Consider it done. The open-source AI agent that works out of the box · 想到,就能做到。开源、开箱即用的 AI Agent。
TypeScript agentai-agentai-assistantandroidclaude-codecodexelectroniosllmmacosreact-nativetypescriptwindows
/**
* subagentDurableStatusRenderItems.test.ts
* ---------------------------------------------------------------------------
* A running background Agent card must stay visible in the chat, not be folded
* into the finished work group, even when the launch receipt's wording is one
* the text matcher does not recognise (Claude Code 2.1.280 changed it once).
* The host's `subagent_runs` status is the structural source that decides.
*/
import { describe, expect, it } from 'vitest';
import { buildSubagentRunStatusIndex } from '@cindy/maker-shared/agent-task';
import { buildRenderItems, groupWorkRuns } from '../components/chat/MessageStream';
import type { AgentTaskUpdate, ChatMessage } from '@/lib/makerChatStore';
const at = (seconds: number): string =>
new Date(Date.UTC(2026, 8, 26, 12, 0, seconds)).toISOString();
const messages: ChatMessage[] = [
{ clientId: 'u1', role: 'user', content: '帮我调研一下', createdAt: at(0) },
{
clientId: 'agent-call',
role: 'tool_use',
content: '',
toolUseId: 'toolu_bg',
toolName: 'Agent',
toolInput: { description: 'background research', prompt: 'research' },
createdAt: at(1),
},
{
clientId: 'agent-receipt',
role: 'tool_result',
content: 'Background agent queued (a receipt wording nobody has matched yet)',
toolUseId: 'toolu_bg',
createdAt: at(2),
},
{
clientId: 'a1',
role: 'assistant',
content: '已经派出一个后台调研,完成后会通知我。',
createdAt: at(3),
turnCompleted: true,
},
];
const runningUpdate: AgentTaskUpdate = {
provider: 'claude-code',
taskId: 'agent-1',
parentToolUseId: 'toolu_bg',
status: 'running',
};
const taskUpdates = new Map<string, AgentTaskUpdate>([
['agent-1', runningUpdate],
['toolu_bg', runningUpdate],
]);
function topLevelAgentTask(items: ReturnType<typeof groupWorkRuns>) {
return items.find((item) => item.type === 'agent_task');
}
describe('buildRenderItems — durable Subagent status', () => {
it('keeps a running background Agent visible when subagent_runs says it is running', () => {
const { items } = buildRenderItems(messages, taskUpdates, undefined, {
subagentRunStatuses: buildSubagentRunStatusIndex([
{ parentToolUseId: 'toolu_bg', logicalAgentId: 'agent-1', status: 'running' },
]),
});
const card = topLevelAgentTask(groupWorkRuns(items, false));
expect(card).toMatchObject({ type: 'agent_task', durableStatus: 'running' });
});
it('without the durable record an unknown receipt still reads as finished work (fallback path)', () => {
const { items } = buildRenderItems(messages, taskUpdates);
expect(topLevelAgentTask(groupWorkRuns(items, false))).toBeUndefined();
});
});