"""cloud worker 任务间互发消息单测：agent 标记块 → 投递 → 接收方排空前缀 / 终态复活。 不跑真实模型、不联网：注入假 runner（假 CLI），任务与 目标都在临时 SQLite 里；env 里的值只是占位用的假名，绝 不是真实密钥。 """ import json import os import sqlite3 import sys import tempfile import time import unittest from pathlib import Path from types import SimpleNamespace try: from cloud.server import init_db from cloud.worker import Worker except ImportError: # discover -s cloud/tests 时把项目根目录补进 sys.path sys.path.insert(0, str(Path(__file__).resolve().parents[2])) from cloud.server import init_db from cloud.worker import Worker MARKER_TO_T2 = ("结果如下。\n```task-message\n" "{\"to\": \"T2\", \"body\": \"请评审这份结论\"}\n```\n") PREFIX = "[来自任务 T1 的消息]\n" def _insert_task(db_path, task_id, created_by="tester", status="queued", project_id="p1"): now = time.time() conn = sqlite3.connect(db_path) try: conn.execute( "INSERT INTO tasks(id, project_id, prompt, status," " idempotency_key, depends_on, priority, created_by," " created_at, updated_at, last_heartbeat)" " VALUES(?, ?, ?, ?, NULL, '[]', 0, ?, ?, ?, NULL)", (task_id, project_id, "do work", status, created_by, now, now), ) conn.commit() finally: conn.close() def _status(db_path, task_id): conn = sqlite3.connect(db_path) try: row = conn.execute( "SELECT status FROM tasks WHERE id = ?", (task_id,)).fetchone() finally: conn.close() return row[0] if row else None def _resume_message(db_path, task_id): conn = sqlite3.connect(db_path) try: row = conn.execute( "SELECT resume_message FROM tasks WHERE id = ?", (task_id,)).fetchone() finally: conn.close() return row[0] if row else None def _followups(db_path, task_id=None): conn = sqlite3.connect(db_path) try: if task_id is None: return conn.execute("SELECT * FROM task_followups").fetchall() return conn.execute( "SELECT * FROM task_followups WHERE task_id = ?", (task_id,)).fetchall() finally: conn.close() def _prompt_values(calls): """假 runner 收到的全部 prompt：dispatch 在 --task 后， run 在 --prompt 后，两个都扫。""" values = [] for cmd in calls: for flag in ("--task", "--prompt"): if flag in cmd: values.append(cmd[cmd.index(flag) + 1]) return values def _marker_runner(calls, marker_text): """第一轮输出含标记块，之后一律 plain done。""" def fake(cmd): calls.append(list(cmd)) text = marker_text if len(calls) == 1 else "done" return SimpleNamespace( returncode=0, stdout=json.dumps({"status": "completed", "text": text}), stderr="") return fake class WorkerPeerTestCase(unittest.TestCase): def setUp(self): self.tmp = tempfile.TemporaryDirectory() self.db_path = os.path.join(self.tmp.name, "control.db") init_db(self.db_path) self.env = {"XCM_API_KEY": "test-placeholder"} self.calls = [] def tearDown(self): self.tmp.cleanup() def _worker(self, runner): return Worker(self.db_path, "/fake/host.py", runner=runner, env=self.env, model="gpt-6.1-sol", retry_backoff_seconds=0) def test_marker_delivered_and_drain_prefixed(self): _insert_task(self.db_path, "T1") _insert_task(self.db_path, "T2") worker = self._worker(_marker_runner(self.calls, MARKER_TO_T2)) result = worker.run_one(task=worker.get_task("T1")) self.assertEqual(result["status"], "completed") # T1 的标记块投递成 T2 的后续消息（发送方 id 在列里）。 fus = _followups(self.db_path, "T2") self.assertEqual(len(fus), 1) row = fus[0] # 列序与 DDL 一致：id, task_id, project_id, body, mode, # status, created_by, sender_task_id, created_at, updated_at。 self.assertEqual(row[3], "请评审这份结论") self.assertEqual(row[5], "pending") self.assertEqual(row[7], "T1") # T2 跑起来：首轮原 prompt，排空轮次 prompt 带发送方前缀。 result = worker.run_one(task=worker.get_task("T2")) self.assertEqual(result["status"], "completed") prompts = _prompt_values(self.calls) self.assertTrue( any(PREFIX in p and "请评审这份结论" in p for p in prompts), prompts) self.assertEqual(_followups(self.db_path, "T2")[0][5], "done") def test_marker_revives_completed_target(self): _insert_task(self.db_path, "T1") _insert_task(self.db_path, "T2", status="completed") worker = self._worker(_marker_runner(self.calls, MARKER_TO_T2)) result = worker.run_one(task=worker.get_task("T1")) self.assertEqual(result["status"], "completed") # T2 被复活：状态回 queued，续跑消息带发送方前缀。 self.assertEqual(_status(self.db_path, "T2"), "queued") self.assertEqual(_resume_message(self.db_path, "T2"), PREFIX + "请评审这份结论") self.assertEqual(_followups(self.db_path, "T2"), []) # 复活后的 T2 续跑：prompt 就是续跑消息，跑完清掉。 result = worker.run_one(task=worker.get_task("T2")) self.assertEqual(result["status"], "completed") prompts = _prompt_values(self.calls) self.assertTrue(any(PREFIX in p for p in prompts), prompts) self.assertIsNone(_resume_message(self.db_path, "T2")) def test_failed_round_markers_not_delivered(self): _insert_task(self.db_path, "T1") _insert_task(self.db_path, "T2") def failing(cmd): self.calls.append(list(cmd)) return SimpleNamespace( returncode=1, stdout=json.dumps({"status": "completed", "text": MARKER_TO_T2}), stderr="boom") worker = self._worker(failing) result = worker.run_one(task=worker.get_task("T1")) self.assertEqual(result["status"], "failed") self.assertEqual(_followups(self.db_path), []) def test_self_and_cross_project_markers_rejected(self): _insert_task(self.db_path, "T1") _insert_task(self.db_path, "T3", project_id="p2") text = ("```task-message\n{\"to\": \"T1\", \"body\": \"自己\"}\n```\n" "```task-message\n{\"to\": \"T3\", \"body\": \"别项目\"}\n```\n") worker = self._worker(_marker_runner(self.calls, text)) result = worker.run_one(task=worker.get_task("T1")) self.assertEqual(result["status"], "completed") self.assertEqual(_followups(self.db_path), []) if __name__ == "__main__": unittest.main()