Move reranker to Serenity, add monitoring and mesh-mqtt-observer stacks

- ai: remove reranker service (moved to Serenity at 10.5.1.5:9787)
- ai: add LiteLLM reranker proxy entry for Serenity
- ai: configure OpenWebUI to use LiteLLM for reranking
- monitoring: add Grafana + Prometheus stack
- mesh-mqtt-observer: add new service
- automation, documents, meshtastic: misc updates

ai/docker-compose.yaml

@@ -52,6 +52,10 @@ services:
       ENABLE_LOGIN_FORM: "true"
       VECTOR_DB: qdrant
       QDRANT_URI: http://qdrant:6333
+      RAG_RERANKING_MODEL: "reranker"
+      RAG_RERANKING_MODEL_AUTO_UPDATE: "false"
+      RAG_OPENAI_API_BASE_URL: "http://litellm:4000"
+      RAG_OPENAI_API_KEY: ${LITELLM_MASTER_KEY}
     volumes:
       - /mnt/docker-ssd/docker/appdata/openwebui:/app/backend/data
     depends_on:
@@ -101,19 +105,6 @@ services:
     extra_hosts:
       - "host.docker.internal:host-gateway"
 
-  reranker:
-    container_name: reranker
-    image: ghcr.io/huggingface/text-embeddings-inference:cpu-latest
-    restart: unless-stopped
-    networks:
-      - ai-net
-      - ai-services
-    ports:
-      - "8787:80"
-    command: ["--model-id", "BAAI/bge-reranker-v2-m3"]
-    volumes:
-      - /mnt/docker-ssd/docker/appdata/reranker:/data
-
   whisper:
     container_name: whisper
     image: fedirz/faster-whisper-server:latest-cuda

ai/litellm/config.yaml

@@ -1,4 +1,9 @@
 model_list:
+  # ---- Reranker (Serenity, 10.5.1.5) ----
+  - model_name: "reranker"
+    litellm_params:
+      model: "huggingface/BAAI/bge-reranker-v2-m3"
+      api_base: "http://10.5.1.5:9787"
   # ---- ROCINANTE (RTX 4090, 24GB) — heavy reasoning ----
   - model_name: "heavy"
     litellm_params: