Shape

interface Input {
  text: string
  attachments: Attachment[]
}

interface Attachment {
  id: string
  name: string
  size: number
  modality: 'document' | 'image' | 'audio' | 'video'
  mimeType?: string
}

Fields

text

The textual content of the user message. If the user sends attachments without text, this will be an empty string.

attachments

A normalized list of uploaded assets associated with the current user turn.

Attachment Fields

• id – asset identifier used to reference the attachment in AfferLab APIs
• name – original file name
• size – file size in bytes
• modality – the attachment modality (document, image, audio, video)
• mimeType – MIME type

Usage

Strategies can pass the full user input directly into the prompt:

ctx.slots.add('input', ctx.input)

This forwards both the message text and its attachments.

Strategies may also access specific fields when needed:

const text = ctx.input.text
const files = ctx.input.attachments

Notes

ctx.input is read-only from the strategy point of view. If a strategy wants to ingest new material into memory, that happens through ctx.memory.ingest(...), not by mutating ctx.input.