Skip to content

LLM Model Ollama

187

The LLM Model node is the execution engine that sends requests to Large Language Models via REST API. This specific node is optimized for local Ollama instances.

Advanced Features: * Tool Support Fallback: If the selected model does not natively support "Function Calling" (Tools), ERGS Studio can automatically inject tool definitions into the system prompt. It then parses the AI's response for specific tags (e.g., [TOOL: tool_name]) to trigger logic events. * Game Cue Integration: When use Cues is enabled, the model automatically incorporates active game hints (Cues) into the prompt. Since Cues in ERGS are tied to player progress, the AI always knows what the player hasn't solved yet and can provide contextually relevant help. Once a task is completed, the Cue is deactivated and removed from the AI's context. * History Connection: This node requires a connection to an LLM Chat history node to maintain the conversation flow.

Technical Note: Use the processing boolean output to drive "Thinking..." animations or UI indicators while waiting for the API response.

Connectors in

  • process [type: action]
  • history [type: pointer]

Connectors out

  • onFinish [type: action]
  • onError [type: action]
  • response [type: string]
  • tool [type: string]
  • error [type: string]
  • processing [type: bool]

Node properties

  • API key [type: string]
  • info [type: string]
  • model_name [type: string]
  • tools supported [type: bool]
  • URL [type: string]
  • use Cues [type: bool]