LogoOctarine Docs
Working with AI

Writing Assistant

Create, rewrite or improve your ideas with the help of an assistant

The Writing Assistant in Octarine is an AI-powered tool designed to help you generate, rewrite, or refine content directly within your notes. This feature is available exclusively to Pro License users.

Getting Started

Before using the Writing Assistant, ensure that at least one AI provider or Ollama is configured. For setup instructions, refer to Configuring AI.

Accessing the Writing Assistant

  • Sidebar Access: Open the Writing Assistant sidebar using Cmd/Ctrl + J.
  • Context Selection: By default, the current note is used as the context. To change the context, use the context selector or press @.
  • Partial Text Context: Select any portion of text and press Cmd/Ctrl + J or click the Add to chat button in the Bubble Menu. The selected text will be shown as "x ch," where x is the number of characters selected.

Chat Context and Context Changes

  • When the context selection or file changes within the same chat session, a "Context Changed" indicator appears. This ensures previous, now-irrelevant AI responses are discarded, maintaining the accuracy of interactions.
  • Each chat session takes previous exchanges into account for consistent and context-aware assistance.

Responding to AI Suggestions

After you receive an AI-generated response, several quick actions are available:

  • Copy: Copy the response to your clipboard.
  • Insert: Insert the response at your current cursor position.
  • Replace: Replace the originally selected text (if any) with the AI-generated reply.
  • Retry: Request a new version of the response.
  • Delete: Remove unwanted responses from the chat.

Slash Commands

  • Customization: You can create and manage custom slash commands for frequent prompts. Configure these in Settings → AI Assistant → Slash Commands.
  • Storage: Custom slash commands are saved in .octarine/ai/slash-commands.json within your workspace.

Recommendations

For optimal performance, select a lightweight AI model not specifically designed for code. This choice ensures efficient assistance while conserving resources.