Skip to main content
Go to the Studio
Artificial intelligence

Ask AI

Ask AI

Asks questions to a generative artificial intelligence model based on a context provided and returns the answer generated by the model. Ideal for summarizing texts, classifying information, extracting structured data or any task that benefits from natural language reasoning.

caution

This action consumes AI credits of the Roberty account. Make sure your account has balance available before using it.


Options​

Context​

Context information that will be passed to the model to guide the answer. It can be a fixed text describing the role of the assistant, data extracted from the flow or any information relevant to the task, for example You are a customer support assistant of the company XYZ..

Prompt​

The question or instruction that will be sent to the model, for example Summarize the following e-mail in one sentence..

Timeout​

Maximum time in milliseconds to wait for the model answer. Use 0 to wait indefinitely.

Advanced settings​

Clean code wrapper​

Automatically removes the code block delimiters (```) from the answer generated by the model. Useful when the model returns code wrapped in Markdown markers and you only want the plain content. Enabled by default.

Format JSON automatically​

Tries to automatically convert the answer into a JSON object when the model returns a text in the JSON format. Disabled by default.

Returns​

Javascript code​

js
actions["action-id"].response // answer generated by the AI model (Text)

Field selection​

  • response — answer generated by the AI model (Text)

Rules and Conditions​

  • The Context and Prompt fields are required.
  • The Timeout field with the value 0 means the action will wait indefinitely until it receives an answer.
  • Clean code wrapper is enabled by default. Disable it if you need to preserve the Markdown delimiters in the answer.
  • Format JSON automatically is disabled by default. Enable it only when the prompt instructs the model to return JSON — otherwise, the action may fail when it tries to interpret a free-text answer as JSON.
  • The quality of the answer depends on the clarity of the context and the prompt provided.
  • If the model does not answer within the timeout configured, the action ends with an error.