First-run setup: RIMAPI integration gaps and setup friction

[CalebisGross]

Summary

Documenting the setup experience getting RLE running against a live RimWorld instance with the RIMAPI mod for the first time.

Issues encountered

1. Steam Workshop RIMAPI is insufficient

The upstream Steam Workshop version of RIMAPI is missing endpoints that RLE depends on. The AppSprout-dev fork (rle-testing branch) has the required additions (research target, pawn job, set-forbidden, etc.).

Users must clone and build from AppSprout-dev/RIMAPI on the rle-testing branch — not the Steam Workshop version.

2. RIMAPI mod requires local build from source

The RIMAPI mod is a C# project targeting net472. Building requires .NET SDK:

sudo apt install dotnet-sdk-8.0
cd RIMAPI/Source/RIMAPI
dotnet build RimApi.csproj -c Release-1.6

Then symlink into RimWorld's Mods folder:

ln -s /path/to/RIMAPI ~/.steam/steam/steamapps/common/RimWorld/Mods/RIMAPI

Note: build via the .csproj directly — the .sln doesn't have the Release-1.6 configuration mapped.

3. .env configuration for local models undocumented

The .env.example only shows Anthropic/OpenAI config. For LM Studio (local inference), the correct config is:

RIMAPI_URL=http://localhost:8765
PROVIDER=local
MODEL=<model-name>
PROVIDER_BASE_URL=http://localhost:1234/v1

No API keys needed for local providers.

4. LM Studio model reload crashes mid-run

LM Studio returns 400 - {'error': 'Model reloaded.'} if the model is swapped or reloaded while RLE is running. This crashes the scenario with an unhandled ProviderError.

Suggestion: Add retry logic in the game loop for transient provider errors, or at minimum catch and skip the tick rather than crashing.

Recommended actions

• Add a first-run setup guide to README covering RIMAPI fork setup, build, and symlink
• Document .env configuration for all three provider types (anthropic, openai, local)
• Consider graceful error handling for transient LLM provider failures
• Note in docs that Steam Workshop RIMAPI is insufficient — the AppSprout fork on rle-testing is required