fix(agent): persist signal incidents with tenant context
Some checks failed
CD Pipeline / workflow-shape (push) Successful in 0s
CD Pipeline / cancel-stale-cd (push) Has been skipped
CD Pipeline / build-and-deploy (push) Has been cancelled
CD Pipeline / post-deploy-checks (push) Has been cancelled
CD Pipeline / tests (push) Has been cancelled

This commit is contained in:
ogt
2026-07-10 21:09:31 +08:00
parent c004615220
commit 8aa8967c2b
5 changed files with 83 additions and 6 deletions

View File

@@ -1120,11 +1120,12 @@ class TimelineDBService:
risk_level: str | None = None,
approval_id: str | None = None,
incident_id: str | None = None,
project_id: str | None = None,
) -> dict[str, Any]:
"""
新增時間軸事件
"""
async with get_db_context() as db:
async with get_db_context(project_id) as db:
event = TimelineEvent(
event_type=event_type,
status=status,

View File

@@ -97,6 +97,8 @@ def _derive_incident_alert_metadata(incident: Incident) -> dict[str, Any]:
async def _add_signal_timeline_event(
incident: Incident,
metadata: dict[str, Any],
*,
project_id: str,
) -> None:
"""Best-effort timeline seed for incidents created outside Alertmanager."""
alertname = metadata.get("alertname")
@@ -115,6 +117,7 @@ async def _add_signal_timeline_event(
actor_role="signal_worker",
risk_level=getattr(incident.severity, "value", incident.severity),
incident_id=incident.incident_id,
project_id=project_id,
)
except Exception as exc:
logger.warning(
@@ -137,10 +140,13 @@ class IncidentDbAdapter:
注入到 lewooogo-brain 的 DualIncidentMemory
"""
def __init__(self, project_id: str = "awoooi") -> None:
self._project_id = project_id
async def load(self, incident_id: str) -> Incident | None:
"""從 PostgreSQL 載入 Incident"""
try:
async with get_db_context() as db:
async with get_db_context(self._project_id) as db:
from sqlalchemy import select
stmt = select(IncidentRecord).where(
IncidentRecord.incident_id == incident_id
@@ -161,7 +167,7 @@ class IncidentDbAdapter:
metadata = _derive_incident_alert_metadata(incident)
created = False
try:
async with get_db_context() as db:
async with get_db_context(self._project_id) as db:
from sqlalchemy import select
stmt = select(IncidentRecord).where(
@@ -191,6 +197,7 @@ class IncidentDbAdapter:
else:
record = IncidentRecord(
incident_id=incident.incident_id,
project_id=self._project_id,
status=incident.status.value,
severity=incident.severity.value,
signals=[
@@ -222,7 +229,11 @@ class IncidentDbAdapter:
created = True
if created:
await _add_signal_timeline_event(incident, metadata)
await _add_signal_timeline_event(
incident,
metadata,
project_id=self._project_id,
)
logger.debug("db_adapter_save_success", incident_id=incident.incident_id)
return True

View File

@@ -436,6 +436,7 @@ async def _add_observation_timeline(
actor_role="signal_worker",
risk_level=_severity_value(getattr(incident, "severity", None)),
incident_id=_incident_id(incident),
project_id=PROJECT_ID,
)
except Exception as exc:
logger.warning(
@@ -477,7 +478,10 @@ async def _ensure_signal_observation_mcp_runtime() -> None:
try:
from src.plugins.mcp.providers import register_all_providers
from src.plugins.mcp.registry import get_provider_registry
from src.services.mcp_tool_registry import get_mcp_tool_registry, init_mcp_tool_registry
from src.services.mcp_tool_registry import (
get_mcp_tool_registry,
init_mcp_tool_registry,
)
provider_registry = get_provider_registry()
if len(provider_registry) == 0:

View File

@@ -1,9 +1,16 @@
from __future__ import annotations
from contextlib import asynccontextmanager
from types import SimpleNamespace
import pytest
from src.models.incident import Severity
from src.services.incident_memory import _derive_incident_alert_metadata
from src.services import incident_memory as incident_memory_module
from src.services.incident_memory import (
IncidentDbAdapter,
_derive_incident_alert_metadata,
)
def test_signal_worker_incident_metadata_uses_signal_alert_name() -> None:
@@ -30,3 +37,31 @@ def test_signal_worker_incident_metadata_uses_signal_alert_name() -> None:
assert metadata["description"] == "29 ERROR log entries in last 5min"
assert metadata["actor"] == "journal"
@pytest.mark.asyncio
async def test_incident_db_adapter_uses_explicit_project_context(monkeypatch) -> None:
project_ids: list[str | None] = []
class FakeResult:
def scalar_one_or_none(self):
return None
class FakeDb:
async def execute(self, _statement):
return FakeResult()
@asynccontextmanager
async def fake_get_db_context(project_id=None):
project_ids.append(project_id)
yield FakeDb()
monkeypatch.setattr(
incident_memory_module,
"get_db_context",
fake_get_db_context,
)
result = await IncidentDbAdapter(project_id="awoooi").load("INC-NOT-FOUND")
assert result is None
assert project_ids == ["awoooi"]

View File

@@ -2,7 +2,10 @@ from __future__ import annotations
from types import SimpleNamespace
import pytest
from src.models.incident import Severity
from src.services import signal_observation_service as signal_observation_module
from src.services.signal_observation_service import (
build_signal_worker_provider_event_id,
enrich_incident_for_signal_observation,
@@ -72,3 +75,26 @@ def test_signal_worker_event_content_is_operator_readable() -> None:
assert "Incident: INC-20260518-TEST01" in content
assert "Alert: HostErrorLogFlood" in content
assert "Telegram: not sent" in content
@pytest.mark.asyncio
async def test_signal_observation_timeline_uses_explicit_project_context(monkeypatch) -> None:
calls: list[dict] = []
class FakeTimelineService:
async def add_event(self, **kwargs):
calls.append(kwargs)
monkeypatch.setattr(
"src.services.approval_db.get_timeline_service",
lambda: FakeTimelineService(),
)
await signal_observation_module._add_observation_timeline(
incident=_host_incident(),
signal_data={"alert_name": "HostErrorLogFlood"},
labels={"target": "ollama"},
raw_evidence_created=True,
)
assert calls[0]["project_id"] == "awoooi"