Livrare LOT 1 - Didi

ai_platform/modules/dashboard/tests/test_monitoring.py

@@ -0,0 +1,118 @@
+"""Tests for the central AI monitoring routes (Val 1).
+
+External sources (service health, RabbitMQ, Prometheus) are mocked with respx /
+monkeypatch — no real infra needed.
+"""
+
+from __future__ import annotations
+
+import httpx
+import pytest
+import respx
+
+from dashboard.api.routes import monitoring
+from dashboard.config import DashboardSettings, SettingsCache
+
+
+def _settings(**overrides) -> DashboardSettings:
+    base = {
+        "database_url": "postgresql+asyncpg://u:p@localhost/db",
+        "external_url": "http://localhost:51300",
+    }
+    base.update(overrides)
+    s = DashboardSettings(**base)
+    SettingsCache.set(s)
+    return s
+
+
+# --- services aggregation -------------------------------------------------
+
+
+async def test_services_summary(monkeypatch):
+    _settings()
+
+    async def fake_probe(client, mid, url):
+        status = "down" if mid == "video" else "healthy"
+        return {"module": mid, "url": url, "status": status, "latency_ms": 1.0}
+
+    monkeypatch.setattr(monitoring, "_probe", fake_probe)
+    out = await monitoring.monitoring_services()
+
+    assert out["summary"]["total"] == len(monitoring._service_endpoints())
+    assert out["summary"]["down"] == 1
+    assert out["summary"]["healthy"] == out["summary"]["total"] - 1
+    assert any(s["module"] == "video" and s["status"] == "down" for s in out["services"])
+
+
+# --- RabbitMQ queues ------------------------------------------------------
+
+
+async def test_queues_disabled_when_unset():
+    _settings(rabbitmq_mgmt_url=None)
+    out = await monitoring.monitoring_queues()
+    assert out["enabled"] is False
+    assert out["queues"] == []
+
+
+@respx.mock
+async def test_queues_parsed_and_sorted():
+    _settings(rabbitmq_mgmt_url="http://rmq:15672")
+    respx.get("http://rmq:15672/api/queues").mock(
+        return_value=httpx.Response(
+            200,
+            json=[
+                {"name": "low", "vhost": "/", "messages": 2, "messages_ready": 2,
+                 "messages_unacknowledged": 0, "consumers": 1, "state": "running"},
+                {"name": "busy", "vhost": "/", "messages": 50, "messages_ready": 48,
+                 "messages_unacknowledged": 2, "consumers": 3, "state": "running"},
+            ],
+        )
+    )
+    out = await monitoring.monitoring_queues()
+    assert out["enabled"] is True
+    # sorted by messages desc
+    assert [q["name"] for q in out["queues"]] == ["busy", "low"]
+    assert out["queues"][0]["unacked"] == 2
+
+
+@respx.mock
+async def test_queues_failopen_on_error():
+    _settings(rabbitmq_mgmt_url="http://rmq:15672")
+    respx.get("http://rmq:15672/api/queues").mock(side_effect=httpx.ConnectError("x"))
+    out = await monitoring.monitoring_queues()
+    assert out["enabled"] is True
+    assert out["queues"] == []
+    assert "error" in out
+
+
+# --- Prometheus latency ---------------------------------------------------
+
+
+async def test_latency_disabled_when_unset():
+    _settings(prometheus_url=None)
+    out = await monitoring.monitoring_latency()
+    assert out["enabled"] is False
+
+
+@respx.mock
+async def test_latency_parsed_per_job():
+    _settings(prometheus_url="http://prom:9090")
+    # Same response for all three quantile queries → equal p50/p90/p99.
+    respx.get("http://prom:9090/api/v1/query").mock(
+        return_value=httpx.Response(
+            200,
+            json={
+                "data": {
+                    "result": [
+                        {"metric": {"job": "llm"}, "value": [0, "0.250"]},
+                        {"metric": {"job": "web"}, "value": [0, "1.500"]},
+                    ]
+                }
+            },
+        )
+    )
+    out = await monitoring.monitoring_latency()
+    assert out["enabled"] is True
+    by_job = {s["job"]: s for s in out["services"]}
+    assert by_job["llm"]["p50_ms"] == 250.0  # 0.250s → 250ms
+    assert by_job["web"]["p99_ms"] == 1500.0