypasses Unity's networking and the main thread's memory limits for large strings.

1. Background Threading: All networking happens in a Task.Run().
2. Standard .NET Library: We use System.Net.Http.HttpClient instead of UnityWebRequest.
3. Synchronous Execution: Inside the background thread, we use synchronous calls (.GetAwaiter().GetResult()) because the Mono runtime's Async State Machine generation is also unstable on background threads in this environment.
4. Background Parsing: We deserialize the massive JSON response from Google inside the background thread using Newtonsoft.Json.Linq.JObject. We extract only the small summary text.
5. Clean Hand-off: Only the small summary text is passed back to the Main Thread via a static flag, preventing memory marshalling crashes.

Code Example

// APIClient.cs
Task.Run(() => {
    using (var client = new HttpClient()) {
        // Synchronous Send
        var response = client.PostAsync(url, content).GetAwaiter().GetResult();
        var body = response.Content.ReadAsStringAsync().GetAwaiter().GetResult();
        
        // Background Parse
        var json = JObject.Parse(body);
        string text = json["candidates"][0]["content"]["parts"][0]["text"].ToString();
        
        // Hand-off
        PendingResult = text;
    }
});

License

MIT