agenthub/tests/work.test.ts
chahinebrini 9551c90597 fix(realtime): discovery self-heal, polling fallback, message ack semantics, CRLF SSE parser
- probe configured server URL (~1s) before remote commands; on failure
  re-discover via LAN discovery, persist and use the resolved URL
- poll tryClaim/trySurfaceMessages every ~4s during the SSE wait in CLI
  work and MCP waitForTask — a lost SSE frame costs seconds, never sleep
- work/MCP no longer auto-mark messages read: surfacing sets delivered,
  explicit ack/read required; loops still wake only on unread (no spin)
- parseSSEBuffer accepts CRLF frame separators and joins multi data: lines
- bump version to 0.10.1 (package.json, CLI --version, MCP server)
2026-07-23 16:53:33 +02:00

186 lines
6.9 KiB
TypeScript

/**
* Tests for `agenthub work` — the auto-claim primitive (TSK-0018).
* - immediate: claims an already-open addressed task without waiting (local).
* - wait: blocks on SSE until a matching task is created, then claims it.
*/
import { describe, it, expect, beforeEach, afterEach } from 'vitest';
import { mkdtempSync, rmSync } from 'fs';
import { tmpdir } from 'os';
import { join } from 'path';
import { init } from '../src/cli/commands/init.js';
import { workAgent } from '../src/cli/commands/work.js';
import { startServer } from '../src/server/index.js';
import { createTask, getTask } from '../src/core/services/taskService.js';
import type { Task } from '../src/core/schema.js';
describe('agenthub work — immediate claim (local)', () => {
let cwd: string;
beforeEach(() => {
cwd = mkdtempSync(join(tmpdir(), 'ah-work-'));
init(cwd, { projectName: 'work-test', yes: true });
});
afterEach(() => {
rmSync(cwd, { recursive: true, force: true });
});
it('claims an already-open addressed task without waiting', async () => {
const mine = createTask(cwd, { title: 'kimi: already open', role: 'implementer' });
await workAgent({ projectCwd: cwd, agent: 'kimi', role: 'implementer', timeoutSec: 2 });
const { task } = getTask(cwd, mine.id);
expect(task.status).toBe('in_progress');
expect(task.assignedTo).toBe('kimi');
}, 4000);
});
describe('agenthub work — wait then auto-claim (server SSE)', () => {
let cwd: string;
let server: Awaited<ReturnType<typeof startServer>>;
let nextServer: Awaited<ReturnType<typeof startServer>> | undefined;
beforeEach(async () => {
cwd = mkdtempSync(join(tmpdir(), 'ah-work-wait-'));
init(cwd, { projectName: 'work-wait', yes: true });
server = await startServer(cwd, { host: '127.0.0.1', port: 0 });
});
afterEach(async () => {
await server.app.close().catch(() => undefined);
await nextServer?.app.close().catch(() => undefined);
rmSync(cwd, { recursive: true, force: true });
});
it('claims a task created AFTER work starts waiting', async () => {
// Start waiting; do NOT await yet — no task is addressed to kimi initially.
const workDone = workAgent({
serverUrl: server.url,
projectCwd: cwd,
agent: 'kimi',
role: 'implementer',
timeoutSec: 3,
});
// Give work a moment to reach the SSE wait, then delegate a task to kimi.
await new Promise((r) => setTimeout(r, 200));
await fetch(`${server.url}/tasks`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ title: 'kimi: delegated after wait', role: 'implementer' }),
});
await workDone; // resolves once work has claimed the task (or on timeout)
const tasks = (await fetch(`${server.url}/tasks`).then((r) => r.json())) as Task[];
const mine = tasks.find((t) => (t.title ?? '').startsWith('kimi:'));
expect(mine).toBeDefined();
expect(mine?.status).toBe('in_progress');
expect(mine?.assignedTo).toBe('kimi');
}, 6000);
it('reconnects after the SSE server moves and claims a later task', async () => {
const workDone = workAgent({
serverUrl: server.url,
projectCwd: cwd,
agent: 'kimi',
role: 'implementer',
timeoutSec: 6,
reconnectBackoffMs: [50, 100],
discoverServer: async () => nextServer?.url,
});
await new Promise((r) => setTimeout(r, 200));
server.app.server.closeAllConnections?.();
await server.app.close();
nextServer = await startServer(cwd, { host: '127.0.0.1', port: 0 });
await fetch(`${nextServer.url}/tasks`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ title: 'kimi: delegated after reconnect', role: 'implementer' }),
});
await workDone;
const tasks = (await fetch(`${nextServer.url}/tasks`).then((r) => r.json())) as Task[];
const mine = tasks.find((t) => (t.title ?? '').startsWith('kimi: delegated after reconnect'));
expect(mine).toBeDefined();
expect(mine?.status).toBe('in_progress');
expect(mine?.assignedTo).toBe('kimi');
}, 8000);
// ── TSK-0119: after a review submit the re-armed work loop must stay reachable ──
it('wakes on an architect follow-up message instead of going dormant', async () => {
// The implementer re-armed `work` after submitting; no task is addressed yet.
const workDone = workAgent({
serverUrl: server.url,
projectCwd: cwd,
agent: 'kimi',
role: 'implementer',
timeoutSec: 4,
});
await new Promise((r) => setTimeout(r, 200));
await fetch(`${server.url}/messages`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ from: 'claude', to: 'kimi', text: 'quick question about your submission' }),
});
await workDone; // resolves because the message woke the loop (not on timeout)
// The loop drained + surfaced the message (unread → delivered) — proving it
// woke on the message, not that it merely timed out (a timeout would leave
// it unread). Since TSK-0224 the loop no longer auto-marks read: the message
// stays visible in the inbox until an explicit ack/read.
const inbox = (await fetch(`${server.url}/messages?agent=kimi`).then((r) => r.json())) as Array<{ text: string; status: string }>;
const m = inbox.find((x) => x.text.includes('quick question'));
expect(m).toBeDefined();
expect(m?.status).toBe('delivered');
}, 7000);
it('wakes and re-claims when a submitted task is reopened', async () => {
// Seed a task addressed to kimi, claim it, submit for review.
await fetch(`${server.url}/tasks`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ title: 'kimi: build the widget', role: 'implementer' }),
});
await fetch(`${server.url}/tasks/TSK-0001`, {
method: 'PATCH',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ status: 'in_progress', assignedTo: 'kimi' }),
});
await fetch(`${server.url}/tasks/TSK-0001`, {
method: 'PATCH',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ status: 'review' }),
});
// Re-armed work loop: the task is in review, so nothing is addressed/open yet.
const workDone = workAgent({
serverUrl: server.url,
projectCwd: cwd,
agent: 'kimi',
role: 'implementer',
timeoutSec: 4,
});
await new Promise((r) => setTimeout(r, 200));
// Architect reopens (send back to implementer).
await fetch(`${server.url}/tasks/TSK-0001`, {
method: 'PATCH',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ status: 'open' }),
});
await workDone;
const { task } = getTask(cwd, 'TSK-0001');
expect(task.status).toBe('in_progress');
expect(task.assignedTo).toBe('kimi');
}, 7000);
});