fix: regenerate nornir inventory from DB at backend startup

groups.yaml only got rewritten as a side effect of group CRUD, while
hosts.yaml rewrites on switch CRUD. The inventory dir is gitignored and
not a Docker volume, so a fresh/recreated container has neither file
until someone happens to touch Groups — until then any switch sync
crashes with KeyError on a missing platform group (e.g. arista_eos).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Author: thangphan205

backend/app/crud/create_nornir.py

@@ -110,3 +110,27 @@ def create_groups(groups_db: any):
     group_dict_nornir["arista_eos"] = {"platform": "eos"}
 
     _write_yaml_atomic(f"{_INVENTORY_DIR}/groups.yaml", group_dict_nornir)
+
+
+def regenerate_inventory() -> None:
+    """Rebuild hosts.yaml and groups.yaml from the DB.
+
+    The inventory dir is gitignored and not a Docker volume, so every
+    container recreate starts with it empty. hosts.yaml is normally
+    rewritten as a side effect of switch CRUD, and groups.yaml as a side
+    effect of group CRUD — until either happens, InitNornir crashes with
+    a KeyError on a missing platform group. Call this at app startup so
+    a fresh container is never in that half-populated state.
+    """
+    from sqlmodel import Session, select
+
+    from app.core.db import engine
+    from app.models import Credential, Group, Switch
+
+    with Session(engine) as session:
+        switches_db = session.exec(
+            select(Switch, Credential).where(Switch.credential_id == Credential.id)
+        ).all()
+        create_hosts(switches_db)
+        groups_db = session.exec(select(Group)).all()
+        create_groups(groups_db)

backend/app/main.py

@@ -9,6 +9,7 @@
 from app.api.main import api_router
 from app.core.config import settings
 from app.core.scheduler import health_check_all_switches, scheduler, sync_all_switches
+from app.crud.create_nornir import regenerate_inventory
 
 
 def custom_generate_unique_id(route: APIRoute) -> str:
@@ -21,6 +22,7 @@ def custom_generate_unique_id(route: APIRoute) -> str:
 
 @asynccontextmanager
 async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
+    regenerate_inventory()
     scheduler.add_job(
         sync_all_switches,
         "interval",