Skip to content

Using the SkyCMS AI Assistant

Summary

Use this guide when you need to:

  • work with AI inside SkyCMS editors,
  • understand which AI tools are available in Monaco and CKEditor,
  • select a model when the configured provider allows it,
  • open the standalone AI Help Chat.

What the AI assistant can do

SkyCMS currently exposes AI in three places:

  • Code Editor (Monaco) for inline completions and code-focused chat,
  • Visual Editor (CKEditor) for region-scoped writing help,
  • AI Help Chat for broader product and site guidance.

All three experiences use the same tenant AI provider settings.

Code Editor AI

The Monaco code editor includes two separate AI behaviors.

Inline completions

As you type, SkyCMS can show ghost-text suggestions.

  • Press Tab to accept the suggestion.
  • Keep typing to dismiss it.
  • Inline completions are best for short markup, CSS, or JavaScript continuations.

AI chat panel

The code editor also includes an AI chat panel for longer requests such as:

  • explain this block,
  • fix syntax,
  • improve structure,
  • convert a selection,
  • suggest head metadata or layout changes.

The AI prompt is enriched with the current editor context, including the active field, document kind, and section kind.

Visual Editor AI

The CKEditor-based Visual Editor includes region-scoped writing assistance.

This is designed for editorial tasks such as:

  • improving tone or clarity,
  • rewriting a selected block,
  • summarizing or shortening copy,
  • generating polished HTML for the current editor region.

Important behavior:

  • the assistant is scoped to the active rich-text region, not the whole page,
  • applying a response may replace the current region or selection depending on the action you choose,
  • if the region changed after the AI response was generated, SkyCMS may ask you to regenerate before applying non-replace actions.

AI Help Chat

AI Help Chat is a broader assistant experience that is not tied to a single editor field.

You can open it from:

  • Editor -> AI Help Chat in the main editor menu,
  • the Detach link in AI chat panels when available.

Current help modes include:

  • SkyCMS help for product guidance,
  • Website guidance for practical content and website advice,
  • This site for site-aware questions using the current page context when available.

Detached window support

AI Help Chat can run in a detached window.

Use this when you want to:

  • keep the help chat open while editing,
  • compare answers across multiple content screens,
  • work with a larger conversation surface.

Model selection inside the editor

When the provider supports it, SkyCMS shows a Model picker inside the editor AI UI.

Current behavior:

  • GitHub Models and OpenAI can expose a live model list,
  • the selected model is saved for your account and editor context,
  • leaving the selector on Default uses the tenant default behavior.

Examples:

  • you may keep a fast model in Monaco for inline work,
  • and use a stronger model in AI Help Chat for longer guidance.

If the provider does not support user model selection, the picker stays disabled and the tenant default is used.

Refreshing available models

Use Refresh when:

  • the administrator changed the provider configuration,
  • the provider published additional models,
  • the picker looks stale.

SkyCMS combines cached results with explicit refresh requests, so a refresh is the safest way to verify the latest provider catalog.

Practical usage guidance

Best uses for AI in Monaco

  • small HTML scaffolds,
  • CSS fixes,
  • JavaScript cleanup,
  • head or meta suggestions,
  • explaining unfamiliar markup or scripts.

Best uses for AI in CKEditor

  • rewriting copy,
  • improving readability,
  • restructuring paragraphs,
  • turning rough notes into cleaner publish-ready copy.

Best uses for AI Help Chat

  • asking where content should live,
  • understanding SkyCMS workflows,
  • getting quick website guidance without opening an editor field.

Limitations to keep in mind

  • AI availability depends on the tenant provider configuration.
  • Discovery and model selection vary by provider.
  • Azure OpenAI usually uses an inferred deployment rather than a live model picker in the current SkyCMS implementation.
  • Direct native Anthropic endpoints are not a standard SkyCMS preset because the current AI proxy expects OpenAI-compatible chat completions.