nextcloud-hpb¶
Primary audience: Nextcloud Talk users (installed by the TAPPaaS admin alongside Nextcloud).
Nextcloud Talk High-Performance Backend — a dedicated nextcloud-spreed-signaling server that offloads Talk's WebSocket signaling from the Nextcloud VM, so calls scale to hundreds of concurrent connections.
What you get¶
| Capability | Access from | How |
|---|---|---|
| Talk signaling backend | Nextcloud Talk clients | wss://<proxyDomain>/spreed (via Caddy) |
| High concurrency | Talk calls | Dedicated Go process, far lower latency than PHP signaling |
| TURN-credential distribution | Talk clients | Served over WebSocket (from coturn's shared secret) |
What is not included¶
- Media relay — that is coturn's job (
coturn:turn); HPB only does signaling. - Standalone use — it only serves the Nextcloud instance it is wired to.
- Public reachability for external calls — multi-party calls with external participants additionally require a publicly reachable coturn (WAN endpoint).
Requirements¶
srvzone- VM sizing (defaults from
nextcloud-hpb.json): 4 vCPU, 4 GB RAM, 16 GB disk ontanka1 nextcloud(nextcloud:fileservice) andcoturn(coturn:turn) deployed- A public DNS record for the HPB proxy domain (external Talk clients)
Dependencies¶
| Depends on | Purpose |
|---|---|
cluster:vm | VM provisioning |
templates:nixos | NixOS base image |
backup:vm | VM registered with Proxmox Backup Server |
network:proxy | wss:// reverse proxy via Caddy |
network:rules | Firewall pass rules (incl. callback egress to the Nextcloud proxy) |
nextcloud:fileservice | The Nextcloud instance whose Talk uses this backend |
coturn:turn | TURN relay whose credentials HPB distributes |
For installation steps see INSTALL.md.