mirror of
https://github.com/danswer-ai/danswer.git
synced 2025-03-18 05:41:58 +01:00
57 lines
1.2 KiB
YAML
57 lines
1.2 KiB
YAML
apiVersion: v1
|
|
kind: Service
|
|
metadata:
|
|
name: inference-model-server-service
|
|
spec:
|
|
selector:
|
|
app: inference-model-server
|
|
ports:
|
|
- name: inference-model-server-port
|
|
protocol: TCP
|
|
port: 9000
|
|
targetPort: 9000
|
|
type: ClusterIP
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: inference-model-server-deployment
|
|
spec:
|
|
replicas: 1
|
|
selector:
|
|
matchLabels:
|
|
app: inference-model-server
|
|
template:
|
|
metadata:
|
|
labels:
|
|
app: inference-model-server
|
|
spec:
|
|
containers:
|
|
- name: inference-model-server
|
|
image: danswer/danswer-model-server:latest
|
|
imagePullPolicy: IfNotPresent
|
|
command: [ "uvicorn", "model_server.main:app", "--host", "0.0.0.0", "--port", "9000" ]
|
|
ports:
|
|
- containerPort: 9000
|
|
envFrom:
|
|
- configMapRef:
|
|
name: env-configmap
|
|
volumeMounts:
|
|
- name: inference-model-storage
|
|
mountPath: /root/.cache
|
|
volumes:
|
|
- name: inference-model-storage
|
|
persistentVolumeClaim:
|
|
claimName: inference-model-pvc
|
|
---
|
|
apiVersion: v1
|
|
kind: PersistentVolumeClaim
|
|
metadata:
|
|
name: inference-model-pvc
|
|
spec:
|
|
accessModes:
|
|
- ReadWriteOnce
|
|
resources:
|
|
requests:
|
|
storage: 3Gi
|