Skip to content
Ragtime Developers

Actions

An action is something that happens in or around a conversation. It can fetch an order status, ask a visitor for their email, tell your website to open a page, or save a summary once the chat is over.

Every action is a short graph of steps. The steps say what it does. One special step, the Start, says when it runs.

Actions are managed in Dashboard → Actions, which has three tabs: Actions, Stages and Page events. They can also be built from code or by an AI tool, on the draft, through the Actions API.

When an action runs

Section titled “When an action runs”
Start says The action runs Typical use
When the AI decides The assistant calls it as a tool, using the description you wrote. Look up an order, book a slot, request a callback.
When the chat starts Automatically, as the conversation begins. Load context, tell the page the chat opened.
Right after the greeting Automatically. Server steps run with the chat-start ones, but the visitor sees the first question only after the greeting has been shown and spoken. A form or choice that the greeting leads into.
When the end is detected Once, as Ragtime notices the session is winding down. Wrap up before the session closes.
When the session has ended Once, after the session is over. Log the visit, send a summary to another system.
When another action runs it Only when a Run another action step calls it. A sequence you reuse in several places.

Chat-start and after-greeting actions also run in voice and avatar conversations, and page events wait for the greeting with them.

What an action can do

Section titled “What an action can do”

Steps are the building blocks. You can combine them freely:

  • Ask the visitor with a message, a choice, or a short form.
  • Call another system with an HTTP request and use the response later.
  • Tell your website something happened, with a page event.
  • Branch on an answer, an input, or a response.
  • Run another action and carry on when it finishes.
  • End by returning to open chat, ending the session, or handing a value back to the AI.

Each step is described in Steps.

Values that steps send or compare are always linked to something earlier: an answer, an input the AI supplied, a response, or a fixed value. Nothing is guessed while the action runs. See Start inputs and values.

Create your first action

Section titled “Create your first action”
  1. Open Dashboard → Actions and start a new action.
  2. Choose a template, or describe what you need and let Ragtime draft it.
  3. Fill in the details the template leaves open, such as a URL or an event name.
  4. Use Try it to run it before the assistant does.
  5. Switch it on.

The templates are:

  • Let the assistant look something up. The assistant fetches an answer from another system and uses it in its reply.
  • Collect details. Ask for a name and email in a short form, then carry on chatting.
  • Tell the website something happened. Send an event to the page around the assistant.
  • Save something when the chat ends. Send details to another system once the conversation is over.
  • Start from scratch. An empty action to build step by step.

Describe what you need generates a draft from a plain description. Check the draft before you switch it on, as you would with any action.

Keeping track of your actions

Section titled “Keeping track of your actions”

The action list shows when each action runs, whether it asks the visitor anything, and what uses it: other actions that run it, conversation stages that offer it, and workflows that start on it. You can switch an action on or off, or delete it, from the same list. Editing actions needs the Edit Actions permission.

Try it runs the server part of a saved action with inputs you type in. It shows every HTTP call with what was sent and what came back, every page event, and how the run ended: it finished (and what it returns to the AI), it stopped, or it reached a step that needs the visitor.

Where to go next

Section titled “Where to go next”