AI Assistant

The AI Assistant in Topol is an integrated writing assistant powered by OpenAI's ChatGPT. It works on one block at a time, taking the text from the selected block and returning a rewritten version of it.

For an example of its use, check out this article.

AI Assistant or AI Chat?

The AI Assistant rewrites the block that is currently selected. AI Chat is conversational, receives the whole template as context, and can suggest changes across several blocks at once. The comparison below covers when each one fits.

How to Enable the AI Assistant

The AI Assistant feature is enabled by default. However, to connect it with ChatGPT, which is essential for its functionality, you must first configure it. Instructions are available in this article.

To disable it, add the disableAiAssistant setting to your TOPOL_OPTIONS object:

disableAiAssistant: true, // default is false

Where It Appears

The assistant sits behind the "stars" icon in the toolbar of text and button blocks. Other block types do not show it.

Text blocks open the full menu of actions. Button blocks show only the Translate option, since the remaining actions are meant for longer copy.

Action	What it does
Improve	Rewrites the text to read better
Rephrase	Restates it while keeping the formatting close to the original
Make longer	Expands the text slightly
Make shorter	Condenses the text slightly
Check grammar	Fixes spelling, grammar and punctuation without rewriting
Translate	Translates the block into a chosen language. See Template Translation.

Each action replaces the content of the selected block. The prompts are localized, so an editor running in another language sends prompts in that language.

Choosing Between the Two

Both features call OpenAI, and the practical difference is what each one sees and what it changes:

• AI Assistant sends the text of one block and returns replacement text for that block.
• AI Chat sends the whole template as context and can return suggestions for several blocks at once, which the user reviews and applies.

Polishing a headline or fixing a typo is faster through the assistant. Questions about the email as a whole, or edits spanning multiple sections, are what AI Chat is built for.

Monitoring Usage

Total token consumption is available from the AI Assistant Logs, which covers both the AI Assistant and AI Chat.

Using Your Own Model

Requests from the AI Assistant can be routed to your own endpoint instead of going through Topol. See Custom AI Model for the configuration and the request format.