refactor web pagination and add partitioned events, flap/timeout showcase agents

server/monlet_server/main.py

@@ -14,19 +14,43 @@ from .middleware.body_limit import BodyLimitMiddleware
 from .middleware.request_id import RequestIdMiddleware
 from .services.detector import run_detector
 from .services.notifier_worker import run_notifier_worker
+from .services.partitioning import ensure_event_partitions, run_partition_maintenance
 from .settings import get_settings
 
 log = get_logger("monlet.main")
 
 
+async def _ensure_partitions_with_retry(engine, attempts: int = 6, base_delay: float = 1.0) -> None:
+    last_exc: Exception | None = None
+    for i in range(attempts):
+        try:
+            await ensure_event_partitions(engine)
+            return
+        except Exception as exc:
+            last_exc = exc
+            log.warning("ensure_event_partitions_retry", attempt=i + 1, error=str(exc))
+            await asyncio.sleep(min(base_delay * (2**i), 10.0))
+    assert last_exc is not None
+    raise last_exc
+
+
 @asynccontextmanager
 async def lifespan(app: FastAPI):
     s = get_settings()
     configure_logging(s.log_level)
-    get_engine()
+    engine = get_engine()
     sm = get_sessionmaker()
+    # Mandatory: current month partition must exist before accepting writes.
+    # Retry briefly so a transient DB unavailability at startup (compose race)
+    # doesn't crash-loop the server.
+    await _ensure_partitions_with_retry(engine)
     stop_event = asyncio.Event()
     tasks: list[asyncio.Task] = []
+    # Partition maintenance is independent of detector — must run even when detector is off,
+    # otherwise events_future_partitions runs out and ingest fails.
+    tasks.append(
+        asyncio.create_task(run_partition_maintenance(sm, stop_event), name="partition_maintenance")
+    )
     if s.enable_detector:
         tasks.append(asyncio.create_task(run_detector(sm, stop_event), name="detector"))
     if s.enable_notifier_worker: