Files
roboco/tests/e2e_smoke/test_war_room_loop.py
T

340 lines
12 KiB
Python
Raw Normal View History

"""Scenario: the War Room (Board Program) loop end to end.
Mirrors test_periscope_loop.py's arm -> originate -> dedup shape (org scope,
no per-project opt-in to RUN, RoboCo-project FK anchor) AND
test_feature_spotlight.py's do_server wiring-regression check + real-route
propose call. The genuinely new piece War Room adds: EVENT trigger with a
REAL originator (unlike Coroner's always-None stub — proven here via
``open_program_cycle`` standing in for the CEO's "run now" call, since this
harness has no HTTP client), an X-credentials gate the origination step must
clear, and a per-item materialization loop (N ordered ``x_campaign`` drafts
from one ``propose_campaign`` call, not one item on the exploration task
itself).
"""
from __future__ import annotations
from datetime import UTC, datetime, timedelta
from typing import TYPE_CHECKING, Any
from uuid import UUID, uuid4
from roboco.foundation import identity as _foundation
from tests.e2e_smoke.harness import ScriptedAgent, expect_ok
if TYPE_CHECKING:
from sqlalchemy.ext.asyncio import AsyncSession
from tests.e2e_smoke.harness import E2EStack
ONE = 1
THREE = 3
def _future(hours: float) -> str:
return (datetime.now(UTC) + timedelta(hours=hours)).isoformat()
def _seed_system_hom_and_secretary(stack: E2EStack) -> tuple[Any, str]:
"""Seed ``system`` + ``head-marketing`` + ``secretary-1`` at their FIXED
foundation UUIDs, plus a project those own — ``WarRoomEngine._originate``
assigns the campaign-planning task to ``head-marketing``'s fixed UUID
(not ``seed_company``'s random ``uuid4()`` row, which would also collide
on the unique agent slug), and ``XEngine._originate_post`` assigns every
materialized draft to ``secretary-1``'s fixed UUID — both FKs need real
rows there. Returns ``(project_id, project_slug)`` — the id scopes every
read below to THIS test's own rows: ``ROBOCO_TEST_DB_*`` is the same
physical Postgres a non-e2e suite run earlier in the same invocation
also wrote to (e.g. test_board_programs_api.py's own war_room run-now
test leaves a CANCELLED ``board_war_room`` row behind, cleaned up but
never deleted), so an unscoped "count every board_war_room row" query
would double-count across runs.
"""
from roboco.db.tables import AgentTable, ProjectTable
from roboco.models import AgentRole, AgentStatus, Team
slug = f"e2e-war-room-{uuid4().hex[:8]}"
project_id = uuid4()
async def _run(session: AsyncSession) -> None:
for agent_uuid, agent_slug, role, team in (
(_foundation.AGENTS["system"].uuid, "system", AgentRole.SYSTEM, None),
(
_foundation.AGENTS["head-marketing"].uuid,
"head-marketing",
AgentRole.HEAD_MARKETING,
Team.BOARD,
),
(
_foundation.AGENTS["secretary-1"].uuid,
"secretary-1",
AgentRole.SECRETARY,
None,
),
):
if await session.get(AgentTable, agent_uuid) is not None:
continue
session.add(
AgentTable(
id=agent_uuid,
name=agent_slug,
slug=agent_slug,
role=role,
team=team,
status=AgentStatus.ACTIVE,
model_config={},
system_prompt=agent_slug,
capabilities=[],
permissions={},
metrics={},
)
)
session.add(
ProjectTable(
id=project_id,
name="RoboCo",
slug=slug,
git_url="https://example.com/roboco.git",
default_branch="master",
protected_branches=["master"],
assigned_cell=Team.BACKEND,
created_by=_foundation.AGENTS["system"].uuid,
is_active=True,
)
)
stack.run_db(_run)
return project_id, slug
def _arm(stack: E2EStack, project_slug: str) -> None:
"""Arm via the settings-store key — the ONLY arming path (no legacy env
flag exists for war_room) — point ``self_heal_project_slug`` at the
seeded project, and seed X credentials: ``WarRoomEngine``'s own creds
gate would otherwise no-op every origination call, mirroring XEngine's
release/spotlight guard."""
from roboco.config import settings as cfg
from roboco.db.tables import SystemSettingTable
from roboco.services.x_credentials import get_x_credentials_service
cfg.self_heal_project_slug = project_slug
async def _run(session: AsyncSession) -> None:
session.add(
SystemSettingTable(key="board_program.war_room.enabled", value="true")
)
await get_x_credentials_service(session).set_credentials(
api_key="ak-test",
api_secret="as-test",
access_token="at-test",
access_token_secret="ats-test",
)
stack.run_db(_run)
def _run_now(stack: E2EStack) -> Any:
"""The CEO's "run now" seam — ``open_program_cycle`` does not check
trigger kind, only ``run_due_programs`` (the cron loop) does, so this is
the real EVENT-program run-now path."""
from roboco.services.board_programs import get_board_program_engine
async def _run(session: AsyncSession) -> Any:
task = await get_board_program_engine(session).open_program_cycle("war_room")
return task.id if task is not None else None
result: Any = stack.run_db(_run)
return result
def _run_due_programs(stack: E2EStack) -> list[str]:
from roboco.services.board_programs import get_board_program_engine
async def _run(session: AsyncSession) -> list[str]:
return await get_board_program_engine(session).run_due_programs()
result: list[str] = stack.run_db(_run)
return result
def _find_war_room_task(stack: E2EStack, project_id: Any) -> dict[str, Any]:
"""Scoped to ``project_id`` — see ``_seed_system_hom_and_secretary``'s
docstring for why an unscoped query double-counts against the shared
test Postgres."""
from roboco.db.tables import TaskTable
from roboco.services.task import WAR_ROOM_SOURCE
from sqlalchemy import select
async def _run(session: AsyncSession) -> dict[str, Any]:
rows = (
(
await session.execute(
select(TaskTable).where(
TaskTable.source == WAR_ROOM_SOURCE,
TaskTable.project_id == project_id,
)
)
)
.scalars()
.all()
)
return {
"count": len(rows),
"rows": [
{
"id": r.id,
"status": str(r.status),
"assigned_to": r.assigned_to,
"source": r.source,
"project_id": r.project_id,
"confirmed_by_human": r.confirmed_by_human,
}
for r in rows
],
}
state: dict[str, Any] = stack.run_db(_run)
return state
def _campaign_drafts(stack: E2EStack, project_id: Any) -> list[dict[str, Any]]:
"""Scoped to ``project_id`` for the same reason as ``_find_war_room_task``."""
from roboco.db.tables import TaskTable
from roboco.foundation.policy.content import markers
from roboco.services.task import X_CAMPAIGN_SOURCE
from sqlalchemy import select
async def _run(session: AsyncSession) -> list[dict[str, Any]]:
rows = (
(
await session.execute(
select(TaskTable).where(
TaskTable.source == X_CAMPAIGN_SOURCE,
TaskTable.project_id == project_id,
)
)
)
.scalars()
.all()
)
return [
{
"id": r.id,
"status": str(r.status),
"confirmed_by_human": r.confirmed_by_human,
"ref": markers.get_x_campaign_ref(r),
}
for r in rows
]
result: list[dict[str, Any]] = stack.run_db(_run)
return result
def test_war_room_loop_run_now_then_propose_campaign(e2e_stack: E2EStack) -> None:
stack = e2e_stack
project_id, project_slug = _seed_system_hom_and_secretary(stack)
_arm(stack, project_slug)
# EVENT programs are never cron-due — proves the loop itself never opens
# war_room, unlike the run-now seam below.
opened_by_loop = _run_due_programs(stack)
assert "war_room" not in opened_by_loop, opened_by_loop
# THE run-now-works-for-EVENT proof: war_room's originator is REAL
# (unlike Coroner's always-None stub), so this must actually originate.
exploration_id = _run_now(stack)
assert exploration_id is not None
state = _find_war_room_task(stack, project_id)
assert state["count"] == ONE, state
row = state["rows"][0]
assert row["id"] == exploration_id
assert row["status"] == "pending"
assert row["assigned_to"] == _foundation.AGENTS["head-marketing"].uuid
assert row["confirmed_by_human"] is False
assert row["project_id"] is not None # resolves against the RoboCo project
# The dispatcher's own dev-work skip recognizes this exact task shape.
from roboco.runtime.orchestrator import _is_non_dev_dispatch_source
assert _is_non_dev_dispatch_source({"source": row["source"]}) is True
# A second run-now while one campaign is open is refused — dedup.
second = _run_now(stack)
assert second is None
# The Head of Marketing authors the campaign through the REAL do_server
# module + /api/v1/do/propose_campaign route + ContentActions + XEngine
# wiring — the wiring-regression class check test_feature_spotlight.py
# guards (a verb granted in role_config but missing from do_server._TOOLS
# is unreachable over MCP).
hom = ScriptedAgent(
stack,
_foundation.AGENTS["head-marketing"].uuid,
"head-marketing",
"head_marketing",
)
do_module = hom._module("roboco.mcp.do_server")
assert "propose_campaign" in do_module._TOOLS, (
"propose_campaign missing from do_server._TOOLS — the MCP server "
"has no way to expose it to any role"
)
assert "propose_campaign" in do_module._REGISTERED_TOOLS, (
"propose_campaign is granted to head_marketing in role_config but "
"absent from this agent's _register_tools() output — the manifest "
"-> _register_tools -> callable chain dropped it"
)
posts = [
{
"body": "Something big is coming to RoboCo. Watch this space.",
"publish_after": _future(1),
"stage_label": "teaser",
},
{
"body": "War Room is live: plan a whole X campaign in one cycle.",
"publish_after": _future(25),
"stage_label": "launch",
},
{
"body": "War Room posts carry a recommended publish time — you "
"still approve each one yourself.",
"publish_after": _future(49),
"stage_label": "follow_up",
},
]
env = expect_ok(
hom.do(
"propose_campaign",
campaign_name="War Room launch",
posts=posts,
),
"hom propose_campaign",
)
assert env.get("status") == "campaign_proposed", env
assert env.get("task_id") == str(exploration_id), (
"propose_campaign completes the SAME exploration task it authored "
f"against, no separate materialized cycle task: {env}"
)
from tests.e2e_smoke.arcs import task_state
assert task_state(stack, UUID(str(exploration_id)))["status"] == "completed"
drafts = _campaign_drafts(stack, project_id)
assert len(drafts) == THREE, drafts
ordered = sorted(drafts, key=lambda d: d["ref"]["sequence"])
assert [d["ref"]["stage_label"] for d in ordered] == [
"teaser",
"launch",
"follow_up",
]
assert [d["ref"]["sequence"] for d in ordered] == [1, 2, THREE]
assert all(d["ref"]["campaign_name"] == "War Room launch" for d in ordered)
assert all(d["status"] == "pending" for d in ordered) # held, awaiting the CEO
assert all(d["confirmed_by_human"] is False for d in ordered)
# publish_after strictly ascends across the campaign, matching the
# order the Head of Marketing proposed it in.
timestamps = [datetime.fromisoformat(d["ref"]["publish_after"]) for d in ordered]
assert timestamps == sorted(timestamps)