How do I use Leiolai in Cursor?
In Cursor’s model settings, enter your Leiolai API key, override the OpenAI base URL with https://api.leiolai.com/v1, and add leiolai-1 as a custom model. Check the selected model before starting a conversation, because the base-URL override is a shared setting.
Where do I configure Leiolai in Cursor?
Open Settings, then Models. Enter the Leiolai key in the OpenAI API-key setting, enable the base-URL override, and use https://api.leiolai.com/v1. Add leiolai-1 as the custom model and select that model for the conversation.
Which key should I use for confidential code?
Use a Leiolai Private-only key. The key permits Private requests without requiring a separate mode field. An unrestricted key needs mode on every request, so the client must support adding that field.
Create a key for the clientWhy should I check the base-URL override when switching models?
Cursor’s override is shared rather than attached only to the Leiolai model. Check the override when moving to another provider, and turn the override off when returning to that provider’s normal connection.
Does this setting replace Cursor Tab completion?
No. Cursor Tab uses Cursor’s own models. The custom API key and base URL connect Leiolai for supported chat conversations, not Tab completion.
How do I check that the model connection works?
Start a small chat with leiolai-1 selected. Ask a simple question using non-sensitive sample text, then inspect the result or returned error. Confirm the key and base URL before expanding the task to larger project context.
What should I do if an agent request is rejected?
Read Cursor’s returned Leiolai error code and inspect the named field. Compare the request with Leiolai’s supported Chat Completions fields, tool definitions, and key-mode restriction. Correct the unsupported or conflicting option, then test the request again.
Supported Chat Completions fieldsHow does thinking effort relate to Cursor’s model setting?
The Leiolai model identifier remains leiolai-1. Effort is a separate field, reasoning_effort for Chat Completions. A client that does not send that field uses the default low effort.
Set API thinking effortHow can I get a useful answer about a code change?
Give Leiolai in Cursor the behavior you want, relevant code, and constraints on the change. Ask for a small change with a clear verification step. Review the resulting diff and run the project’s appropriate checks.
Frame a coding question