65 lines
1.6 KiB
YAML
65 lines
1.6 KiB
YAML
node:
|
|
host_id: "singlebox-ollama"
|
|
display_name: "SingleBox Ollama"
|
|
listen_host: "127.0.0.1"
|
|
listen_port: 8891
|
|
address: "127.0.0.1"
|
|
labels:
|
|
topology: "singlebox"
|
|
runtime: "ollama"
|
|
|
|
control_plane:
|
|
base_url: "http://127.0.0.1:8800"
|
|
node_api_key: "change-me-node-key"
|
|
heartbeat_interval_s: 5
|
|
|
|
inventory:
|
|
model_roots: []
|
|
cpu_threads: 24
|
|
ram_gb: 64
|
|
capabilities:
|
|
cpu: true
|
|
cuda: false
|
|
|
|
managed_runtimes:
|
|
enabled: false
|
|
|
|
services:
|
|
- service_id: "singlebox/chat/qwen3"
|
|
kind: "chat"
|
|
endpoint: "http://127.0.0.1:11434"
|
|
runtime:
|
|
engine: "ollama"
|
|
launcher: "external"
|
|
# discover_protocol: "ollama" queries GET /api/tags and GET /api/ps on each heartbeat.
|
|
# Assets listed here serve as a static baseline; discovered models are merged in and
|
|
# their loaded state is corrected from /api/ps (VRAM-resident = loaded: true).
|
|
# loaded_model_count and vram_used_bytes are populated in observed from /api/ps.
|
|
discover_protocol: "ollama"
|
|
assets:
|
|
- asset_id: "qwen3"
|
|
loaded: true
|
|
state:
|
|
health: "healthy"
|
|
load_state: "loaded"
|
|
accept_requests: true
|
|
observed:
|
|
p50_latency_ms: 900
|
|
|
|
- service_id: "singlebox/embeddings/nomic-embed-text"
|
|
kind: "embeddings"
|
|
endpoint: "http://127.0.0.1:11434"
|
|
runtime:
|
|
engine: "ollama"
|
|
launcher: "external"
|
|
discover_protocol: "ollama"
|
|
assets:
|
|
- asset_id: "nomic-embed-text"
|
|
loaded: true
|
|
state:
|
|
health: "healthy"
|
|
load_state: "loaded"
|
|
accept_requests: true
|
|
observed:
|
|
p50_latency_ms: 150
|