相对第一版 46fc976 的完整变更。组员迁移对照表见 docs/20。
一、对外契约对齐 docs/05(破坏性,共 4 处,组员需按 docs/20 调整)
1) 配置发布端点改为文档规定的复数资源名:submit→validations、
approve→reviews(需 body decision)、activate→activations、
rollback→rollbacks;第一版这 4 个动词式路径 docs/05 从未定义过。
2) 错误码由 8 个笼统码改为 15 个具体语义码(FORBIDDEN→AGENT_PERMISSION_DENIED、
UNAUTHORIZED→AUTHENTICATION_REQUIRED、CONFLICT→RESOURCE_VERSION_CONFLICT、
RESOURCE_NOT_FOUND→RUN_NOT_FOUND/SESSION_NOT_FOUND 等),
输入类错误状态码 400→422。
3) POST /api/v1/agent-runs 与 GET /api/v1/agent-runs/{run_id} 统一为
{data, meta} 信封(data 内字段名与语义未变)。
4) 错误响应体统一为 {error:{code,message,retryable,field_errors}, meta:{trace_id}},
不再返回 FastAPI 默认的 {"detail": ...}。
二、数据库基线与约束
新增 39 张表的基线迁移(链根)与联合唯一键纠偏(4 张表、删 8 增 4,幂等收敛);
撤下 config_release 的双人复核 CHECK(应用层已允许自审,审核节点保留,
自审如实写入 reviewer_id);记忆 active key 生成列与唯一键;
activate 开始记录 supersedes_release_id 使版本链可追溯。
docs/00 基线未修改,未重命名或删除任何表与字段。
三、修复会静默出错或无报错的缺陷
- 跑完集成测试后平台会静默失去生效配置:清理只删自己创建的版本,却没有恢复被它
顶成 superseded 的原生效版本,且审计一并删除因而完全无痕,表现为所有工具被拒
但没有任何报错。已修清理逻辑并加恢复。
- Worker 单轮异常导致进程退出;记忆抽取调用方的“事务已开始”异常;
召回缓存丢失 degraded 标记;连接时区未生效导致 created_at/updated_at 差 8 小时;
.env 与 os.getenv 密钥来源分裂导致“没有可用的已批准模型端点”。
- 记忆信号识别漏判与跨键误命中;SSE 未带 Accept 的协商行为。
四、功能补齐
记忆链路 P1/P2/P3(抽取、受控词表、召回与缓存、生命周期级联及投影事件)、
fin_* 场内交易只读 ORM 层、agent_intent_config 状态流转并在运行期真正生效、
限流(Redis 固定窗口、故障一律放行)、游标校验、trace_id 中间件、
示例业务 Agent fund_query_demo 与一键端到端验证脚本,以及审计/指纹/迁移状态工具。
五、文档与验证
新增 docs/19(业务 Agent 接入实操)、docs/20(第一版迁移指南)与 docs/evidence 证据;
docs/01/02/06/08/09/17 同步实现现状。
验证结果:ruff 通过、mypy 103 文件无错、unit+contract 447 passed、
integration 29 passed、acceptance_check --production 7 PASS、
demo_agent_e2e 9/9 PASS(含失败关闭反证)。
145 lines
5.0 KiB
Python
145 lines
5.0 KiB
Python
"""运行取消路径的错误码契约测试(不连数据库)。
|
||
|
||
`POST /api/v1/agent-runs/{run_id}/cancellations` 的语义来自 `docs/05-接口文档.md` §6.4:
|
||
|
||
- 运行不存在或不属于当前用户 → `404 RUN_NOT_FOUND`(不泄露运行是否存在);
|
||
- 已成功、已失败或进入最终提交事务 → `409 RUN_NOT_CANCELLABLE`;
|
||
- **重复取消返回同一状态**:`cancel_requested` 与 `cancelled` 都必须幂等返回当前快照,
|
||
不得把重复取消当成错误(`RUN_CANCELLED` 不是 HTTP 响应码,见下方幂等断言);
|
||
- 取消成功后 `request_idempotency` 落 `failed + RUN_CANCELLED`。
|
||
|
||
测试手法与 `test_public_platform_service.py` 一致:替换权限闸门与事务服务,
|
||
直接驱动 `_cancel`,因此既不需要数据库,也不需要真实 repository。
|
||
"""
|
||
|
||
from datetime import UTC, datetime
|
||
from typing import Any
|
||
|
||
import pytest
|
||
|
||
from app.core.errors import RunNotCancellableError, RunNotFoundError
|
||
|
||
NOW = datetime.now(UTC).replace(tzinfo=None)
|
||
|
||
|
||
class FakeRun:
|
||
def __init__(self, status: str) -> None:
|
||
self.run_id = "run-1"
|
||
self.user_id = 9001
|
||
self.status = status
|
||
self.cancel_requested_at: datetime | None = None
|
||
self.session_id = "session-1"
|
||
self.idempotency_id = 7
|
||
|
||
|
||
class FakeSession:
|
||
"""只记录 `execute` 的绑定参数,用来断言 `request_idempotency` 的落库取值。"""
|
||
|
||
def __init__(self, run: FakeRun | None) -> None:
|
||
self._run = run
|
||
self.updates: list[dict[str, Any]] = []
|
||
|
||
async def scalar(self, _query: Any) -> FakeRun | None:
|
||
return self._run
|
||
|
||
async def execute(self, statement: Any) -> None:
|
||
self.updates.append(dict(statement.compile().params))
|
||
|
||
|
||
def service() -> Any:
|
||
from app.service.public_platform_service import PublicPlatformService
|
||
|
||
return PublicPlatformService.__new__(PublicPlatformService)
|
||
|
||
|
||
async def cancel(run: FakeRun | None) -> tuple[tuple[dict[str, Any], str], FakeSession]:
|
||
session = FakeSession(run)
|
||
result = await service()._cancel(session, "run-1", 9001, NOW)
|
||
return result, session
|
||
|
||
|
||
def idempotency_writes(session: FakeSession) -> list[dict[str, Any]]:
|
||
return [
|
||
update for update in session.updates
|
||
if update.get("status") == "failed" and "error_code" in update
|
||
]
|
||
|
||
|
||
async def test_missing_or_foreign_run_is_run_not_found() -> None:
|
||
"""越权与不存在必须返回同一码,否则可以据状态码探测运行是否存在。"""
|
||
with pytest.raises(RunNotFoundError) as excinfo:
|
||
await cancel(None)
|
||
|
||
assert excinfo.value.code == "RUN_NOT_FOUND"
|
||
assert excinfo.value.status_code == 404
|
||
|
||
|
||
@pytest.mark.parametrize("status", ["succeeded", "failed"])
|
||
async def test_terminal_run_is_not_cancellable(status: str) -> None:
|
||
with pytest.raises(RunNotCancellableError) as excinfo:
|
||
await cancel(FakeRun(status))
|
||
|
||
assert excinfo.value.code == "RUN_NOT_CANCELLABLE"
|
||
assert excinfo.value.status_code == 409
|
||
|
||
|
||
@pytest.mark.parametrize(
|
||
("status", "expected_status"),
|
||
[("cancel_requested", "cancel_requested"), ("cancelled", "cancelled")],
|
||
)
|
||
async def test_repeated_cancel_is_idempotent_and_never_an_error(
|
||
status: str, expected_status: str
|
||
) -> None:
|
||
"""重复取消必须返回同一状态,不得以 `409 RUN_CANCELLED` 之类的错误回应。"""
|
||
run = FakeRun(status)
|
||
run.cancel_requested_at = NOW
|
||
|
||
(payload, session_id), session = await cancel(run)
|
||
|
||
assert payload == {
|
||
"run_id": "run-1",
|
||
"status": expected_status,
|
||
"cancel_requested_at": NOW.isoformat() + "Z",
|
||
}
|
||
assert session_id == "session-1"
|
||
# 重复取消不改写幂等记录,也不产生新的状态迁移。
|
||
assert idempotency_writes(session) == []
|
||
|
||
|
||
async def test_queued_run_is_marked_cancel_requested_and_terminates_original_request() -> None:
|
||
run = FakeRun("queued")
|
||
|
||
(payload, session_id), session = await cancel(run)
|
||
|
||
assert run.status == "cancel_requested"
|
||
assert run.cancel_requested_at == NOW
|
||
assert payload["run_id"] == "run-1"
|
||
assert payload["status"] == "cancel_requested"
|
||
assert session_id == "session-1"
|
||
# 文档 §6.4:原请求以 failed + RUN_CANCELLED 结束,不扩展状态枚举。
|
||
writes = idempotency_writes(session)
|
||
assert writes == [{"id_1": 7, "status": "failed", "error_code": "RUN_CANCELLED",
|
||
"updated_at": NOW}]
|
||
|
||
|
||
async def test_running_run_is_cancellable_too() -> None:
|
||
run = FakeRun("running")
|
||
|
||
(payload, _session_id), session = await cancel(run)
|
||
|
||
assert payload["status"] == "cancel_requested"
|
||
assert len(idempotency_writes(session)) == 1
|
||
|
||
|
||
async def test_cancelled_run_never_reports_run_cancelled_as_http_error() -> None:
|
||
"""`RUN_CANCELLED` 是 request_idempotency 的状态标识,不是客户端错误码。"""
|
||
from app.core import errors
|
||
|
||
assert not hasattr(errors, "RunCancelledError")
|
||
run = FakeRun("cancelled")
|
||
run.cancel_requested_at = NOW
|
||
|
||
payload, _session_id = (await cancel(run))[0]
|
||
|
||
assert payload["status"] == "cancelled"
|