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 { ChatMessage } from '../../core/types'
import { projectMessage } from '../imageSource'
// An empty contentParts array is "no parts": the turn stays V1 byte-identical.
describe('projectMessage with an empty contentParts array', () => {
it('keeps a user turn byte-identical to V1 and keeps its content', () => {
const message = { role: 'user', content: 'hello', contentParts: [] } as ChatMessage
expect(JSON.stringify(projectMessage(message, 'codex_request_invalid'))).toBe(
JSON.stringify({ role: 'user', content: 'hello' })
)
})
it('does not refuse an assistant turn that carries no parts', () => {
const message = { role: 'assistant', content: 'done', contentParts: [] } as ChatMessage
expect(projectMessage(message, 'codex_request_invalid')).toEqual({
role: 'assistant',
content: 'done',
})
})
})