feat: LLM routing by tier (free→Ollama, pro→Timeweb)

- Add tier-based provider routing in llm-svc - free tier → Ollama (local qwen3.5:9b) - pro/business → Timeweb Cloud AI - Add /api/v1/embed endpoint for embeddings via Ollama - Update Ollama client: qwen3.5:9b default, remove auth - Add GenerateEmbedding() function for qwen3-embedding:0.6b - Add Ollama K8s deployment with GPU support (RTX 4060 Ti) - Add monitoring stack (Prometheus, Grafana, Alertmanager) - Add Grafana dashboards for LLM and security metrics - Update deploy.sh with monitoring and Ollama deployment Made-with: Cursor
2026-03-03 02:25:22 +03:00
parent 5ac082a7c6
commit 7a40ff629e
19 changed files with 1759 additions and 35 deletions
--- a/backend/deploy/k8s/api-gateway.yaml
+++ b/backend/deploy/k8s/api-gateway.yaml
@@ -16,6 +16,10 @@ spec:
    metadata:
      labels:
        app: api-gateway
+      annotations:
+        prometheus.io/scrape: "true"
+        prometheus.io/port: "3015"
+        prometheus.io/path: "/metrics"
    spec:
      containers:
      - name: api-gateway
--- a/backend/deploy/k8s/configmap.yaml
+++ b/backend/deploy/k8s/configmap.yaml
@@ -23,6 +23,10 @@ data:
  AUTH_SVC_URL: "http://auth-svc:3050"
  TRAVEL_SVC_URL: "http://travel-svc:3035"
  ADMIN_SVC_URL: "http://admin-svc:3040"
+  OLLAMA_BASE_URL: "http://ollama:11434"
+  OLLAMA_MODEL: "qwen3.5:9b"
+  OLLAMA_EMBEDDING_MODEL: "qwen3-embedding:0.6b"
+  OLLAMA_NUM_PARALLEL: "2"
  DEFAULT_LLM_MODEL: "${DEFAULT_LLM_MODEL}"
  DEFAULT_LLM_PROVIDER: "${DEFAULT_LLM_PROVIDER}"
  TIMEWEB_API_BASE_URL: "${TIMEWEB_API_BASE_URL}"
@@ -50,5 +54,6 @@ stringData:
  GEMINI_API_KEY: "${GEMINI_API_KEY}"
  JWT_SECRET: "${JWT_SECRET}"
  TIMEWEB_API_KEY: "${TIMEWEB_API_KEY}"
+  OLLAMA_API_TOKEN: "${OLLAMA_API_TOKEN}"
  POSTGRES_USER: "gooseek"
  POSTGRES_PASSWORD: "gooseek"
--- a/backend/deploy/k8s/deploy.sh
+++ b/backend/deploy/k8s/deploy.sh
@@ -22,6 +22,21 @@ if [ -f "$ENV_FILE" ]; then
  set +a
 fi

+# Check required secrets
+if [ -z "$OLLAMA_API_TOKEN" ]; then
+  echo "Warning: OLLAMA_API_TOKEN not set. Generating random token..."
+  OLLAMA_API_TOKEN=$(openssl rand -hex 32)
+  echo "OLLAMA_API_TOKEN=$OLLAMA_API_TOKEN" >> "$ENV_FILE"
+  echo "Token saved to .env"
+fi
+
+if [ -z "$GRAFANA_ADMIN_PASSWORD" ]; then
+  echo "Warning: GRAFANA_ADMIN_PASSWORD not set. Generating random password..."
+  GRAFANA_ADMIN_PASSWORD=$(openssl rand -base64 24)
+  echo "GRAFANA_ADMIN_PASSWORD=$GRAFANA_ADMIN_PASSWORD" >> "$ENV_FILE"
+  echo "Grafana password saved to .env"
+fi
+
 # Check kubectl
 if ! command -v kubectl &> /dev/null; then
    echo "Error: kubectl not found"
@@ -62,6 +77,9 @@ echo "=== Generating K8s manifests from .env ==="
 if command -v envsubst &> /dev/null && [ -f "$ENV_FILE" ]; then
    envsubst < "$SCRIPT_DIR/configmap.yaml" > "$SCRIPT_DIR/_generated_configmap.yaml"
    kubectl apply -f "$SCRIPT_DIR/_generated_configmap.yaml" -n gooseek
+
+    # Generate monitoring manifests
+    envsubst < "$SCRIPT_DIR/monitoring.yaml" > "$SCRIPT_DIR/_generated_monitoring.yaml"
 fi

 # Apply kustomization
@@ -70,6 +88,14 @@ echo "=== Applying K8s manifests ==="
 cd "$SCRIPT_DIR"
 kubectl apply -k .

+# Apply monitoring stack
+echo ""
+echo "=== Deploying Monitoring Stack ==="
+if [ -f "$SCRIPT_DIR/_generated_monitoring.yaml" ]; then
+    kubectl apply -f "$SCRIPT_DIR/_generated_monitoring.yaml"
+    kubectl apply -f "$SCRIPT_DIR/grafana-dashboards.yaml"
+fi
+
 # Rolling restart to pull new images
 echo ""
 echo "=== Rolling restart deployments ==="
@@ -79,11 +105,16 @@ kubectl -n gooseek rollout restart deployment/chat-svc
 kubectl -n gooseek rollout restart deployment/agent-svc
 kubectl -n gooseek rollout restart deployment/discover-svc
 kubectl -n gooseek rollout restart deployment/search-svc
+kubectl -n gooseek rollout restart deployment/llm-svc
 kubectl -n gooseek rollout restart deployment/learning-svc
 kubectl -n gooseek rollout restart deployment/medicine-svc
 kubectl -n gooseek rollout restart deployment/travel-svc
 kubectl -n gooseek rollout restart deployment/sandbox-svc

+# Ollama: не рестартим без необходимости (модели хранятся на PVC)
+# Модели загружаются один раз и сохраняются между деплоями
+# Для загрузки новых моделей: kubectl apply -f ollama-models.yaml
+
 # Wait for rollout
 echo ""
 echo "=== Waiting for rollouts ==="
@@ -100,7 +131,18 @@ kubectl -n gooseek get svc
 echo ""
 kubectl -n gooseek get ingress

+# Show monitoring status
+echo ""
+echo "=== Monitoring Status ==="
+kubectl -n monitoring get pods 2>/dev/null || echo "Monitoring namespace not ready yet"
+kubectl -n monitoring get ingress 2>/dev/null || true
+
 echo ""
 echo "=== Done ==="
-echo "API: https://api.gooseek.ru"
-echo "Web: https://gooseek.ru"
+echo "API:     https://api.gooseek.ru"
+echo "Web:     https://gooseek.ru"
+echo "Grafana: https://grafana.gooseek.ru"
+echo ""
+echo "Grafana credentials:"
+echo "  User: admin"
+echo "  Pass: (see GRAFANA_ADMIN_PASSWORD in .env)"
--- a/backend/deploy/k8s/grafana-dashboards.yaml
+++ b/backend/deploy/k8s/grafana-dashboards.yaml
@@ -0,0 +1,266 @@
+# Grafana Dashboards ConfigMap
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: grafana-dashboards
+  namespace: monitoring
+data:
+  security.json: |
+    {
+      "annotations": {
+        "list": []
+      },
+      "editable": true,
+      "fiscalYearStartMonth": 0,
+      "graphTooltip": 0,
+      "id": 1,
+      "links": [],
+      "liveNow": false,
+      "panels": [
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {
+            "defaults": {
+              "color": {"mode": "palette-classic"},
+              "mappings": [],
+              "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 80}]}
+            }
+          },
+          "gridPos": {"h": 4, "w": 6, "x": 0, "y": 0},
+          "id": 1,
+          "options": {"colorMode": "value", "graphMode": "area", "justifyMode": "auto", "orientation": "auto", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto"},
+          "pluginVersion": "10.3.3",
+          "targets": [{"datasource": {"type": "prometheus", "uid": "prometheus"}, "expr": "sum(rate(llm_unauthorized_requests_total[5m]))", "legendFormat": "Unauthorized/sec", "refId": "A"}],
+          "title": "🚨 Неавторизованные запросы",
+          "type": "stat"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "yellow", "value": 10}, {"color": "red", "value": 50}]}}},
+          "gridPos": {"h": 4, "w": 6, "x": 6, "y": 0},
+          "id": 2,
+          "options": {"colorMode": "value", "graphMode": "area", "justifyMode": "auto", "orientation": "auto", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto"},
+          "targets": [{"expr": "sum(rate(llm_free_tier_limit_exceeded_total[5m]))", "legendFormat": "Limit exceeded/sec", "refId": "A"}],
+          "title": "⚠️ Превышение лимитов free",
+          "type": "stat"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}]}}},
+          "gridPos": {"h": 4, "w": 6, "x": 12, "y": 0},
+          "id": 3,
+          "options": {"colorMode": "value", "graphMode": "area", "justifyMode": "auto", "orientation": "auto", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto"},
+          "targets": [{"expr": "sum(rate(llm_requests_total[5m]))", "legendFormat": "Requests/sec", "refId": "A"}],
+          "title": "📊 LLM запросы/сек",
+          "type": "stat"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 10}]}}},
+          "gridPos": {"h": 4, "w": 6, "x": 18, "y": 0},
+          "id": 4,
+          "options": {"colorMode": "value", "graphMode": "area", "justifyMode": "auto", "orientation": "auto", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto"},
+          "targets": [{"expr": "sum(rate(llm_errors_total[5m]))", "legendFormat": "Errors/sec", "refId": "A"}],
+          "title": "❌ Ошибки LLM",
+          "type": "stat"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "custom": {"axisBorderShow": false, "axisCenteredZero": false, "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, "drawStyle": "line", "fillOpacity": 10, "gradientMode": "none", "hideFrom": {"legend": false, "tooltip": false, "viz": false}, "insertNulls": false, "lineInterpolation": "linear", "lineWidth": 1, "pointSize": 5, "scaleDistribution": {"type": "linear"}, "showPoints": "auto", "spanNulls": false, "stacking": {"group": "A", "mode": "none"}, "thresholdsStyle": {"mode": "off"}}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 80}]}}},
+          "gridPos": {"h": 8, "w": 12, "x": 0, "y": 4},
+          "id": 5,
+          "options": {"legend": {"calcs": [], "displayMode": "list", "placement": "bottom", "showLegend": true}, "tooltip": {"mode": "single", "sort": "none"}},
+          "targets": [
+            {"expr": "sum by (reason) (rate(llm_unauthorized_requests_total[5m]))", "legendFormat": "{{reason}}", "refId": "A"},
+            {"expr": "sum by (limit_type) (rate(llm_free_tier_limit_exceeded_total[5m]))", "legendFormat": "limit: {{limit_type}}", "refId": "B"}
+          ],
+          "title": "🔐 События безопасности",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "custom": {"axisBorderShow": false, "axisCenteredZero": false, "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, "drawStyle": "line", "fillOpacity": 10, "gradientMode": "none", "hideFrom": {"legend": false, "tooltip": false, "viz": false}, "insertNulls": false, "lineInterpolation": "linear", "lineWidth": 1, "pointSize": 5, "scaleDistribution": {"type": "linear"}, "showPoints": "auto", "spanNulls": false, "stacking": {"group": "A", "mode": "none"}, "thresholdsStyle": {"mode": "off"}}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 80}]}}},
+          "gridPos": {"h": 8, "w": 12, "x": 12, "y": 4},
+          "id": 6,
+          "options": {"legend": {"calcs": [], "displayMode": "list", "placement": "bottom", "showLegend": true}, "tooltip": {"mode": "single", "sort": "none"}},
+          "targets": [
+            {"expr": "sum by (provider) (rate(llm_requests_total[5m]))", "legendFormat": "{{provider}}", "refId": "A"}
+          ],
+          "title": "📈 Запросы по провайдерам",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "custom": {"axisBorderShow": false, "axisCenteredZero": false, "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, "drawStyle": "line", "fillOpacity": 10, "gradientMode": "none", "hideFrom": {"legend": false, "tooltip": false, "viz": false}, "insertNulls": false, "lineInterpolation": "linear", "lineWidth": 1, "pointSize": 5, "scaleDistribution": {"type": "linear"}, "showPoints": "auto", "spanNulls": false, "stacking": {"group": "A", "mode": "none"}, "thresholdsStyle": {"mode": "off"}}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 80}]}}},
+          "gridPos": {"h": 8, "w": 12, "x": 0, "y": 12},
+          "id": 7,
+          "options": {"legend": {"calcs": [], "displayMode": "list", "placement": "bottom", "showLegend": true}, "tooltip": {"mode": "single", "sort": "none"}},
+          "targets": [
+            {"expr": "sum by (tier) (rate(llm_tokens_used_total[5m]))", "legendFormat": "{{tier}}", "refId": "A"}
+          ],
+          "title": "🎫 Токены по тарифам",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "custom": {"axisBorderShow": false, "axisCenteredZero": false, "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, "drawStyle": "line", "fillOpacity": 10, "gradientMode": "none", "hideFrom": {"legend": false, "tooltip": false, "viz": false}, "insertNulls": false, "lineInterpolation": "linear", "lineWidth": 1, "pointSize": 5, "scaleDistribution": {"type": "linear"}, "showPoints": "auto", "spanNulls": false, "stacking": {"group": "A", "mode": "none"}, "thresholdsStyle": {"mode": "off"}}, "mappings": [], "unit": "s", "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 80}]}}},
+          "gridPos": {"h": 8, "w": 12, "x": 12, "y": 12},
+          "id": 8,
+          "options": {"legend": {"calcs": [], "displayMode": "list", "placement": "bottom", "showLegend": true}, "tooltip": {"mode": "single", "sort": "none"}},
+          "targets": [
+            {"expr": "histogram_quantile(0.95, sum by (le, provider) (rate(llm_request_latency_seconds_bucket[5m])))", "legendFormat": "p95 {{provider}}", "refId": "A"},
+            {"expr": "histogram_quantile(0.50, sum by (le, provider) (rate(llm_request_latency_seconds_bucket[5m])))", "legendFormat": "p50 {{provider}}", "refId": "B"}
+          ],
+          "title": "⏱️ Latency LLM (p50, p95)",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "thresholds"}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "yellow", "value": 5}, {"color": "red", "value": 20}]}}},
+          "gridPos": {"h": 8, "w": 24, "x": 0, "y": 20},
+          "id": 9,
+          "options": {"displayMode": "lcd", "minVizHeight": 10, "minVizWidth": 0, "orientation": "horizontal", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "showUnfilled": true, "valueMode": "color"},
+          "targets": [
+            {"expr": "topk(10, sum by (user_id) (rate(llm_free_tier_limit_exceeded_total[1h])))", "legendFormat": "{{user_id}}", "refId": "A"}
+          ],
+          "title": "🚫 Top-10 пользователей превышающих лимиты (за час)",
+          "type": "bargauge"
+        }
+      ],
+      "refresh": "10s",
+      "schemaVersion": 39,
+      "tags": ["security", "llm"],
+      "templating": {"list": []},
+      "time": {"from": "now-1h", "to": "now"},
+      "timepicker": {},
+      "timezone": "",
+      "title": "🔐 Security & LLM Monitoring",
+      "uid": "security-llm",
+      "version": 1,
+      "weekStart": ""
+    }
+
+  system-health.json: |
+    {
+      "annotations": {"list": []},
+      "editable": true,
+      "fiscalYearStartMonth": 0,
+      "graphTooltip": 0,
+      "id": 2,
+      "links": [],
+      "liveNow": false,
+      "panels": [
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "thresholds"}, "mappings": [], "max": 100, "min": 0, "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "yellow", "value": 60}, {"color": "red", "value": 80}]}, "unit": "percent"}},
+          "gridPos": {"h": 6, "w": 6, "x": 0, "y": 0},
+          "id": 1,
+          "options": {"orientation": "auto", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "showThresholdLabels": false, "showThresholdMarkers": true},
+          "targets": [{"expr": "100 - (avg(rate(node_cpu_seconds_total{mode=\"idle\"}[5m])) * 100)", "legendFormat": "CPU Usage", "refId": "A"}],
+          "title": "💻 CPU",
+          "type": "gauge"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "thresholds"}, "mappings": [], "max": 100, "min": 0, "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "yellow", "value": 70}, {"color": "red", "value": 85}]}, "unit": "percent"}},
+          "gridPos": {"h": 6, "w": 6, "x": 6, "y": 0},
+          "id": 2,
+          "options": {"orientation": "auto", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "showThresholdLabels": false, "showThresholdMarkers": true},
+          "targets": [{"expr": "(1 - (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes)) * 100", "legendFormat": "Memory Usage", "refId": "A"}],
+          "title": "🧠 Memory",
+          "type": "gauge"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "thresholds"}, "mappings": [], "max": 100, "min": 0, "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "yellow", "value": 70}, {"color": "red", "value": 85}]}, "unit": "percent"}},
+          "gridPos": {"h": 6, "w": 6, "x": 12, "y": 0},
+          "id": 3,
+          "options": {"orientation": "auto", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "showThresholdLabels": false, "showThresholdMarkers": true},
+          "targets": [{"expr": "(1 - (node_filesystem_avail_bytes{mountpoint=\"/\"} / node_filesystem_size_bytes{mountpoint=\"/\"})) * 100", "legendFormat": "Disk Usage", "refId": "A"}],
+          "title": "💾 Disk",
+          "type": "gauge"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}]}}},
+          "gridPos": {"h": 6, "w": 6, "x": 18, "y": 0},
+          "id": 4,
+          "options": {"colorMode": "value", "graphMode": "area", "justifyMode": "auto", "orientation": "auto", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto"},
+          "targets": [{"expr": "sum(rate(http_requests_total[5m]))", "legendFormat": "Requests/sec", "refId": "A"}],
+          "title": "🌐 HTTP запросы/сек",
+          "type": "stat"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "custom": {"axisBorderShow": false, "axisCenteredZero": false, "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, "drawStyle": "line", "fillOpacity": 10, "gradientMode": "none", "hideFrom": {"legend": false, "tooltip": false, "viz": false}, "insertNulls": false, "lineInterpolation": "linear", "lineWidth": 1, "pointSize": 5, "scaleDistribution": {"type": "linear"}, "showPoints": "auto", "spanNulls": false, "stacking": {"group": "A", "mode": "none"}, "thresholdsStyle": {"mode": "off"}}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 80}]}}},
+          "gridPos": {"h": 8, "w": 12, "x": 0, "y": 6},
+          "id": 5,
+          "options": {"legend": {"calcs": [], "displayMode": "list", "placement": "bottom", "showLegend": true}, "tooltip": {"mode": "single", "sort": "none"}},
+          "targets": [
+            {"expr": "sum by (service) (rate(http_requests_total[5m]))", "legendFormat": "{{service}}", "refId": "A"}
+          ],
+          "title": "📊 Запросы по сервисам",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "custom": {"axisBorderShow": false, "axisCenteredZero": false, "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, "drawStyle": "line", "fillOpacity": 10, "gradientMode": "none", "hideFrom": {"legend": false, "tooltip": false, "viz": false}, "insertNulls": false, "lineInterpolation": "linear", "lineWidth": 1, "pointSize": 5, "scaleDistribution": {"type": "linear"}, "showPoints": "auto", "spanNulls": false, "stacking": {"group": "A", "mode": "none"}, "thresholdsStyle": {"mode": "off"}}, "mappings": [], "unit": "s", "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 80}]}}},
+          "gridPos": {"h": 8, "w": 12, "x": 12, "y": 6},
+          "id": 6,
+          "options": {"legend": {"calcs": [], "displayMode": "list", "placement": "bottom", "showLegend": true}, "tooltip": {"mode": "single", "sort": "none"}},
+          "targets": [
+            {"expr": "histogram_quantile(0.95, sum by (le, service) (rate(http_request_duration_seconds_bucket[5m])))", "legendFormat": "p95 {{service}}", "refId": "A"}
+          ],
+          "title": "⏱️ Latency p95 по сервисам",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "custom": {"axisBorderShow": false, "axisCenteredZero": false, "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, "drawStyle": "line", "fillOpacity": 10, "gradientMode": "none", "hideFrom": {"legend": false, "tooltip": false, "viz": false}, "insertNulls": false, "lineInterpolation": "linear", "lineWidth": 1, "pointSize": 5, "scaleDistribution": {"type": "linear"}, "showPoints": "auto", "spanNulls": false, "stacking": {"group": "A", "mode": "none"}, "thresholdsStyle": {"mode": "off"}}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}, {"color": "red", "value": 80}]}}},
+          "gridPos": {"h": 8, "w": 12, "x": 0, "y": 14},
+          "id": 7,
+          "options": {"legend": {"calcs": [], "displayMode": "list", "placement": "bottom", "showLegend": true}, "tooltip": {"mode": "single", "sort": "none"}},
+          "targets": [
+            {"expr": "sum by (service) (rate(http_requests_total{status=~\"5..\"}[5m]))", "legendFormat": "5xx {{service}}", "refId": "A"},
+            {"expr": "sum by (service) (rate(http_requests_total{status=~\"4..\"}[5m]))", "legendFormat": "4xx {{service}}", "refId": "B"}
+          ],
+          "title": "❌ Ошибки HTTP",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "palette-classic"}, "custom": {"axisBorderShow": false, "axisCenteredZero": false, "axisColorMode": "text", "axisLabel": "", "axisPlacement": "auto", "barAlignment": 0, "drawStyle": "line", "fillOpacity": 10, "gradientMode": "none", "hideFrom": {"legend": false, "tooltip": false, "viz": false}, "insertNulls": false, "lineInterpolation": "linear", "lineWidth": 1, "pointSize": 5, "scaleDistribution": {"type": "linear"}, "showPoints": "auto", "spanNulls": false, "stacking": {"group": "A", "mode": "none"}, "thresholdsStyle": {"mode": "off"}}, "mappings": [], "thresholds": {"mode": "absolute", "steps": [{"color": "green", "value": null}]}}},
+          "gridPos": {"h": 8, "w": 12, "x": 12, "y": 14},
+          "id": 8,
+          "options": {"legend": {"calcs": [], "displayMode": "list", "placement": "bottom", "showLegend": true}, "tooltip": {"mode": "single", "sort": "none"}},
+          "targets": [
+            {"expr": "sum by (service) (http_requests_in_flight)", "legendFormat": "{{service}}", "refId": "A"}
+          ],
+          "title": "🔄 Активные запросы",
+          "type": "timeseries"
+        },
+        {
+          "datasource": {"type": "prometheus", "uid": "prometheus"},
+          "fieldConfig": {"defaults": {"color": {"mode": "thresholds"}, "mappings": [{"options": {"0": {"color": "red", "index": 0, "text": "DOWN"}, "1": {"color": "green", "index": 1, "text": "UP"}}, "type": "value"}], "thresholds": {"mode": "absolute", "steps": [{"color": "red", "value": null}, {"color": "green", "value": 1}]}}},
+          "gridPos": {"h": 6, "w": 24, "x": 0, "y": 22},
+          "id": 9,
+          "options": {"colorMode": "background", "graphMode": "none", "justifyMode": "auto", "orientation": "horizontal", "reduceOptions": {"calcs": ["lastNotNull"], "fields": "", "values": false}, "textMode": "auto"},
+          "targets": [
+            {"expr": "up{job=\"gooseek-services\"}", "legendFormat": "{{service}}", "refId": "A"}
+          ],
+          "title": "🏥 Статус сервисов",
+          "type": "stat"
+        }
+      ],
+      "refresh": "10s",
+      "schemaVersion": 39,
+      "tags": ["system", "health"],
+      "templating": {"list": []},
+      "time": {"from": "now-1h", "to": "now"},
+      "timepicker": {},
+      "timezone": "",
+      "title": "🏥 System Health",
+      "uid": "system-health",
+      "version": 1,
+      "weekStart": ""
+    }
--- a/backend/deploy/k8s/kustomization.yaml
+++ b/backend/deploy/k8s/kustomization.yaml
@@ -24,6 +24,7 @@ resources:
  - travel-svc.yaml
  - sandbox-svc.yaml
  - opensandbox.yaml
+  - ollama.yaml
  - ingress.yaml

 labels:
--- a/backend/deploy/k8s/llm-svc.yaml
+++ b/backend/deploy/k8s/llm-svc.yaml
@@ -16,6 +16,10 @@ spec:
    metadata:
      labels:
        app: llm-svc
+      annotations:
+        prometheus.io/scrape: "true"
+        prometheus.io/port: "3020"
+        prometheus.io/path: "/metrics"
    spec:
      containers:
      - name: llm-svc
--- a/backend/deploy/k8s/monitoring.yaml
+++ b/backend/deploy/k8s/monitoring.yaml
@@ -0,0 +1,674 @@
+# Monitoring Stack: Prometheus + Grafana + AlertManager
+# Отслеживание безопасности, ресурсов, здоровья системы
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: monitoring
+  labels:
+    app.kubernetes.io/name: monitoring
+---
+# Prometheus ConfigMap
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: prometheus-config
+  namespace: monitoring
+data:
+  prometheus.yml: |
+    global:
+      scrape_interval: 15s
+      evaluation_interval: 15s
+
+    alerting:
+      alertmanagers:
+        - static_configs:
+            - targets:
+              - alertmanager:9093
+
+    rule_files:
+      - /etc/prometheus/rules/*.yml
+
+    scrape_configs:
+      # Prometheus self-monitoring
+      - job_name: 'prometheus'
+        static_configs:
+          - targets: ['localhost:9090']
+
+      # Kubernetes API server
+      - job_name: 'kubernetes-apiservers'
+        kubernetes_sd_configs:
+          - role: endpoints
+        scheme: https
+        tls_config:
+          ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
+        bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token
+        relabel_configs:
+          - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_service_name, __meta_kubernetes_endpoint_port_name]
+            action: keep
+            regex: default;kubernetes;https
+
+      # Kubernetes nodes
+      - job_name: 'kubernetes-nodes'
+        scheme: https
+        tls_config:
+          ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
+        bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token
+        kubernetes_sd_configs:
+          - role: node
+        relabel_configs:
+          - action: labelmap
+            regex: __meta_kubernetes_node_label_(.+)
+          - target_label: __address__
+            replacement: kubernetes.default.svc:443
+          - source_labels: [__meta_kubernetes_node_name]
+            regex: (.+)
+            target_label: __metrics_path__
+            replacement: /api/v1/nodes/${1}/proxy/metrics
+
+      # Kubernetes pods (auto-discovery)
+      - job_name: 'kubernetes-pods'
+        kubernetes_sd_configs:
+          - role: pod
+        relabel_configs:
+          - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_scrape]
+            action: keep
+            regex: true
+          - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_path]
+            action: replace
+            target_label: __metrics_path__
+            regex: (.+)
+          - source_labels: [__address__, __meta_kubernetes_pod_annotation_prometheus_io_port]
+            action: replace
+            regex: ([^:]+)(?::\d+)?;(\d+)
+            replacement: $1:$2
+            target_label: __address__
+          - action: labelmap
+            regex: __meta_kubernetes_pod_label_(.+)
+          - source_labels: [__meta_kubernetes_namespace]
+            action: replace
+            target_label: kubernetes_namespace
+          - source_labels: [__meta_kubernetes_pod_name]
+            action: replace
+            target_label: kubernetes_pod_name
+
+      # GooSeek services (direct)
+      - job_name: 'gooseek-services'
+        static_configs:
+          - targets:
+            - api-gateway.gooseek.svc:3015
+            - llm-svc.gooseek.svc:3020
+            - agent-svc.gooseek.svc:3018
+            - chat-svc.gooseek.svc:3005
+            - search-svc.gooseek.svc:3001
+            - learning-svc.gooseek.svc:3034
+            - travel-svc.gooseek.svc:3035
+            - medicine-svc.gooseek.svc:3037
+        metrics_path: /metrics
+        relabel_configs:
+          - source_labels: [__address__]
+            regex: (.+)\.gooseek\.svc:(\d+)
+            replacement: $1
+            target_label: service
+
+  alerts.yml: |
+    groups:
+      - name: security
+        rules:
+          - alert: HighUnauthorizedRequests
+            expr: rate(llm_unauthorized_requests_total[5m]) > 10
+            for: 2m
+            labels:
+              severity: critical
+            annotations:
+              summary: High unauthorized LLM requests
+              description: More than 10 unauthorized requests per second
+
+          - alert: FreeTierAbuse
+            expr: rate(llm_free_tier_limit_exceeded_total[5m]) > 5
+            for: 5m
+            labels:
+              severity: warning
+            annotations:
+              summary: Free tier limit exceeded
+              description: Users are exceeding LLM free tier limits
+
+          - alert: SuspiciousActivity
+            expr: sum by (client_ip) (rate(http_requests_total[5m])) > 100
+            for: 5m
+            labels:
+              severity: warning
+            annotations:
+              summary: Suspicious activity detected
+              description: High request rate from single IP
+
+      - name: resources
+        rules:
+          - alert: HighCPUUsage
+            expr: 100 - (avg by(instance) (rate(node_cpu_seconds_total{mode="idle"}[5m])) * 100) > 80
+            for: 10m
+            labels:
+              severity: warning
+            annotations:
+              summary: High CPU usage
+              description: CPU usage is above 80 percent
+
+          - alert: LowMemory
+            expr: (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes) * 100 < 20
+            for: 5m
+            labels:
+              severity: critical
+            annotations:
+              summary: Low memory available
+              description: Less than 20 percent memory available
+
+          - alert: DiskSpaceLow
+            expr: (node_filesystem_avail_bytes / node_filesystem_size_bytes) * 100 < 15
+            for: 5m
+            labels:
+              severity: critical
+            annotations:
+              summary: Low disk space
+              description: Less than 15 percent disk space available
+
+      - name: availability
+        rules:
+          - alert: ServiceDown
+            expr: up{job="gooseek-services"} == 0
+            for: 2m
+            labels:
+              severity: critical
+            annotations:
+              summary: Service is down
+              description: A GooSeek service is not responding
+
+          - alert: HighLatency
+            expr: histogram_quantile(0.95, rate(http_request_duration_seconds_bucket[5m])) > 5
+            for: 5m
+            labels:
+              severity: warning
+            annotations:
+              summary: High latency detected
+              description: P95 latency is above 5 seconds
+
+          - alert: HighErrorRate
+            expr: rate(http_requests_total{status=~"5.."}[5m]) / rate(http_requests_total[5m]) > 0.05
+            for: 5m
+            labels:
+              severity: critical
+            annotations:
+              summary: High error rate
+              description: Error rate is above 5 percent
+---
+# Prometheus RBAC
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: prometheus
+  namespace: monitoring
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: prometheus
+rules:
+  - apiGroups: [""]
+    resources:
+      - nodes
+      - nodes/proxy
+      - services
+      - endpoints
+      - pods
+    verbs: ["get", "list", "watch"]
+  - apiGroups: ["extensions"]
+    resources:
+      - ingresses
+    verbs: ["get", "list", "watch"]
+  - nonResourceURLs: ["/metrics"]
+    verbs: ["get"]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: prometheus
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: prometheus
+subjects:
+  - kind: ServiceAccount
+    name: prometheus
+    namespace: monitoring
+---
+# Prometheus Deployment
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: prometheus
+  namespace: monitoring
+  labels:
+    app: prometheus
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: prometheus
+  template:
+    metadata:
+      labels:
+        app: prometheus
+    spec:
+      serviceAccountName: prometheus
+      containers:
+        - name: prometheus
+          image: prom/prometheus:v2.50.0
+          args:
+            - "--config.file=/etc/prometheus/prometheus.yml"
+            - "--storage.tsdb.path=/prometheus"
+            - "--storage.tsdb.retention.time=30d"
+            - "--web.enable-lifecycle"
+          ports:
+            - containerPort: 9090
+          volumeMounts:
+            - name: config
+              mountPath: /etc/prometheus
+            - name: rules
+              mountPath: /etc/prometheus/rules
+            - name: data
+              mountPath: /prometheus
+          resources:
+            requests:
+              cpu: 200m
+              memory: 512Mi
+            limits:
+              cpu: 1000m
+              memory: 2Gi
+      volumes:
+        - name: config
+          configMap:
+            name: prometheus-config
+        - name: rules
+          configMap:
+            name: prometheus-config
+            items:
+              - key: alerts.yml
+                path: alerts.yml
+        - name: data
+          persistentVolumeClaim:
+            claimName: prometheus-pvc
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: prometheus-pvc
+  namespace: monitoring
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 20Gi
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: prometheus
+  namespace: monitoring
+spec:
+  type: ClusterIP
+  selector:
+    app: prometheus
+  ports:
+    - port: 9090
+      targetPort: 9090
+---
+# AlertManager ConfigMap
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: alertmanager-config
+  namespace: monitoring
+data:
+  alertmanager.yml: |
+    global:
+      resolve_timeout: 5m
+
+    route:
+      group_by: ['alertname', 'severity']
+      group_wait: 30s
+      group_interval: 5m
+      repeat_interval: 4h
+      receiver: 'telegram'
+      routes:
+        - match:
+            severity: critical
+          receiver: 'telegram'
+          continue: true
+
+    receivers:
+      - name: 'telegram'
+        webhook_configs:
+          - url: 'http://api-gateway.gooseek.svc:3015/api/v1/alerts/webhook'
+            send_resolved: true
+
+    inhibit_rules:
+      - source_match:
+          severity: 'critical'
+        target_match:
+          severity: 'warning'
+        equal: ['alertname']
+---
+# AlertManager Deployment
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: alertmanager
+  namespace: monitoring
+  labels:
+    app: alertmanager
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: alertmanager
+  template:
+    metadata:
+      labels:
+        app: alertmanager
+    spec:
+      containers:
+        - name: alertmanager
+          image: prom/alertmanager:v0.27.0
+          args:
+            - "--config.file=/etc/alertmanager/alertmanager.yml"
+            - "--storage.path=/alertmanager"
+          ports:
+            - containerPort: 9093
+          volumeMounts:
+            - name: config
+              mountPath: /etc/alertmanager
+            - name: data
+              mountPath: /alertmanager
+          resources:
+            requests:
+              cpu: 50m
+              memory: 64Mi
+            limits:
+              cpu: 200m
+              memory: 256Mi
+      volumes:
+        - name: config
+          configMap:
+            name: alertmanager-config
+        - name: data
+          emptyDir: {}
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: alertmanager
+  namespace: monitoring
+spec:
+  type: ClusterIP
+  selector:
+    app: alertmanager
+  ports:
+    - port: 9093
+      targetPort: 9093
+---
+# Grafana ConfigMap
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: grafana-config
+  namespace: monitoring
+data:
+  grafana.ini: |
+    [server]
+    root_url = https://grafana.gooseek.ru
+    
+    [security]
+    admin_user = admin
+    admin_password = ${GRAFANA_ADMIN_PASSWORD}
+    
+    [auth.anonymous]
+    enabled = false
+    
+    [dashboards]
+    default_home_dashboard_path = /var/lib/grafana/dashboards/security.json
+
+  datasources.yml: |
+    apiVersion: 1
+    datasources:
+      - name: Prometheus
+        type: prometheus
+        access: proxy
+        url: http://prometheus:9090
+        isDefault: true
+        editable: false
+---
+# Grafana Deployment
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: grafana
+  namespace: monitoring
+  labels:
+    app: grafana
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: grafana
+  template:
+    metadata:
+      labels:
+        app: grafana
+    spec:
+      containers:
+        - name: grafana
+          image: grafana/grafana:10.3.3
+          ports:
+            - containerPort: 3000
+          env:
+            - name: GF_SECURITY_ADMIN_PASSWORD
+              valueFrom:
+                secretKeyRef:
+                  name: grafana-secrets
+                  key: admin-password
+            - name: GF_INSTALL_PLUGINS
+              value: "grafana-piechart-panel,grafana-clock-panel"
+          volumeMounts:
+            - name: config
+              mountPath: /etc/grafana/grafana.ini
+              subPath: grafana.ini
+            - name: datasources
+              mountPath: /etc/grafana/provisioning/datasources
+            - name: dashboards-config
+              mountPath: /etc/grafana/provisioning/dashboards
+            - name: dashboards
+              mountPath: /var/lib/grafana/dashboards
+            - name: data
+              mountPath: /var/lib/grafana
+          resources:
+            requests:
+              cpu: 100m
+              memory: 256Mi
+            limits:
+              cpu: 500m
+              memory: 512Mi
+      volumes:
+        - name: config
+          configMap:
+            name: grafana-config
+        - name: datasources
+          configMap:
+            name: grafana-config
+            items:
+              - key: datasources.yml
+                path: datasources.yml
+        - name: dashboards-config
+          configMap:
+            name: grafana-dashboards-config
+        - name: dashboards
+          configMap:
+            name: grafana-dashboards
+        - name: data
+          persistentVolumeClaim:
+            claimName: grafana-pvc
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: grafana-pvc
+  namespace: monitoring
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 5Gi
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: grafana-secrets
+  namespace: monitoring
+type: Opaque
+stringData:
+  admin-password: "${GRAFANA_ADMIN_PASSWORD}"
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: grafana
+  namespace: monitoring
+spec:
+  type: ClusterIP
+  selector:
+    app: grafana
+  ports:
+    - port: 3000
+      targetPort: 3000
+---
+# Grafana Dashboards Config
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: grafana-dashboards-config
+  namespace: monitoring
+data:
+  dashboards.yml: |
+    apiVersion: 1
+    providers:
+      - name: 'default'
+        orgId: 1
+        folder: ''
+        type: file
+        disableDeletion: false
+        editable: true
+        options:
+          path: /var/lib/grafana/dashboards
+---
+# Grafana Ingress
+apiVersion: networking.k8s.io/v1
+kind: Ingress
+metadata:
+  name: grafana-ingress
+  namespace: monitoring
+  annotations:
+    nginx.ingress.kubernetes.io/ssl-redirect: "true"
+    cert-manager.io/cluster-issuer: "letsencrypt-prod"
+spec:
+  ingressClassName: nginx
+  tls:
+    - hosts:
+        - grafana.gooseek.ru
+      secretName: grafana-tls
+  rules:
+    - host: grafana.gooseek.ru
+      http:
+        paths:
+          - path: /
+            pathType: Prefix
+            backend:
+              service:
+                name: grafana
+                port:
+                  number: 3000
+---
+# Node Exporter DaemonSet (для метрик хоста)
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: node-exporter
+  namespace: monitoring
+  labels:
+    app: node-exporter
+spec:
+  selector:
+    matchLabels:
+      app: node-exporter
+  template:
+    metadata:
+      labels:
+        app: node-exporter
+      annotations:
+        prometheus.io/scrape: "true"
+        prometheus.io/port: "9100"
+    spec:
+      hostNetwork: true
+      hostPID: true
+      containers:
+        - name: node-exporter
+          image: prom/node-exporter:v1.7.0
+          args:
+            - "--path.procfs=/host/proc"
+            - "--path.sysfs=/host/sys"
+            - "--path.rootfs=/host/root"
+            - "--collector.filesystem.mount-points-exclude=^/(sys|proc|dev|host|etc)($$|/)"
+          ports:
+            - containerPort: 9100
+          volumeMounts:
+            - name: proc
+              mountPath: /host/proc
+              readOnly: true
+            - name: sys
+              mountPath: /host/sys
+              readOnly: true
+            - name: root
+              mountPath: /host/root
+              readOnly: true
+          resources:
+            requests:
+              cpu: 50m
+              memory: 64Mi
+            limits:
+              cpu: 200m
+              memory: 128Mi
+      volumes:
+        - name: proc
+          hostPath:
+            path: /proc
+        - name: sys
+          hostPath:
+            path: /sys
+        - name: root
+          hostPath:
+            path: /
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: node-exporter
+  namespace: monitoring
+  annotations:
+    prometheus.io/scrape: "true"
+    prometheus.io/port: "9100"
+spec:
+  type: ClusterIP
+  selector:
+    app: node-exporter
+  ports:
+    - port: 9100
+      targetPort: 9100
--- a/backend/deploy/k8s/ollama-models.yaml
+++ b/backend/deploy/k8s/ollama-models.yaml
@@ -0,0 +1,85 @@
+# Job для загрузки моделей Ollama после деплоя
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: ollama-model-loader
+  namespace: gooseek
+  labels:
+    app: ollama-model-loader
+spec:
+  ttlSecondsAfterFinished: 3600
+  backoffLimit: 3
+  template:
+    metadata:
+      labels:
+        app: ollama-model-loader
+    spec:
+      restartPolicy: OnFailure
+      initContainers:
+      - name: wait-for-ollama
+        image: curlimages/curl:latest
+        command:
+        - /bin/sh
+        - -c
+        - |
+          echo "Waiting for Ollama to be ready..."
+          until curl -sf http://ollama.gooseek.svc.cluster.local:11434/api/tags; do
+            echo "Ollama not ready, retrying in 5s..."
+            sleep 5
+          done
+          echo "Ollama is ready!"
+      containers:
+      - name: model-loader
+        image: ollama/ollama:latest
+        env:
+        - name: OLLAMA_HOST
+          value: "http://ollama.gooseek.svc.cluster.local:11434"
+        command:
+        - /bin/sh
+        - -c
+        - |
+          set -e
+          
+          OLLAMA_URL="http://ollama.gooseek.svc.cluster.local:11434"
+          
+          pull_model() {
+            MODEL=$1
+            echo "=== Pulling model: $MODEL ==="
+            
+            # Check if model already exists
+            EXISTING=$(curl -sf "$OLLAMA_URL/api/tags" | grep -o "\"name\":\"$MODEL\"" || true)
+            if [ -n "$EXISTING" ]; then
+              echo "Model $MODEL already exists, skipping..."
+              return 0
+            fi
+            
+            # Pull model via API
+            echo "Downloading $MODEL..."
+            curl -sf "$OLLAMA_URL/api/pull" \
+              -H "Content-Type: application/json" \
+              -d "{\"name\": \"$MODEL\", \"stream\": false}" \
+              --max-time 1800
+            
+            echo "Model $MODEL downloaded successfully!"
+          }
+          
+          echo "=== Ollama Model Loader ==="
+          echo "Target: $OLLAMA_URL"
+          
+          # Основная модель генерации (4 параллельных воркера)
+          pull_model "qwen3.5:9b"
+          
+          # Embedding модель (быстрые эмбеддинги)
+          pull_model "qwen3-embedding:0.6b"
+          
+          echo ""
+          echo "=== All models loaded ==="
+          curl -sf "$OLLAMA_URL/api/tags" | head -c 500
+          echo ""
+        resources:
+          requests:
+            cpu: 100m
+            memory: 256Mi
+          limits:
+            cpu: 500m
+            memory: 512Mi
--- a/backend/deploy/k8s/ollama-pull-models.sh
+++ b/backend/deploy/k8s/ollama-pull-models.sh
@@ -0,0 +1,38 @@
+#!/bin/bash
+# Скрипт для загрузки моделей в Ollama
+# Запустить ОДИН РАЗ после первого деплоя
+# Модели сохраняются в PVC и не нужно скачивать повторно
+
+set -e
+
+NAMESPACE="${NAMESPACE:-gooseek}"
+MODELS="${@:-llama3.2:3b}"
+
+echo "=== Ollama Model Loader ==="
+echo "Namespace: $NAMESPACE"
+echo "Models:    $MODELS"
+
+# Проверить что Ollama pod запущен
+echo ""
+echo "Checking Ollama pod status..."
+kubectl -n $NAMESPACE wait --for=condition=ready pod -l app=ollama --timeout=120s
+
+# Получить имя пода
+POD=$(kubectl -n $NAMESPACE get pod -l app=ollama -o jsonpath='{.items[0].metadata.name}')
+echo "Pod: $POD"
+
+# Скачать модели
+for MODEL in $MODELS; do
+  echo ""
+  echo "=== Pulling model: $MODEL ==="
+  kubectl -n $NAMESPACE exec -it $POD -c ollama -- ollama pull $MODEL
+done
+
+# Показать список моделей
+echo ""
+echo "=== Installed models ==="
+kubectl -n $NAMESPACE exec -it $POD -c ollama -- ollama list
+
+echo ""
+echo "=== Done! ==="
+echo "Models are stored in PVC and will persist across restarts."
--- a/backend/deploy/k8s/ollama.yaml
+++ b/backend/deploy/k8s/ollama.yaml
@@ -0,0 +1,130 @@
+# Ollama Deployment with GPU
+# Требования: NVIDIA GPU Operator установлен в кластере
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ollama
+  namespace: gooseek
+  labels:
+    app: ollama
+    app.kubernetes.io/name: ollama
+    app.kubernetes.io/part-of: gooseek
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: ollama
+  template:
+    metadata:
+      labels:
+        app: ollama
+    spec:
+      runtimeClassName: nvidia
+      containers:
+      # Ollama server (только GPU)
+      - name: ollama
+        image: ollama/ollama:latest
+        ports:
+        - containerPort: 11434
+          name: http
+        env:
+        - name: OLLAMA_HOST
+          value: "0.0.0.0:11434"
+        - name: OLLAMA_KEEP_ALIVE
+          value: "24h"
+        - name: OLLAMA_MODELS
+          value: "/root/.ollama/models"
+        # Параллельная обработка для SaaS
+        - name: OLLAMA_NUM_PARALLEL
+          value: "4"
+        - name: OLLAMA_MAX_LOADED_MODELS
+          value: "2"
+        - name: OLLAMA_FLASH_ATTENTION
+          value: "true"
+        # GPU
+        - name: NVIDIA_VISIBLE_DEVICES
+          value: "all"
+        - name: NVIDIA_DRIVER_CAPABILITIES
+          value: "compute,utility"
+        volumeMounts:
+        - name: ollama-data
+          mountPath: /root/.ollama
+        resources:
+          requests:
+            cpu: 1000m
+            memory: 8Gi
+            nvidia.com/gpu: 1
+          limits:
+            cpu: 4000m
+            memory: 16Gi
+            nvidia.com/gpu: 1
+        livenessProbe:
+          httpGet:
+            path: /
+            port: 11434
+          initialDelaySeconds: 30
+          periodSeconds: 30
+          timeoutSeconds: 5
+        readinessProbe:
+          httpGet:
+            path: /
+            port: 11434
+          initialDelaySeconds: 10
+          periodSeconds: 10
+          timeoutSeconds: 5
+
+      volumes:
+      - name: ollama-data
+        persistentVolumeClaim:
+          claimName: ollama-pvc
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: ollama-pvc
+  namespace: gooseek
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 20Gi
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: ollama
+  namespace: gooseek
+spec:
+  type: ClusterIP
+  selector:
+    app: ollama
+  ports:
+  - port: 11434
+    targetPort: 11434
+    name: http
+---
+# NetworkPolicy: llm-svc и model-loader могут обращаться к ollama
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: ollama-access
+  namespace: gooseek
+spec:
+  podSelector:
+    matchLabels:
+      app: ollama
+  policyTypes:
+  - Ingress
+  ingress:
+  - from:
+    - podSelector:
+        matchLabels:
+          app: llm-svc
+    - podSelector:
+        matchLabels:
+          app: ollama-model-loader
+    ports:
+    - protocol: TCP
+      port: 11434