mirror of
https://github.com/ThearchyHelios/AgentLab.git
synced 2026-11-15 02:21:58 +00:00
测试说明补上多 Agent 协作和让模型分类的条件分支,过长的一行拆开
This commit is contained in:
@@ -3,7 +3,7 @@
|
||||
以前点名的接入不存在时 get_chat_model 悄悄换成演示模型(mock-fast):报告由它写成假文本,双盲复算、反方审稿、审批
|
||||
质询由它来「核查」,正式运行照样出具。现在(providers/factory.named_provider):
|
||||
|
||||
- 模型调用、Agent、报告撰写节点:节点失败,报错带原因和去哪里改,一次模型调用都不发;
|
||||
- 模型调用、Agent、报告撰写、多 Agent 协作、让模型分类的条件分支:节点失败,报错带原因和去哪里改,一次模型调用都不发;
|
||||
- 双盲复算:该指标无法复算,原因「核查模型不可用(…)」,记缺口;反方审稿:这几句未能反查,原因同上;
|
||||
- 审批质询:回答 ok 为假,error 说明原因;结论句裁判:按「裁判模型不可用」照实记未裁判和缺口;
|
||||
- 助手:报错给用户,查当前用哪个模型的接口照常返回、写明原因;门控模型:转人工审批、写明原因;
|
||||
@@ -458,7 +458,8 @@ async def test_a_scheduled_formal_run_naming_a_broken_provider_fails_and_is_stil
|
||||
try:
|
||||
await loop.fire_due(NINE + timedelta(seconds=5))
|
||||
async with SessionLocal() as session:
|
||||
[fire] = list((await session.execute(select(ScheduleFire).where(ScheduleFire.schedule_id == sid))).scalars())
|
||||
fired = await session.execute(select(ScheduleFire).where(ScheduleFire.schedule_id == sid))
|
||||
[fire] = list(fired.scalars())
|
||||
assert fire.status == "started" and fire.run_id
|
||||
run = await wait(fire.run_id, ("succeeded", "failed", "cancelled"))
|
||||
assert run.status == "failed" and run.run_class == "formal"
|
||||
|
||||
Reference in New Issue
Block a user