config.yml

Global configuration options for Genius

Default config.yml

# *****************************************
# *         Genius Configuration          *
# *****************************************
# For reference, see https://github.com/fletchly/genius/wiki/Configuration

# Properties for the agent's appearance and behavior
genius:
  # The name for the Genius agent to use in chat when responding to messages
  agentName: Genius

# Ollama client configuration
ollama:
  # The base URL for the Ollama API. If using a cloud model, this should be set to https://ollama.com/.
  # Defaults to the Ollama API's default listening location. For more info, see https://github.com/fletchly/genius/wiki/hosting
  baseUrl: http://localhost:11434/

  # Your key for the Ollama cloud API. This only needs to be set if you are using an Ollama cloud model.
  # For more info, see https://github.com/fletchly/genius/wiki/hosting
  apiKey: API_KEY

  # The name of the model to use for response generation.
  # If you are using Ollama cloud, the model must be an ollama cloud model
  model: deepseek-v3.1:671b

  # Controls how random or deterministic Genius's output is.
  # Lower values make responses more predictable and precise,
  # while higher values make them more creative and varied.
  temperature: 0.5

  # Top-K sampling limits the model’s next-token choices to the K most likely options, filtering out all others.
  # This reduces randomness while still allowing some creativity compared to always picking the single most probable token.
  topK: 40

  # Top-P controls how many likely tokens the model considers when generating text.
  # Lower values make output more focused and predictable, while higher values allow more randomness and creativity.
  topP: 0.85

  # The number of predictions (max output tokens) controls how much text the model is allowed to generate.
  # A lower limit keeps responses short and concise, while a higher limit allows longer, more detailed answers.
  numPredict: 400

# Conversation context configuration
context:
  # Maximum number of messages to store per player
  # A higher limit will allow players to continue conversations for longer,
  # however this could cause performance impacts with high player counts.
  maxPlayerMessages: 20

genius

Properties for the agent's appearance and behavior

genius.agentName

The name for the Genius agent to use in chat when responding to messages

• Type: string
• Default: Genius

ollama

Ollama client configuration

ollama.baseUrl

The base URL for the Ollama API. If using a cloud model, this should be set to https://ollama.com/. Defaults to the Ollama API's default listening location. For more info, see the hosting guide

• Type: string
• Default: http://localhost:11434/

ollama.apiKey

Your key for the Ollama cloud API. This only needs to be set if you are using an Ollama cloud model. For more info, see the hosting guide

• Type: string
• Default: (blank)

ollama.model

The name of the model to use for response generation. If you are using Ollama cloud, the model must be an ollama cloud model

• Type: string
• Default: deepseek-v3.1:671b

ollama.temperature

Controls how random or deterministic Genius's output is. Lower values make responses more predictable and precise, while higher values make them more creative and varied.

• Type: number
• Default: 0.5

ollama.topK

Top-K sampling limits the model’s next-token choices to the K most likely options, filtering out all others. This reduces randomness while still allowing some creativity compared to always picking the single most probable token.

• Type: integer
• Default: 40

ollama.topP

Top-P controls how many likely tokens the model considers when generating text. Lower values make output more focused and predictable, while higher values allow more randomness and creativity.

• Type: number
• Default: 0.85

ollama.numPredict

The number of predictions (max output tokens) controls how much text the model is allowed to generate. A lower limit keeps responses short and concise, while a higher limit allows longer, more detailed answers.

• Type: integer
• Default: 400

context

Conversation context configuration

context.maxPlayerMessages

Maximum number of messages to store per player. A higher limit will allow players to continue conversations for longer, however this could cause performance impacts with high player counts.

• Type: integer
• Default: 20