Skip to main content
A prompt is the fundamental building block in Prompt Bank. It captures the instruction you send to an AI model — the text description, configuration, and any reference media — so you can store it, reuse it, and trigger generations from it at any time. Rather than rewriting the same prompt every time you call a model, you save it once and reference it by ID across your workflows.

Prompt types

Every prompt has a promptType that describes the kind of output it is intended to produce. Choosing the correct type keeps your vault organized and lets the API surface the right generation options when you use a prompt to queue a job.

Key fields

Prompt object

When you read a prompt from the API, the returned object includes the following fields:

Prompt lifecycle

Working with a prompt follows a predictable three-step pattern:
  1. Create — Send a POST /api/v1/prompts request with your instruction text, type, and any model configuration. The API returns a prompt object with a unique id.
  2. Save to a vault — Prompts are saved into a vault automatically when you supply a vaultId, or land in your General vault by default. You can move them between vaults at any time.
  3. Use in a generation — Reference the prompt’s text and configuration when calling POST /api/v1/generations. The generation runs asynchronously and saves its output back to a vault.

Example prompt object

The following shows a complete image prompt as returned by the API:

Next steps

See the Manage Prompts guide for step-by-step instructions on creating, updating, filtering, and deleting prompts via the API.