AI

Table of contents

The AI API provides functions for sending requests to Large Language Models (LLMs) from your scripts. Use it to generate text, analyze data, or work with attached images directly within Memento Database.

AI requests must be executed asynchronously in the last Phase of an Event

Each AI request consumes account credits. For simple tasks, use light models like gpt-4o-mini or gpt-4.1-nano. Write scripts efficiently to manage usage costs

Global functions

ai()

Creates and returns an AI object used to send requests to an AI model.

Returns

A new AI object instance for making AI requests

AI object

The AI object provides methods for making AI requests within your app scripts. Access it using the global ai() function.

model(model_name)

Sets the AI model to be used for requests.

Parameters

Parameter Type Description
model_name String The name of the AI mode

Supported models:

  • gpt-4o-mini
  • gpt-4.1
  • gpt-4.1-mini
  • gpt-4.1-nano
  • claude-sonnet-4-0
  • claude-3-5-haiku-latest

Returns

The AI object.

system( message )

Sets a system instruction message for the model. System messages define the model’s behavior, style, or role.

Parameters

Parameter Type Description
message String string containing the system instruction.

Returns

The AI object.

ask(message , attachments)

Sends a message to the AI model.

Parameters

Parameter Type Description
message String string, the user message
attachments (optional) String or Array one or more image file paths

Returns

An AIResult object containing the model’s response. AIResult can be used directly as a string.

AIResult Object

AIResult objects represent the model’s response and can be used as strings.

asJson()

Converts the model response into a JavaScript object (parsed from JSON). Use this when expecting structured data.

Returns

A JavaScript object.

AI credits

Each request consumes AI credits from your account:

  • 1 credit per 1000 tokens for lightweight models (e.g. gpt-4o-mini, gpt-4.1-nano)
  • 5 credits per 1000 tokens for premium models (e.g. gpt-4.1, claude-sonnet-4-0)

Examples

Simple text request 

var response = ai()
    .model("gpt-4o-mini")
    .ask("Write a haiku about databases.");
message(response);

Using system message 

var response = ai()
    .model("gpt-4.1")
    .system("You are a professional assistant specialized in finance.")
    .ask("Explain compound interest in simple terms.");
message(response);

Passing images as attachments 

var img = entry().images("Photo");
var response = ai()
    .ask("Describe this image.", [img]);
message(response);

Getting structured JSON output 

var response = ai()
    .system("Return your answer strictly as JSON.")
    .ask("Give me a shopping list of 3 items.");
var json = response.asJson();
message(json); // first item in the list

Collect context from entries 

var entries = lib().entries();
var context = "";
for (var i = 0; i < entries.length; i++) {
    var e = entries[i];
    context += "- " + e.field("Notes") + "\n";
}

// Ask the AI with context included
var response = ai()
    .model("gpt-4.1")
    .system("You are an assistant that answers questions using the provided context.")
    .ask("Context:\n" + context + "\n\nQuestion: What are the main themes across these notes?");

message(response);

Passing entry values to AI 

var data = entry().values();
var response = ai()
    .model("gpt-4o-mini")
    .system("You are a data assistant. Summarize the entry in plain English.")
    .ask("Here is the entry data:\n" + JSON.stringify(data, null, 2));
message(response);