Connect a model for your Palworld server tasks

You choose the model. GeekCrest connects to a cloud AI provider through a standard API endpoint, to a model running locally on your PC, or to the built-in model, and the Test button checks the connection before you give it a task.

Three ways to connect a model

A cloud AI provider

Pick this when a task is hard enough to need a large model and you accept that the prompt goes to the provider. The key stays local. The prompt text, including file text the agent reads, is sent to that provider, and the provider may charge separately for it.

A model on your PC

Choose a model running locally on your PC when the files you work with should not leave the machine. Then nothing leaves the PC. Its quality and context size depend on the model and your hardware, so run the Test button before you rely on it.

The built-in model

The built-in model is the starting point if you have not connected anything yet. Begin with one folder. Ask it for a small task, such as listing the folders of your server, and note that prompts go only to the model you picked, as with every option; the privacy page explains the details.

What the Test button shows

The Test button checks an endpoint before you spend a session on it. It shows the response time and the context window, which is how much text the model can take in at once. A slow answer or a small window is a reason to choose another model for long logs. Read the result yourself.

Matching the model to the task

Use a bigger model for refactors and complex logic, such as reading a long log together with your settings file, and a smaller one for small edits like changing a single value. run_on_model sends one hard step to a bigger model while the rest stays on your main model. Decide per task.

No model is always right. Review the diff before you accept it, and use Ask first so that every change to a file waits for your OK before anything is written to your server folder.