feat(self-heal): open a PENDING fix task on regression, then stop

Behind the second opt-in (self_heal_originate_enabled), a detected regression
also opens a fix task into RoboCo's own delivery lifecycle and STOPS: PENDING,
unassigned, confirmed_by_human=False, team=main_pm, source=self_heal, with
synthesized acceptance criteria and a self_heal_fp= dedupe marker. It rides the
normal dispatchers only once the CEO Approve-&-Starts it; the loop itself never
calls start / approve / merge / deploy.

- TaskService: SELF_HEAL_SOURCE, extract_self_heal_fingerprint, and
  list_open_self_heal_tasks (the dedupe + open-cap basis)
- SelfHealEngine._originate: per-signal fingerprint dedupe, per-cycle and
  rolling open-task caps, repo resolved to RoboCo's own project (notify-only
  when it can't be resolved)
- 7 DB-backed tests including the never-start / never-approve invariant
This commit is contained in:
Renn F
2026-06-17 21:00:45 +02:00
parent 606ccc3327
commit bd1fb84198
3 changed files with 345 additions and 3 deletions
@@ -0,0 +1,210 @@
"""Self-heal task origination against a real Postgres DB.
The loop opens a fix task only when ``self_heal_originate_enabled``, dedupes one
open task per regression fingerprint, honors the per-cycle and rolling open-task
caps, and creates the task PENDING + UNASSIGNED + ``confirmed_by_human=False`` so
it sits inert until the CEO Approve-&-Starts it. Crucially it NEVER calls
start / approve / merge / deploy — asserted here.
"""
from __future__ import annotations
from typing import TYPE_CHECKING
from unittest.mock import AsyncMock
import pytest
from roboco.config import settings as cfg
from roboco.db.tables import AgentTable, ProjectTable
from roboco.foundation import identity as _foundation
from roboco.models.base import AgentRole, AgentStatus, TaskStatus, Team
from roboco.services.notification import NotificationService
from roboco.services.self_heal_engine import SelfHealEngine
from roboco.services.task import TaskService, get_task_service
from roboco.services.telemetry import TelemetrySample
if TYPE_CHECKING:
from sqlalchemy.ext.asyncio import AsyncSession
SYSTEM_UUID = _foundation.AGENTS["system"].uuid
SLUG = "roboco"
ONE = 1
class _FakeSource:
def __init__(self, samples: list[TelemetrySample]) -> None:
self._samples = samples
async def fetch(self) -> list[TelemetrySample]:
return list(self._samples)
def _breach(signal: str) -> TelemetrySample:
return TelemetrySample(
signal_name=signal,
value=1.0,
threshold=1.0,
window="latest_completed_run",
repo_hint=SLUG,
observed_at="2026-06-17T00:00:00Z",
raw_ref="https://github.com/x/roboco/actions/runs/1",
detail=f"{signal} concluded 'failure'",
)
async def _seed_project(session: AsyncSession, slug: str = SLUG) -> None:
"""Seed the system agent (FK target for created_by) + RoboCo's own project."""
session.add(
AgentTable(
id=SYSTEM_UUID,
name="System",
slug=f"system-{slug}",
role=AgentRole.SYSTEM,
team=None,
status=AgentStatus.ACTIVE,
model_config={},
system_prompt="system",
capabilities=[],
permissions={},
metrics={},
)
)
await session.flush()
session.add(
ProjectTable(
name="RoboCo",
slug=slug,
git_url="https://github.com/x/roboco.git",
default_branch="master",
protected_branches=["master"],
assigned_cell=Team.BACKEND,
created_by=SYSTEM_UUID,
is_active=True,
)
)
await session.flush()
def _enable(monkeypatch: pytest.MonkeyPatch, **overrides: object) -> None:
monkeypatch.setattr(cfg, "self_heal_enabled", True)
monkeypatch.setattr(cfg, "self_heal_originate_enabled", True)
monkeypatch.setattr(cfg, "self_heal_max_open_tasks", 5)
monkeypatch.setattr(cfg, "self_heal_max_per_cycle", 5)
for key, value in overrides.items():
monkeypatch.setattr(cfg, key, value)
# Keep notification a no-op (its own session/IO is out of scope here).
monkeypatch.setattr(NotificationService, "send_ack_notification", AsyncMock())
@pytest.mark.asyncio
async def test_disabled_originate_creates_no_task(
db_session: AsyncSession, monkeypatch: pytest.MonkeyPatch
) -> None:
await _seed_project(db_session)
_enable(monkeypatch, self_heal_originate_enabled=False)
engine = SelfHealEngine(db_session, source=_FakeSource([_breach("ci:roboco")]))
obs = await engine.run_cycle()
assert len(obs) == ONE # detected + notified
assert await get_task_service(db_session).list_open_self_heal_tasks() == []
@pytest.mark.asyncio
async def test_originate_creates_pending_unassigned_task(
db_session: AsyncSession, monkeypatch: pytest.MonkeyPatch
) -> None:
await _seed_project(db_session)
_enable(monkeypatch)
engine = SelfHealEngine(db_session, source=_FakeSource([_breach("ci:roboco")]))
await engine.run_cycle()
open_tasks = await get_task_service(db_session).list_open_self_heal_tasks()
assert len(open_tasks) == ONE
task = open_tasks[0]
assert task.status == TaskStatus.PENDING
assert task.assigned_to is None # inert until the CEO Approve-&-Starts it
assert task.confirmed_by_human is False
assert task.team == Team.MAIN_PM
assert task.source == "self_heal"
assert task.acceptance_criteria # non-empty (AC-guardrail)
assert "self_heal_fp=" in (task.quick_context or "")
@pytest.mark.asyncio
async def test_dedupe_no_second_task_same_fingerprint(
db_session: AsyncSession, monkeypatch: pytest.MonkeyPatch
) -> None:
await _seed_project(db_session)
_enable(monkeypatch)
src = _FakeSource([_breach("ci:roboco")])
await SelfHealEngine(db_session, source=src).run_cycle()
await SelfHealEngine(db_session, source=src).run_cycle() # same fingerprint
open_tasks = await get_task_service(db_session).list_open_self_heal_tasks()
assert len(open_tasks) == ONE
@pytest.mark.asyncio
async def test_per_cycle_cap_limits_origination(
db_session: AsyncSession, monkeypatch: pytest.MonkeyPatch
) -> None:
await _seed_project(db_session)
_enable(monkeypatch, self_heal_max_per_cycle=1)
# Two distinct regressions in one cycle, cap = 1 → only one task opens.
src = _FakeSource([_breach("ci:roboco:a"), _breach("ci:roboco:b")])
await SelfHealEngine(db_session, source=src).run_cycle()
assert len(await get_task_service(db_session).list_open_self_heal_tasks()) == ONE
@pytest.mark.asyncio
async def test_open_task_cap_blocks_further_origination(
db_session: AsyncSession, monkeypatch: pytest.MonkeyPatch
) -> None:
await _seed_project(db_session)
_enable(monkeypatch, self_heal_max_open_tasks=1)
await SelfHealEngine(
db_session, source=_FakeSource([_breach("ci:roboco:a")])
).run_cycle()
# A different regression next cycle, but one task is already open → blocked.
await SelfHealEngine(
db_session, source=_FakeSource([_breach("ci:roboco:b")])
).run_cycle()
assert len(await get_task_service(db_session).list_open_self_heal_tasks()) == ONE
@pytest.mark.asyncio
async def test_unresolved_project_is_notify_only(
db_session: AsyncSession, monkeypatch: pytest.MonkeyPatch
) -> None:
await _seed_project(db_session)
_enable(monkeypatch)
ghost = TelemetrySample(
signal_name="ci:ghost",
value=1.0,
threshold=1.0,
window="latest_completed_run",
repo_hint="not-a-registered-project",
observed_at="",
raw_ref="",
)
# No crash, no task — the regression can't be tied to a repo.
await SelfHealEngine(db_session, source=_FakeSource([ghost])).run_cycle()
assert await get_task_service(db_session).list_open_self_heal_tasks() == []
@pytest.mark.asyncio
async def test_loop_never_starts_or_approves(
db_session: AsyncSession, monkeypatch: pytest.MonkeyPatch
) -> None:
await _seed_project(db_session)
_enable(monkeypatch)
approve = AsyncMock()
ceo_approve = AsyncMock()
monkeypatch.setattr(TaskService, "approve_and_start", approve)
monkeypatch.setattr(TaskService, "ceo_approve", ceo_approve)
await SelfHealEngine(
db_session, source=_FakeSource([_breach("ci:roboco")])
).run_cycle()
approve.assert_not_awaited()
ceo_approve.assert_not_awaited()
open_tasks = await get_task_service(db_session).list_open_self_heal_tasks()
assert len(open_tasks) == ONE
assert open_tasks[0].status == TaskStatus.PENDING # never advanced by the loop