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:
@@ -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"],
|
||||
},
|
||||
|
||||
Reference in New Issue
Block a user