Orca is the place for working with a fleet of agents (e.g. Grok Bot, Claude Code). Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.
TypeScript adeagent-ideai-agentsautomationbotclaude-codecodexghosttygrok-bothermes-agentmobile-appmuseopenclawopencodeorchestrationpiruntimeterminalworktreesyc-backed
import { describe, expect, it } from 'vitest'
import {
nativeChatReasoningHeadline,
nativeChatReasoningHeadlineText,
selectNativeChatLiveReasoning
} from './native-chat-reasoning-row'
import type { NativeChatMessage } from '../../native-chat-types'
function row(
id: string,
role: NativeChatMessage['role'],
text: string,
fields: Partial<NativeChatMessage> = {}
): NativeChatMessage {
return {
id,
role,
blocks: [{ type: 'text', text }],
timestamp: 1_000,
source: 'transcript',
...fields
}
}
const prompt = row('user-1', 'user', 'Start the task')
const open = row('r-1', 'reasoning', 'Weighing two approaches', { state: 'running' })
const inTurn = (): boolean => true
describe('the open block the live line discloses', () => {
const select = (messages: NativeChatMessage[], live: (index: number) => boolean = inTurn) =>
selectNativeChatLiveReasoning(messages, live)?.message ?? null
it('is the newest root reasoning row with text, with the text it has so far', () => {
expect(selectNativeChatLiveReasoning([prompt, open], inTurn)).toEqual({
message: open,
markdown: 'Weighing two approaches'
})
// A host that keeps no lifecycle gets the same single live slot.
const stateless = row('r-1', 'reasoning', 'Weighing two approaches')
expect(select([prompt, stateless])).toBe(stateless)
})
it('is nothing when the block is blank or ended', () => {
expect(select([prompt, row('r-1', 'reasoning', ' \n', { state: 'running' })])).toBeNull()
expect(select([prompt, { ...open, state: 'completed' }])).toBeNull()
})
it('is nothing once a tool or the answer is newer than the block', () => {
expect(select([prompt, open, row('a-1', 'assistant', 'Here it is')])).toBeNull()
const tool = row('t-1', 'assistant', '', {
blocks: [{ type: 'tool-call', name: 'Read', input: { file_path: 'a.ts' }, state: 'running' }]
})
expect(select([prompt, open, tool])).toBeNull()
})
it('looks past notices, empty rows and rows outside the live working turn', () => {
const notice = row('s-1', 'system', 'Compacting')
const empty = row('a-0', 'assistant', '')
const waiting = row('user-2', 'user', 'Also say banana')
expect(select([prompt, open, notice, empty, waiting], (index) => index < 4)).toBe(open)
})
it('stops at the live turn prompt and ignores a subagent reasoning', () => {
expect(select([open, prompt])).toBeNull()
const child = row('r-2', 'reasoning', 'Child thinking', { state: 'running', agentId: 'sub-1' })
expect(select([prompt, child])).toBeNull()
expect(select([prompt, open, child])).toBe(open)
})
})
describe('the reasoning headline', () => {
const text = (
fields: { state?: 'running' | 'completed'; completedAt?: number },
live = false
): string =>
nativeChatReasoningHeadlineText(
nativeChatReasoningHeadline({ timestamp: 1_000, ...fields }, { live })
)
it('reads the span the host saw, at least a second, and claims none it did not see', () => {
expect(text({ state: 'completed', completedAt: 66_000 })).toBe('Thought for 1m 5s')
expect(text({ state: 'completed', completedAt: 1_300 })).toBe('Thought for 1s')
expect(text({ state: 'completed' })).toBe('Thought')
expect(text({ state: 'running' })).toBe('Thought')
expect(text({})).toBe('Reasoning')
})
it('claims no past tense for a block not yet ended inside its working turn', () => {
expect(text({ state: 'running' }, true)).toBe('Reasoning')
expect(text({}, true)).toBe('Reasoning')
expect(text({ state: 'completed', completedAt: 13_000 }, true)).toBe('Thought for 12s')
})
})