RimTalk allows your RimWorld colonists to “talk” using an AI brain. You can use either a Cloud Brain (online) or a Local Brain (on your computer).

---

Pros: Powerful, easy setup, low local CPU usage
Cons: Requires an API key, may cost money

• Recommended default: Google Gemini
  
• Obtain a Google API key at https://aistudio.google.com/app/apikey
  
• Paste the API key into RimWorld → Options → Mod Settings → RimTalk → Basic → API Key
  
• Your colonists will now talk using the cloud AI.

---

Pros: Free to run, works offline, completely private
Cons: Needs a good CPU/GPU, setup is slightly more involved
To use a Local Brain, you need to run an AI model on your own machine. RimTalk connects to it via a local API.

1. Switch to Advanced Settings: In RimTalk settings → Basic tab → click the settings button to toggle from "Simple" to "Advanced".
   
2. Select the "Local Provider" radio button.

---

Running an AI model locally requires significant computer resources, primarily Video RAM (VRAM) from your GPU. Choosing the right model is a balance between performance and the quality of the dialogue.

• What is a Model Size? A model's size is measured in "parameters" (e.g., 3B for 3 billion). Larger models are often "smarter" but require more VRAM.
  
• What is Quantization? To run on consumer hardware, models are compressed in a process called quantization. This makes them much smaller. You will see different versions of the same model (e.g., Q4_K_M, Q5_K_M). Higher numbers generally mean better quality but larger VRAM usage.
  
• How to Check Your VRAM: In Windows, open Task Manager (Ctrl+Shift+Esc), go to the "Performance" tab, and select your GPU to see your dedicated VRAM.

General VRAM Guidelines:

• Less than 6 GB VRAM: Stick to smaller models, typically in the 3 Billion (3B) parameter range. Look for heavily quantized versions (e.g., Q4_K_M or lower).
  
• 8 GB - 12 GB VRAM: This is the sweet spot for many gaming PCs. You can comfortably run popular 7B to 9B parameter models, which offer a great balance of quality and performance.
  
• 16 GB VRAM or more: You can run larger, more capable models, such as those in the 13B parameter range, with high-quality quantization for better dialogue.

---

• Download LM Studio here: https://lmstudio.ai/
  
• Open LM Studio and go to the "Search" tab (🔍).
  
  • Search for a model that fits your hardware. Good options for dialogue - "Gemma3 12B GGUF 4_K_M".
    
  • Check VRAM requirements: Before downloading, look at the list on the right. LM Studio provides an estimated VRAM/RAM usage for each file version. Choose one that fits comfortably within your VRAM.
    
  • Click "Download".
• Go to the "Local Server" tab (🔌).
  
  • Select your downloaded model from the dropdown menu at the top.
    
  • Click "Start Server".
• LM Studio will show your Base URL, which is usually: http://localhost:1234 or http://127.0.0.1:1234
  
• In RimTalk settings → API tab → Local Provider Configuration:
  
  • Base URL: Paste LM Studio’s URL.
    
  • Model: The model name is not required for LM Studio, as the loaded model is handled by the server. You can leave this field blank.
    
  • AI Cooldown: Adjust to your preference (e.g., 1–2 seconds).
• Keep LM Studio running in the background while you play RimWorld.

---

• Install Ollama here: https://ollama.com/
  
• Find a model on the Ollama Library[ollama.com]. Choose one appropriate for your VRAM (e.g., gemma3:12b).
  
• Open a terminal (Command Prompt or PowerShell) and pull the model. For example:
  
  ollama pull gemma3:12b
• IMPORTANT: Find your exact model name. Ollama uses specific tags (like :latest or :12b). In the terminal, run the following command to see all the models you have installed:
  
  ollama list
• You will see a list of your models. Copy the exact name from the NAME column.
  
• Start the server in the terminal (if it's not already running) with this command:
  
  ollama serve
• The Base URL for Ollama is usually: http://localhost:11434 or http://127.0.0.1:1234
  
• In RimTalk settings → API tab → Local Provider Configuration:
  
  • Base URL: Paste the Ollama server URL.
    
  • Model: Enter the exact model name you found using the ollama list command (e.g., gemma3:12b).
    
  • AI Cooldown: Adjust to your preference.
• Keep the Ollama terminal window running in the background while you play.