Skip to main content
  1. Be specific in your requests as details will help the model understand what you need best.
  2. Add context to the context panel or reference scripts specifically by name to have the agent look for them.
  3. Minimize requests to optimize results
    1. Working in one step at a time is better than asking for the entirety of the game at once. 
    2. Moving from system to system is more optimal for the AI Agent overall.
  4. Choose the right model for your development needs. Newer models are more thorough.
  5. Use the appropriate Agent mode to optimize results depending on the scenario.
  6. Use the appropriate processing mode to optimize results
    1. Fast uses the least amount of context possible. Great for quick requests cheap in token cost that do not need heavy scene context. 
    2. Default uses a moderate amount of context in each request and does not include full scene hierarchy. Not good for debugging.
    3. Thinking uses the full amount of context available. Great for debugging and all requests but consumes more tokens each message.