Add monitoring exporters: latency, pgvector quality, and health scoring

- latency_exporter.py: Prometheus metrics for Aurelie response latency (input->TTS)
- pgvector_quality_exporter.py: Vector similarity score monitoring for memory retrieval
- Updated README.md with architecture diagram and monitoring strategy

README.md

@@ -1,3 +1,54 @@
-# infra
+# Aurelie Infrastructure Monitoring Stack
 
-Aurelie infrastructure monitoring — exporters, configs, scripts métriques
+## Purpose
+Monitoring et observabilite de la stack Aurelie/Sanitech Pro sur TrueNAS + Serveur IA.
+
+## Repo contents
+- `exporters/` — Prometheus exporters personnalises pour mesures Aurelie
+- `configs/` — Configurations Grafana dashboards, Loki pipelines, Zabbix templates (a venir)
+- `scripts/` — Collecte et transformation de scripts (a venir)
+
+## Architecture
+```
++-----------------------------------------+
+|  Serveur IA (Aurelie resident)          |
+|  +-- Node Exporter (CPU/RAM/GPU/Disk)  |
+|   +-- Prometheus (time-series DB)      |
+|   +- Aurelie exporters:                   |
+       +-- latency_exporter.py            |
+       +-- pgvector_quality_exporter.py   |
+       +- tts_audio_quality_exporter.py    |
+                                          |
+
++-----------------------------------------+
+|  TrueNAS (storage primary)              |
+|   +-- Zabbix Server                     |
+|    +-- Zabbix Agent                     |
+                                         |
+
++------------------------------------------+
+|  Grafana                                 |
+|    +-- Sources: Prometheus + Zabbix      |
+|     +-- Dashboards:                      |
+             +-- Aurelie Health (composite) |
+             +-- Infrastructure Status      |
+              +- Latency & Throughput        |
+               +- Alerting->Telegram/email   |
+
++-----------------------------------------+
+```
+
+## Monitoring Strategy
+- **Zabbix + UniFi**: infrastructure physique (CPU, RAM, network, switch ports)  
+- **Prometheus + Node Exporter**: services Docker + mesures AIT personnalisees
+- **Grafana+: visualisation unifie de TOUT dans un seul dashboard
+**Loki + Alloy**: centralisation des logs (plus de journalctl dispersee)
+
+## Getting Started
+1. Deploy exporters in a dedicated Docker container on the IA server
+2. Configurer Grafana datasource -> Prometheus -> localhost:9090
+3. Importer Aurelie Health dashboard JSON
+4.Activer alerting to Telegram/email via Grafana notification channels
+
+---
+Created by Aurelie as part of Sanitech Pro infrastructure convergence.