Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import { describe, expect, it } from 'vitest'
import type { Attachment } from '../../core/types'
import { ResultStore } from '../../resultStore'
import { entryHasFileAttachment, markFileAttachmentsDelivered } from '../fileAttachmentDelivery'
interface TestEntry {
attachments?: Attachment[]
storedAt: number
}
const fileAttachment: Attachment = {
id: 'file-1',
kind: 'file',
mimeType: 'application/pdf',
encoding: 'base64',
dataBase64: 'redacted-test-file-data',
filename: 'report.pdf',
sourceTool: 'workflow_result',
}
const imageAttachment: Attachment = {
id: 'image-1',
kind: 'image',
mimeType: 'image/jpeg',
encoding: 'base64',
dataBase64: 'redacted-test-image-data',
filename: 'preview.jpg',
sourceTool: 'image_generation',
}
function createStore(): ResultStore<TestEntry> {
return new ResultStore<TestEntry>(10_000, entry => entry.storedAt)
}
describe('file attachment delivery state', () => {
it('detects file attachments', () => {
expect(entryHasFileAttachment({ attachments: [imageAttachment] })).toBe(false)
expect(entryHasFileAttachment({ attachments: [fileAttachment] })).toBe(true)
})
it('removes file attachments from pending state after first delivery', () => {
const store = createStore()
const entry: TestEntry = {
attachments: [fileAttachment, imageAttachment],
storedAt: Date.now(),
}
store.set('task-1', entry)
markFileAttachmentsDelivered(store, 'task-1', entry)
expect(store.get('task-1')?.attachments).toEqual([imageAttachment])
})
it('removes the attachment field when only file attachments were pending', () => {
const store = createStore()
const entry: TestEntry = {
attachments: [fileAttachment],
storedAt: Date.now(),
}
store.set('task-1', entry)
markFileAttachmentsDelivered(store, 'task-1', entry)
expect(store.get('task-1')?.attachments).toBeUndefined()
})
})