mirror of
https://github.com/ThearchyHelios/AgentLab.git
synced 2026-11-15 15:01:59 +00:00
ThearchyHelios
2e7d6103ea
- 配置：database_url、checkpoint_url（启动时校验写法，只认 PostgreSQL）、连接池四项；启动日志写用的是哪种库（主机、库名、用户，不写密码） - 引擎（db/base.make_engine）：SQLite 照旧 WAL、外键、忙等；PostgreSQL 用连接池、探活、会话时区 UTC，装上方言调整（db/postgres.py）：建表时 String(n) 是不限长的 VARCHAR（SQLite 从不查长度，导过来的存量也可能超长）、ORDER BY 里可空列的 NULL 位置照 SQLite 补写（升序在前、降序在后，窗口函数里同样）；Integer 列在 PostgreSQL 上是 BIGINT（列挂到表上时换成 with_variant，绑定参数跟着列按 BIGINT 转换）；JSON 列两边都存原文（PostgreSQL 的 json，不用 JSONB：JSONB 重排对象里的键，快照的列序、表的先后读回来就变了）；JSON 里的 NaN 写成 null、NUL 去掉（PostgreSQL 不收） - 迁移按方言：已有的列和索引用 inspector 认；PostgreSQL 上补列的类型取模型里这一列的、缺省值取迁移表里的（布尔 0/1 换成 false/true），ADD COLUMN IF NOT EXISTS；去重语句里的布尔按方言写；部分唯一索引补 postgresql_where；init_db 在 PostgreSQL 上拿咨询锁（几个进程同时启动不撞） - 应用库上的 SQLite 专用写法改成按方言编译的表达式（db/portable.py）：json_extract 的十几处（口径召回、出具、对账、个人应用、幂等键、访问控制、数据源变化通知、剖析任务、待办超时）换成 json_eq / json_flag / json_text，json_type 换成 json_is_string（两种库同一套带类型的比较）；能力登记的 sqlite insert 换成按会话方言的 insert_ignore；能力使用次数的 json_each 按方言分支；附件「没绑记录」的条件不靠 json 的等号；按报错文字认唯一约束的地方换成 unique_violation（按表和列认，两种库的报错都认得） - 断点（engine/checkpoints.py）：PostgreSQL 时开 psycopg 连接池、在咨询锁里 setup；runner 只认 BaseCheckpointSaver，读断点仍只走 saver 的公开接口 - 依赖：新 extra postgres（asyncpg、psycopg[binary] 3、psycopg-pool、langgraph-checkpoint-postgres 3.1.2）；langgraph-checkpoint 卡在 <4.3（断点格式就在这个包里，postgres 包新版要 4.3，不动已装的 4.2.0）；锁文件补上新包 - 运行收尾：记 action.unknown、安排核实做完才把运行摘掉（以前先摘掉，等「不在跑了」再读事件的一方在库慢时读不到这一条） - 测试：AGENTLAB_TEST_DATABASE_URL 时跑在 PostgreSQL 上（每个 xdist 进程建自己的库、按事件循环分的连接池 tests/pgpool.py），不设时盖掉环境里的数据库地址照旧用 SQLite；scripts/test-postgres.sh 起临时的 postgres:16 容器跑全量；sqlite_only 标记；另开连接改库的几个竞态测试、触发器测试改成两种库都能跑（tests/appdb.py）；新增 test_db_portable 守着方言调整、JSON 键序、超长字符串、大整数、NULL 排序、时区、PostgreSQL 老库迁移
376 lines
17 KiB
Python
376 lines
17 KiB
Python
"""发起运行的规则搬到 engine/launch.py：HTTP 发起和定时任务到点发起共用同一个 launch_run。
|
||
|
||
- 每一种拒绝：POST /api/runs 的状态码、detail、code 先按搬家之前的原话写死（这一组先在旧代码上通过）；
|
||
直接调 launch_run 抛 LaunchRefused，和 HTTP 的返回逐条一致
|
||
- 发起时定下的治理事件（受管级别那条日志、口径升版处置）在服务里写：不经 HTTP 发起的正式运行一样有；
|
||
它们和运行记录同一个事务提交，排在 run.started 前面、在封存范围里，写不进去就不建运行
|
||
- 旧的导入路径（app.api.runs 的 missing_tools、node_setting、tool_bindings、TOOL_MISSING）照旧可用；
|
||
runner 不再从接口层导入
|
||
"""
|
||
from __future__ import annotations
|
||
|
||
import asyncio
|
||
import inspect
|
||
import uuid
|
||
|
||
import pytest
|
||
from httpx import ASGITransport, AsyncClient
|
||
from sqlalchemy import select
|
||
|
||
from app.db.base import SessionLocal
|
||
from app.db.models import Run, RunEvent, Workflow, WorkflowVersion
|
||
from app.engine.runner import run_manager
|
||
from app.engine.schema import GraphSpec
|
||
from app.main import app
|
||
|
||
|
||
@pytest.fixture(autouse=True)
|
||
async def engine_up():
|
||
await run_manager.setup()
|
||
yield
|
||
await run_manager.shutdown()
|
||
|
||
|
||
@pytest.fixture
|
||
async def client():
|
||
async with AsyncClient(transport=ASGITransport(app=app), base_url="http://test") as c:
|
||
yield c
|
||
|
||
|
||
def node(nid, ntype, label=None, **config):
|
||
return {"id": nid, "type": ntype, "position": {"x": 0, "y": 0}, "data": {"label": label or nid, "config": config}}
|
||
|
||
|
||
def chain(*nodes):
|
||
return {"nodes": list(nodes), "edges": [{"source": a["id"], "target": b["id"]} for a, b in zip(nodes, nodes[1:])]}
|
||
|
||
|
||
SIMPLE = chain(node("start", "input"), node("out", "output", fields=[{"name": "r", "value": "完成"}]))
|
||
|
||
|
||
def _tag() -> str:
|
||
return uuid.uuid4().hex[:8]
|
||
|
||
|
||
async def workflow(graph=SIMPLE, *, published=None, versions=(), snapshot=None, name=None) -> tuple[str, str]:
|
||
"""建一个工作流。published 是当前发布版本号，versions 是要建快照的版本号，snapshot 是快照里的图（默认同 graph）。"""
|
||
name = name or f"发起测试·周报-{_tag()}"
|
||
async with SessionLocal() as session:
|
||
wf = Workflow(name=name, graph=graph, status="published" if published else "draft",
|
||
published_version=published, version=max(versions, default=1))
|
||
session.add(wf)
|
||
await session.flush()
|
||
for v in versions:
|
||
session.add(WorkflowVersion(workflow_id=wf.id, version=v,
|
||
graph=snapshot if snapshot is not None else graph))
|
||
await session.commit()
|
||
return wf.id, name
|
||
|
||
|
||
def tool_graph(*tools: tuple[str, str]) -> dict:
|
||
"""每个 (节点名, 工具名) 一个调用工具节点。"""
|
||
middle = [node(f"t{i}", "tool", label, tool=tool, args={}) for i, (label, tool) in enumerate(tools)]
|
||
return chain(node("start", "input"), *middle, node("out", "output", fields=[{"name": "r", "value": "完成"}]))
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 每一种拒绝：(请求体, (状态码, detail, code))。detail 是 None 时只比两边一致，另由 check 查要点
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
async def case_formal_without_workflow():
|
||
return {"run_class": "formal"}, (400, "正式运行需要从已保存的工作流发起：请求中没有指定工作流", None)
|
||
|
||
|
||
async def case_formal_with_canvas_graph():
|
||
wf, _ = await workflow(published=1, versions=(1,))
|
||
return ({"run_class": "formal", "workflow_id": wf, "graph": SIMPLE},
|
||
(400, "正式运行只能使用已发布的版本，画布上未发布的修改不会生效。请先保存并发布", None))
|
||
|
||
|
||
async def case_formal_workflow_gone():
|
||
return {"run_class": "formal", "workflow_id": f"gone{_tag()}"}, (404, "工作流不存在，可能已被删除", None)
|
||
|
||
|
||
async def case_formal_never_published():
|
||
wf, name = await workflow()
|
||
return {"run_class": "formal", "workflow_id": wf}, (409, f"「{name}」还没有发布版本，请先在编排页发布", None)
|
||
|
||
|
||
async def case_formal_not_the_published_version():
|
||
"""v1 是草稿快照（没有发布级别，也不是当前发布版）。应用平台之后正式运行接受「已发布且未退役」的版本，草稿照旧不行，
|
||
拒绝带上机器码 run_version_unpublished（F1）。"""
|
||
wf, _ = await workflow(published=2, versions=(1, 2))
|
||
return ({"run_class": "formal", "workflow_id": wf, "version": 1},
|
||
(409, "v1 没有发布过（当前发布版本为 v2）。正式运行只能使用已发布的版本；如需运行这一版，请先发布它。",
|
||
"run_version_unpublished"))
|
||
|
||
|
||
async def case_formal_snapshot_missing():
|
||
wf, _ = await workflow(published=3, versions=(1, 2))
|
||
return {"run_class": "formal", "workflow_id": wf}, (404, "版本 v3 不存在", None)
|
||
|
||
|
||
async def case_formal_snapshot_malformed():
|
||
"""快照里的图不成形：以前由 main.py 的 ValueError 兜底回 400，detail 是校验的原话。"""
|
||
broken = {"nodes": "坏掉的图"}
|
||
wf, _ = await workflow(published=1, versions=(1,), snapshot=broken)
|
||
with pytest.raises(ValueError) as caught:
|
||
GraphSpec.model_validate(broken)
|
||
return {"run_class": "formal", "workflow_id": wf}, (400, str(caught.value), None)
|
||
|
||
|
||
def library(version="v3"):
|
||
return chain(node("start", "input"),
|
||
node("caliber", "metrics", "周报口径卡", caliber="周报口径", caliber_version=version,
|
||
metrics=[{"id": "gmv", "name": "销售额", "unit": "元", "expression": "vars.kpi.gmv"}]),
|
||
node("out", "output", fields=[{"name": "r", "value": "{{ nodes.caliber.text }}"}]))
|
||
|
||
|
||
def weekly(lib_id: str, **card) -> dict:
|
||
return chain(node("start", "input"),
|
||
node("fetch", "transform", mode="expression", expression="{'gmv': 4321.5}", assign_to="kpi"),
|
||
node("card", "metrics", "本地口径卡",
|
||
caliber_from={"workflow_id": lib_id, "workflow_version": 1, "node_id": "caliber"}, **card),
|
||
node("out", "output", fields=[{"name": "r", "value": "{{ nodes.card.text }}"}]))
|
||
|
||
|
||
async def upstream_with_two_versions() -> str:
|
||
async with SessionLocal() as session:
|
||
lib = Workflow(name=f"口径库-{_tag()}", graph=library(), status="published", published_version=2, version=2)
|
||
session.add(lib)
|
||
await session.flush()
|
||
# 两版都发布过（level 记着）：口径卡固定的版本须「已发布且未退役」（D-q）
|
||
session.add(WorkflowVersion(workflow_id=lib.id, version=1, graph=library(), level="published"))
|
||
session.add(WorkflowVersion(workflow_id=lib.id, version=2, graph=library(version="v4"), level="published"))
|
||
await session.commit()
|
||
return lib.id
|
||
|
||
|
||
async def case_formal_caliber_upgrade_undeclared():
|
||
lib = await upstream_with_two_versions()
|
||
graph = weekly(lib)
|
||
wf, _ = await workflow(graph, published=1, versions=(1,))
|
||
return {"run_class": "formal", "workflow_id": wf}, (409, None, None)
|
||
|
||
|
||
def check_formal_caliber_upgrade_undeclared(detail: str) -> None:
|
||
assert "「本地口径卡」" in detail and "v1" in detail and "v2" in detail and "「上游发布新版本时」" in detail
|
||
|
||
|
||
async def case_saved_workflow_gone():
|
||
return {"workflow_id": f"gone{_tag()}"}, (404, "工作流不存在，可能已被删除", None)
|
||
|
||
|
||
async def case_nothing_to_run():
|
||
return {}, (400, "没有要运行的内容：请求中既没有工作流 ID，也没有工作流内容", None)
|
||
|
||
|
||
async def case_saved_workflow_is_empty():
|
||
wf, _ = await workflow({})
|
||
return {"workflow_id": wf}, (400, "没有要运行的内容：请求中既没有工作流 ID，也没有工作流内容", None)
|
||
|
||
|
||
async def case_missing_datasource_tool():
|
||
source = f"gone{_tag()}"
|
||
return ({"graph": tool_graph(("取数", f"db_query__{source}"))},
|
||
(422, f"绑定的工具不存在：「取数」（调用工具）绑定的 db_query__{source}。请到「数据」页接入，或在节点中重新选择",
|
||
"run_tool_missing"))
|
||
|
||
|
||
async def case_missing_custom_tool():
|
||
tool = f"ghost_{_tag()}"
|
||
return ({"graph": tool_graph(("查编码", tool))},
|
||
(422, f"绑定的工具不存在：「查编码」（调用工具）绑定的 {tool}。请到「工具」页接入，或在节点中重新选择",
|
||
"run_tool_missing"))
|
||
|
||
|
||
async def case_missing_tools_of_both_kinds():
|
||
source, tool = f"gone{_tag()}", f"ghost_{_tag()}"
|
||
return ({"graph": tool_graph(("取数", f"db_schema__{source}"), ("查编码", tool))},
|
||
(422, f"绑定的工具不存在：「取数」（调用工具）绑定的 db_schema__{source}；「查编码」（调用工具）绑定的 {tool}。"
|
||
"数据源工具请到「数据」页接入，其他工具请到「工具」页接入，或在节点中重新选择", "run_tool_missing"))
|
||
|
||
|
||
async def case_more_than_eight_missing_tools():
|
||
tag = _tag()
|
||
tools = [(f"第{i}步", f"ghost_{tag}_{i}") for i in range(10)]
|
||
listed = "；".join(f"「{label}」（调用工具）绑定的 {tool}" for label, tool in tools[:8])
|
||
return ({"graph": tool_graph(*tools)},
|
||
(422, f"绑定的工具不存在：{listed}等 10 处。请到「工具」页接入，或在节点中重新选择", "run_tool_missing"))
|
||
|
||
|
||
async def case_formal_missing_tool():
|
||
source = f"gone{_tag()}"
|
||
wf, _ = await workflow(tool_graph(("取数", f"db_query__{source}")), published=1, versions=(1,))
|
||
return ({"run_class": "formal", "workflow_id": wf},
|
||
(422, f"绑定的工具不存在：「取数」（调用工具）绑定的 db_query__{source}。请到「数据」页接入，或在节点中重新选择",
|
||
"run_tool_missing"))
|
||
|
||
|
||
async def case_graph_fails_validation():
|
||
return ({"graph": {"nodes": [], "edges": []}},
|
||
(400, "工作流未通过校验，请先在画布上修正标红的节点再运行：画布上还没有节点，请先从节点库拖入一个节点", None))
|
||
|
||
|
||
#: 「不存在」的工作流、数据源、自定义工具都带随机后缀，库里不会恰好有同名的
|
||
CASES = {name[len("case_"):]: fn for name, fn in dict(globals()).items() if name.startswith("case_")}
|
||
|
||
|
||
def _body(status: int, detail: str, code: str | None) -> dict:
|
||
return {"detail": detail, **({"code": code} if code else {})}
|
||
|
||
|
||
async def _run_count() -> int:
|
||
async with SessionLocal() as session:
|
||
return len((await session.execute(select(Run.id))).all())
|
||
|
||
|
||
@pytest.mark.parametrize("case", list(CASES))
|
||
async def test_http_refusals_are_unchanged(client, case):
|
||
"""HTTP 这一侧逐字写死：搬家前后都得是这些状态码和原话。"""
|
||
body, (status, detail, code) = await CASES[case]()
|
||
before = await _run_count()
|
||
res = await client.post("/api/runs", json=body)
|
||
assert res.status_code == status, res.text
|
||
got = res.json()
|
||
if detail is None:
|
||
globals()[f"check_{case}"](got["detail"])
|
||
assert set(got) == ({"detail", "code"} if code else {"detail"}), got
|
||
else:
|
||
assert got == _body(status, detail, code)
|
||
assert await _run_count() == before, "被拒的请求不该留下运行记录"
|
||
|
||
|
||
def _launch_kwargs(body: dict) -> dict:
|
||
"""请求体换成 launch_run 的参数：input 对应 input_payload，其余同名。"""
|
||
out = {k: v for k, v in body.items() if k != "input"}
|
||
if "input" in body:
|
||
out["input_payload"] = body["input"]
|
||
return out
|
||
|
||
|
||
@pytest.mark.parametrize("case", list(CASES))
|
||
async def test_launch_run_refuses_exactly_like_http(client, case):
|
||
from app.engine.launch import LaunchRefused, launch_run
|
||
|
||
body, (status, detail, code) = await CASES[case]()
|
||
before = await _run_count()
|
||
async with SessionLocal() as session:
|
||
with pytest.raises(LaunchRefused) as caught:
|
||
await launch_run(session, actor="定时任务", **_launch_kwargs(body))
|
||
refused = caught.value
|
||
assert (refused.status_code, refused.code) == (status, code)
|
||
if detail is not None:
|
||
assert refused.message == detail
|
||
res = await client.post("/api/runs", json=body)
|
||
assert res.status_code == refused.status_code
|
||
assert res.json() == _body(refused.status_code, refused.message, refused.code)
|
||
assert await _run_count() == before
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 发起成功：参数照转、治理事件在服务里写
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
async def _settle(run_id: str) -> Run:
|
||
for _ in range(400):
|
||
async with SessionLocal() as session:
|
||
row = await session.get(Run, run_id)
|
||
if row.status in ("succeeded", "failed", "cancelled") and row.manifest_seq is not None:
|
||
return row
|
||
await asyncio.sleep(0.03)
|
||
raise AssertionError(f"运行 {run_id} 没有结束：{row.status}")
|
||
|
||
|
||
async def _events(run_id: str, etype: str) -> list[RunEvent]:
|
||
async with SessionLocal() as session:
|
||
return list((await session.execute(
|
||
select(RunEvent).where(RunEvent.run_id == run_id, RunEvent.type == etype).order_by(RunEvent.seq)
|
||
)).scalars())
|
||
|
||
|
||
async def test_a_formal_launch_records_the_governance_notes():
|
||
"""不经 HTTP 发起的正式运行（定时任务）也记下按什么级别出具、口径升版怎么处置的。"""
|
||
from app.engine.launch import launch_run
|
||
|
||
lib = await upstream_with_two_versions()
|
||
wf, name = await workflow(weekly(lib, upgrade_policy="dual"), published=1, versions=(1,))
|
||
async with SessionLocal() as session:
|
||
run = await launch_run(session, workflow_id=wf, run_class="formal", input_payload={"week": "第 40 周"},
|
||
actor="定时任务")
|
||
assert (run.run_class, run.version, run.workflow_name, run.started_by) == ("formal", 1, name, "定时任务")
|
||
assert run.version_hash and run.input == {"week": "第 40 周"}
|
||
row = await _settle(run.id)
|
||
assert row.status == "succeeded", row.error
|
||
notes = [e for e in await _events(run.id, "log") if e.data.get("code") == "run_governance"]
|
||
assert len(notes) == 1 and notes[0].data["governed"] is False and "v1" in notes[0].data["message"]
|
||
[upgrade] = await _events(run.id, "caliber.upgrade")
|
||
assert upgrade.node_id == "card" and upgrade.data["policy"] == "dual" and upgrade.data["latest"] == 2
|
||
# 发起时就有的事件排在 run.started 前面，都在封存范围里
|
||
[started] = await _events(run.id, "run.started")
|
||
assert notes[0].seq < upgrade.seq < started.seq <= row.manifest_seq
|
||
|
||
|
||
async def test_launch_events_are_written_with_the_run_or_not_at_all():
|
||
"""发起时定下的事件和运行记录同一个事务提交：写不进去就不建运行，launch_run 抛出异常就是没有发起。以前运行
|
||
建好、开始执行之后才另外写，写失败时运行已经在执行，调用方（定时任务）只拿到一个异常，会把一次正在执行的
|
||
正式运行记成未能发起，封存范围里也少了受管级别那条日志。"""
|
||
from sqlalchemy import text
|
||
from sqlalchemy.exc import IntegrityError
|
||
|
||
from app.db.base import engine
|
||
from app.engine.launch import launch_run
|
||
from tests.appdb import refuse_inserts
|
||
|
||
lib = await upstream_with_two_versions()
|
||
wf, _ = await workflow(weekly(lib, upgrade_policy="dual"), published=1, versions=(1,))
|
||
before = await _run_count()
|
||
# 口径升版处置那条写不进去。走异步引擎建删触发器：同步连接等写锁时会卡住事件循环，而持锁的正是这个循环里的写入
|
||
create, drop = refuse_inserts("run_events", "NEW.type = 'caliber.upgrade'", name="refuse_upgrade_note")
|
||
async with engine.begin() as conn:
|
||
for sql in create:
|
||
await conn.execute(text(sql))
|
||
try:
|
||
async with SessionLocal() as session:
|
||
with pytest.raises(IntegrityError):
|
||
await launch_run(session, workflow_id=wf, run_class="formal", actor="定时任务")
|
||
assert await _run_count() == before, "没写进治理事件，却留下了一条运行"
|
||
finally:
|
||
async with engine.begin() as conn:
|
||
for sql in drop:
|
||
await conn.execute(text(sql))
|
||
|
||
|
||
async def test_an_exploratory_launch_takes_the_saved_graph_and_the_run_defaults():
|
||
from app.engine.launch import launch_run
|
||
|
||
wf, name = await workflow()
|
||
async with SessionLocal() as session:
|
||
run = await launch_run(session, workflow_id=wf, actor=None)
|
||
assert (run.run_class, run.workflow_id, run.workflow_name, run.version) == ("exploratory", wf, name, None)
|
||
assert (run.memory_scope, run.collection) == ("default", "default") and run.input == {}
|
||
row = await _settle(run.id)
|
||
assert row.status == "succeeded", row.error
|
||
assert [e for e in await _events(run.id, "log") if e.data.get("code") == "run_governance"] == []
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 旧的导入路径
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
def test_the_old_import_paths_still_work():
|
||
from app.api import runs as runs_api
|
||
from app.engine import launch
|
||
|
||
for name in ("missing_tools", "node_setting", "tool_bindings", "TOOL_MISSING"):
|
||
assert getattr(runs_api, name) is getattr(launch, name), name
|
||
|
||
|
||
def test_the_runner_no_longer_imports_from_the_api_layer():
|
||
from app.engine import runner
|
||
|
||
assert "app.api.runs" not in inspect.getsource(runner)
|