Configure an AI Model

Add a model

1. Open Model Settings.
2. Choose an existing provider or add a compatible one.
3. Enter the service URL, model name, and API key.
4. Enable thinking mode only when the model supports it.
5. Save and run Check Model.

A successful check covers only a short request. Also verify Chinese output, structured output, completion of thinking mode, a suitable timeout, and the exact model name. Compatible DeepSeek models may take several minutes for long chapters and multi-stage review; avoid clicking Send or Write repeatedly. After a reset, empty response, or timeout, reload the conversation or retry the failed task so completed stages can resume from checkpoints.

Changing the chat model affects the next task, not a task already running. Maintain keys only in Model Settings, and exclude them from screenshots, logs, and exports.