fix(alerts): Add new alerts for submit latency and kudos accounting components

- Mapped new alerts for HordeSubmitLatencyHigh for both image and text components.
- Introduced internal alerts for kudos accounting, including silent applier and backlog conditions.
- Added application background job failure alerts to improve monitoring of internal processes.

Author: tazlin

packages/ai-horde-service-alerts/config/alert_component_map.yaml

@@ -37,6 +37,11 @@ mappings:
     label_match:
       horde_gentype: image
     status: degraded
+  - component: image
+    alertname: HordeSubmitLatencyHigh
+    label_match:
+      horde_gentype: image
+    status: degraded
 
   # ---- Text generation ----
   - component: text
@@ -47,6 +52,11 @@ mappings:
     label_match:
       horde_gentype: text
     status: degraded
+  - component: text
+    alertname: HordeSubmitLatencyHigh
+    label_match:
+      horde_gentype: text
+    status: degraded
 
   # ---- Worker coordination ----
   - component: workers
@@ -61,6 +71,30 @@ mappings:
     alertname: HordeWebhookFailureBurst
     status: degraded
 
+  # ---- Kudos accounting (internal component) ----
+  # A silent applier freezes balance materialization entirely; sustained lag or
+  # backlog means balances and counters are stale by that amount. Postings stay
+  # durable throughout, so these drive the internal component rather than any
+  # public pill. HordeKudosFloorCreationHigh is a finance anomaly signal, not a
+  # service-health one, and is deliberately unmapped.
+  - component: kudos
+    alertname: HordeKudosApplierSilent
+    status: down
+  - component: kudos
+    alertname: HordeKudosApplierLagHigh
+    status: degraded
+  - component: kudos
+    alertname: HordeKudosBacklogHigh
+    status: degraded
+  - component: kudos
+    alertname: HordeKudosReservationStuck
+    status: degraded
+
+  # ---- Application background jobs (internal component) ----
+  - component: app-jobs
+    alertname: HordeBackgroundJobFailures
+    status: degraded
+
   # alchemy: pill driven only by the external prober — the deployed rules
   # carry no alchemy/interrogation-specific alert, so there is nothing curated
   # to map. Add an entry here once a public-determinable alchemy alert exists.

packages/ai-horde-service-alerts/config/components.yaml

@@ -46,6 +46,18 @@ components:
     display_order: 60
 
   # ---- Internal-only infrastructure ----
+  - id: kudos
+    name: Kudos Accounting
+    description: Ledger applier folding balances and derived counters.
+    audience: internal
+    display_order: 100
+
+  - id: app-jobs
+    name: Application Background Jobs
+    description: Quorum-node periodic jobs (queue checks, stats, kudos applier).
+    audience: internal
+    display_order: 105
+
   - id: postgres
     name: PostgreSQL
     description: Primary application database.

packages/ai-horde-service-alerts/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "ai-horde-service-alerts"
-version = "0.7.0"
+version = "0.8.0"
 description = "Exposes AI Horde service alert signals as a FastAPI service."
 readme = "README.md"
 requires-python = ">=3.12"