Skip to main content

OpenAI Connections

Lis Novel uses the OpenAI API for AI generation. You can create multiple named connections with separate API keys and model settings.

Set up a connection​

  1. Open LLM Providers from the app status menu and choose Add a provider. This creates an OpenAI connection directly.
  2. Enter your API Key and optionally your Organization ID.
  3. Leave Custom Base URL at its default for OpenAI, or enter a compatible endpoint.
  4. Enable the connection and click Test to fetch its model list.
  5. Choose a Default Model and enable the models you want available in chat and prompt pickers.

Custom endpoints must support model listing. Beat generation and rewrites require the OpenAI Responses API. Novel outlines and chat also support Chat Completions, selected from model defaults or a per-model protocol override. Outlines can switch protocols after an explicit unsupported-protocol rejection before output begins.

Manage connections and models​

Each connection can be named, reordered, enabled or disabled, tested, and removed independently. Streaming can be switched off per connection. Background checks refresh connection health and model catalogs.

Known model IDs receive bundled catalog defaults on any connection. Conflicting capabilities remain unknown, and unknown models use name-based hints. Use Model Traits to override capability hints for a model, especially when connecting to a custom endpoint. See Model Traits.

Prose draft token counts are estimates. Chat context usage comes from the chat runtime and may be shown as Unknown, including after context compaction. Context limits use model overrides, reported runtime limits, or bundled catalog estimates when available. See Chat history and context.

AI requests send the selected manuscript context, prompts, and attachments to the configured API endpoint.