add FPM worker tuning and architecture diagram

Increase PHP-FPM max_children from 5 to 12 to handle concurrent
requests without queuing, sized for a ~3GB VPS. Add Mermaid
architecture diagram to README.

README.md

@@ -2,6 +2,48 @@
 
 Docker Compose setup for a self-hosted VPS running Nextcloud, Gitea, and monitoring — managed as a GitOps-style repo.
 
+## Architecture
+
+```mermaid
+graph TB
+    Internet([Internet])
+
+    subgraph VPS["VPS (t-gstone.de)"]
+        subgraph proxy_net["proxy network"]
+            Caddy["Caddy<br/>reverse proxy + auto HTTPS"]
+        end
+
+        subgraph nc_stack["Nextcloud Stack"]
+            Nginx["Nginx<br/>static files + FastCGI proxy"]
+            NC["Nextcloud FPM<br/>PHP processing"]
+            Cron["Cron<br/>background jobs"]
+            PG["PostgreSQL 17"]
+            Redis["Redis 8"]
+        end
+
+        subgraph gitea_stack["Gitea Stack"]
+            Gitea["Gitea<br/>rootless, SQLite"]
+        end
+
+        subgraph mon_stack["Monitoring Stack"]
+            Alloy["Grafana Alloy"]
+        end
+    end
+
+    GrafanaCloud([Grafana Cloud])
+
+    Internet -->|":443 HTTPS"| Caddy
+    Internet -->|":2222 SSH"| Gitea
+    Caddy -->|"nextcloud.t-gstone.de"| Nginx
+    Caddy -->|"git.t-gstone.de"| Gitea
+    Nginx -->|":9000 FastCGI"| NC
+    NC --> PG
+    NC --> Redis
+    Cron --> PG
+    Cron --> Redis
+    Alloy -->|"logs + metrics"| GrafanaCloud
+```
+
 ## Prerequisites
 
 - A VPS with SSH access