Components & Versions
Current deployed versions of all PAIS platform components. The authoritative source is deployed.md in the repository root.
Kubernetes infrastructure
| Component |
Version |
Notes |
| Kubernetes |
1.35.2 |
kubeadm, single control plane |
| Ubuntu |
24.04 LTS |
All nodes |
| Calico CNI |
— |
Pod CIDR 192.168.0.0/16 |
| NVIDIA GPU Operator |
25.10.1 |
Time-slicing ×4 |
| KAI Scheduler |
0.13.1 |
Fair-share GPU allocation |
| Kueue |
0.11.4 |
Workload admission control |
| Component |
Version |
Port / URL |
| Kong Gateway |
3.9 (PostgreSQL-backed) |
30080 / api.pais.auckland.ac.nz |
| LiteLLM Proxy |
1.82.3 |
30180 (internal) |
| vLLM (chat) |
— |
pais-inference namespace |
| vLLM (embeddings) |
— |
pais-inference namespace |
| Open WebUI |
— |
chat.pais.auckland.ac.nz |
| Keycloak |
26.3.3 |
pais realm, Tuakiri SAML |
| PostgreSQL |
18.3.0 |
Shared backend |
| Component |
Version |
URL |
| Kubeflow |
1.11.0 |
kubeflow.pais.auckland.ac.nz |
| MLflow |
3.10.1 |
mlflow.pais.auckland.ac.nz |
| VAST CSI Driver |
2.6.5 |
VIP 10.60.200.240 |
Inference models
| Model |
ID |
Type |
| Llama 3.1 8B Instruct (AWQ INT4) |
llama-3-1-8b |
Chat |
| Qwen3-VL-Embedding-8B |
qwen3-vl-embedding-8b |
Embeddings |
Version currency
This page may lag the actual deployed state. Run ./check-docs.sh from the repository root to detect drift, or check deployed.md directly.