monitoring: llama-swap GPU/LLM stack (v250) — PrometheusRule, Grafana dashboard, scrape config, VRAM exporter

ansible/roles/llm-inference-multimodel/templates/llama-swap-prometheus-scrape.yml.j2

@@ -0,0 +1,36 @@
+# ==============================================================================
+# FILE: roles/llm-inference-multimodel/templates/llama-swap-prometheus-scrape.yml.j2
+# DESCRIPTION: Prometheus scrape job configuration for llama-swap's native
+#              /metrics endpoint (OpenMetrics format).
+#              
+#              This template is rendered and deployed to the Prometheus
+#              config via GitOps (cluster/applications/monitoring/values.yaml).
+#              Does NOT include this file inline here; it is referenced and
+#              rendered by Ansible roles/llm-inference-multimodel/tasks/*.yml.
+#
+# TARGET HOST: astro-orbiter ({{ llm_bind_address }}:{{ llm_swapmode_port }})
+# METRICS:     llamacpp_tokens_predicted_total, llamacpp_kv_cache_usage_ratio,
+#              llamacpp_time_predict_ms, llamacpp_queue_size, etc. (per llama.cpp)
+#
+# AUTHOR:      Wong (Infrastructure Automation Specialist)
+# DATE:        2026-08-18
+# ==============================================================================
+
+---
+- job_name: llama-swap
+  static_configs:
+    - targets: ["{{ llm_bind_address }}:{{ llm_swapmode_port }}"]
+      labels:
+        component: llm-inference
+        service: llama-swap
+        environment: homelab
+  scrape_interval: 30s
+  scrape_timeout: 10s
+  honor_labels: true
+  metrics_path: /metrics
+  
+  # Relabeling: extract model name from metric labels for dashboard grouping
+  metric_relabel_configs:
+    - source_labels: [__name__]
+      regex: 'llamacpp_.*'
+      action: keep