feat: Authentik login + switchable GPU prod target

Add OIDC auth for Command Center and runtime GPU endpoint selection
pointed at atc-gpu-prod (10.0.10.106), matching what is currently deployed.
This commit is contained in:
mo
2026-07-21 23:20:24 +00:00
parent f36c8906bc
commit 9008fbd512
31 changed files with 4667 additions and 139 deletions
+9 -9
View File
@@ -183,21 +183,21 @@ NODE_REGISTRY: dict[str, dict[str, Any]] = {
},
"gpu": {
"label": "GPU Lab",
"vm": "atc-gpu-dev",
"vmid": 303,
"vm": "atc-gpu-prod",
"vmid": 306,
"pve": "atc-gpu",
"ip": "10.0.20.106",
"ssh": "ssh root@10.0.20.106",
"ip": "10.0.10.106",
"ssh": "ssh root@10.0.10.106",
"role": "inference",
"color": "#3fb950",
"description": "4× V100 GPU lab. vLLM serves the active model (Llama 3 70B GPTQ) — powers agent reasoning in this Command Center.",
"description": "4× V100 GPU lab (VM306). vLLM serves the active model (Llama 3 70B GPTQ) — powers agent reasoning in this Command Center.",
"links": [
{"label": "GPU Lab UI", "url": "http://10.0.20.106:9000"},
{"label": "vLLM API", "url": "http://10.0.20.106:8001/v1"},
{"label": "GPU Lab UI", "url": "http://10.0.10.106:9000"},
{"label": "vLLM API", "url": "http://10.0.10.106:8001/v1"},
],
"endpoints": [
{"name": "gpu-lab", "host": "10.0.20.106", "port": "9000", "proto": "http"},
{"name": "vllm", "host": "10.0.20.106", "port": "8001", "proto": "http"},
{"name": "gpu-lab", "host": "10.0.10.106", "port": "9000", "proto": "http"},
{"name": "vllm", "host": "10.0.10.106", "port": "8001", "proto": "http"},
],
"commands": ["gpu metrics", "model status", "vram usage"],
},