feat: add router health metrics to node_cache and node-guardian

- Add migration 042_node_cache_router_metrics.sql
- Node guardian now collects router health and sends in heartbeat
- City-service uses cached router_healthy from node_cache
- This allows NODE2 router status to be displayed correctly

migrations/042_node_cache_router_metrics.sql

@@ -0,0 +1,71 @@
+-- Migration: Add router health metrics to node_cache
+-- Purpose: Store router health status collected by node-guardian
+
+ALTER TABLE node_cache
+    ADD COLUMN IF NOT EXISTS router_healthy boolean DEFAULT false,
+    ADD COLUMN IF NOT EXISTS router_version text;
+
+COMMENT ON COLUMN node_cache.router_healthy IS 'Whether DAGI Router is healthy on this node';
+COMMENT ON COLUMN node_cache.router_version IS 'Version of DAGI Router on this node';
+
+-- Update fn_node_heartbeat to handle router metrics
+CREATE OR REPLACE FUNCTION fn_node_heartbeat(
+    p_node_id text,
+    p_metrics jsonb DEFAULT '{}'::jsonb
+) RETURNS jsonb AS $$
+DECLARE
+    v_result jsonb;
+    v_node_exists boolean;
+    v_swapper_state jsonb;
+BEGIN
+    -- Check if node exists
+    SELECT EXISTS(SELECT 1 FROM node_cache WHERE node_id = p_node_id) INTO v_node_exists;
+    
+    IF NOT v_node_exists THEN
+        RETURN jsonb_build_object(
+            'success', false,
+            'should_self_register', true,
+            'message', 'Node not found in cache'
+        );
+    END IF;
+    
+    -- Handle swapper_state - only update if provided and not empty
+    v_swapper_state := CASE 
+        WHEN p_metrics ? 'swapper_state' AND p_metrics->'swapper_state' != '{}'::jsonb 
+        THEN p_metrics->'swapper_state'
+        ELSE NULL
+    END;
+    
+    -- Update node_cache with metrics
+    UPDATE node_cache SET
+        last_heartbeat = NOW(),
+        status = 'online',
+        cpu_usage = COALESCE((p_metrics->>'cpu_usage')::float, cpu_usage),
+        gpu_vram_used = COALESCE((p_metrics->>'gpu_vram_used')::int, gpu_vram_used),
+        ram_used = COALESCE((p_metrics->>'ram_used')::bigint, ram_used),
+        disk_used = COALESCE((p_metrics->>'disk_used')::bigint, disk_used),
+        agent_count_router = COALESCE((p_metrics->>'agent_count_router')::int, agent_count_router),
+        agent_count_system = COALESCE((p_metrics->>'agent_count_system')::int, agent_count_system),
+        dagi_router_url = COALESCE(p_metrics->>'dagi_router_url', dagi_router_url),
+        -- Swapper metrics
+        swapper_healthy = COALESCE((p_metrics->>'swapper_healthy')::boolean, swapper_healthy),
+        swapper_models_loaded = COALESCE((p_metrics->>'swapper_models_loaded')::int, swapper_models_loaded),
+        swapper_models_total = COALESCE((p_metrics->>'swapper_models_total')::int, swapper_models_total),
+        swapper_state = COALESCE(v_swapper_state, swapper_state),
+        -- Router metrics
+        router_healthy = COALESCE((p_metrics->>'router_healthy')::boolean, router_healthy),
+        router_version = COALESCE(p_metrics->>'router_version', router_version),
+        -- Node-specific URLs (update if provided)
+        swapper_url = COALESCE(p_metrics->>'swapper_url', swapper_url),
+        router_url = COALESCE(p_metrics->>'router_url', router_url),
+        updated_at = NOW()
+    WHERE node_id = p_node_id;
+    
+    RETURN jsonb_build_object(
+        'success', true,
+        'node_id', p_node_id,
+        'timestamp', NOW()
+    );
+END;
+$$ LANGUAGE plpgsql;
+

scripts/node-guardian-loop.py

@@ -256,6 +256,20 @@ class NodeGuardian:
         except Exception as e:
             logger.warning(f"Swapper metrics collection failed: {e}")
         
+        # Collect Router Metrics using node-specific URL
+        metrics["router_healthy"] = False
+        metrics["router_version"] = None
+        try:
+            r = await self.client.get(f"{self.router_url}/health", timeout=3.0)
+            if r.status_code == 200:
+                data = r.json()
+                status = data.get("status", "").lower()
+                metrics["router_healthy"] = status in ("healthy", "ok")
+                metrics["router_version"] = data.get("version")
+                logger.debug(f"🔀 Router metrics: healthy={metrics['router_healthy']}, version={metrics['router_version']}")
+        except Exception as e:
+            logger.debug(f"Router health check failed: {e}")
+            
         return metrics
     
     async def run_health_check(self) -> Dict[str, Any]:

services/city-service/repo_city.py

@@ -3373,7 +3373,7 @@ async def get_node_endpoints(node_id: str) -> Dict[str, str]:
 
 async def get_node_metrics(node_id: str) -> Optional[Dict[str, Any]]:
     """
-    Отримати розширені метрики ноди (включаючи Swapper).
+    Отримати розширені метрики ноди (включаючи Swapper та Router).
     """
     pool = await get_pool()
     
@@ -3385,7 +3385,9 @@ async def get_node_metrics(node_id: str) -> Optional[Dict[str, Any]]:
             swapper_models_total,
             swapper_state,
             router_url,
-            swapper_url
+            swapper_url,
+            router_healthy,
+            router_version
         FROM node_cache
         WHERE node_id = $1
     """, node_id)

services/city-service/routes_city.py

@@ -4086,13 +4086,29 @@ async def get_node_swapper_detail(node_id: str):
 async def get_dagi_router_health(node_id: str):
     """
     Get DAGI Router health status for a node.
-    Always returns 200 with status="down" if router is unavailable.
-    Uses node-specific router_url from node_cache.
+    First checks node_cache for cached router_healthy status (from node-guardian).
+    Falls back to direct health check if node is local (NODE1).
     """
     import httpx
     import time
     
-    # Get router URL from database (node-specific)
+    # First, try to get cached router health from node_cache
+    # This is populated by node-guardian which has direct access to the router
+    try:
+        metrics = await repo_city.get_node_metrics(node_id)
+        if metrics and metrics.get("router_healthy") is not None:
+            return {
+                "node_id": node_id,
+                "status": "up" if metrics.get("router_healthy") else "down",
+                "version": metrics.get("router_version"),
+                "agent_count": 0,  # TODO: get from node_cache
+                "latency_ms": None,
+                "source": "node_cache"
+            }
+    except Exception as e:
+        logger.debug(f"Failed to get cached router health for {node_id}: {e}")
+    
+    # Fallback: try direct health check (only works for NODE1 which is local to city-service)
     endpoints = await repo_city.get_node_endpoints(node_id)
     base_url = endpoints.get("router_url")