Merge branch 'dev'

README.md

@@ -85,17 +85,29 @@ If you don't have Ollama installed yet, you can use the provided bash script for
 
 For cpu-only container
 ```bash
-chmod +x run-compose.sh && ./run-compose.sh
+./run-compose.sh
 ```
 
-For gpu-enabled container (to enable this you must have your gpu driver for docker, it mostly works with nvidia so this is the official install guide: [nvidia-container-toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html))
+For GPU enabled container (to enable this you must have your gpu driver for docker, it mostly works with nvidia so this is the official install guide: [nvidia-container-toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html))
+Warning! A GPU-enabled installation has only been tested using linux and nvidia GPU, full functionalities are not guaranteed under Windows or Macos or using a different GPU
 ```bash
-chmod +x run-compose.sh && ./run-compose.sh --enable-gpu[count=1]
+./run-compose.sh --enable-gpu
 ```
 
 Note that both the above commands will use the latest production docker image in repository, to be able to build the latest local version you'll need to append the `--build` parameter, for example:
 ```bash
-./run-compose.sh --build --enable-gpu[count=1]
+./run-compose.sh --enable-gpu --build
+```
+
+### Installing Both Ollama and Ollama Web UI Using Docker Compose
+To install using docker compose script as CPU-only installation simply run this command
+```bash
+docker compose up -d
+```
+
+for a GPU-enabled installation (provided you installed the necessary gpu drivers and you are using nvidia)
+```bash
+docker compose -f docker-compose.yaml -f docker-compose.gpu.yaml up -d
 ```
 
 ### Installing Both Ollama and Ollama Web UI Using Kustomize

kubernetes/helm/templates/ollama-service.yaml

@@ -4,6 +4,7 @@ metadata:
   name: ollama-service
   namespace: {{ .Values.namespace }}
 spec:
+  type: {{ .Values.ollama.service.type }}
   selector:
     app: ollama
   ports:

kubernetes/helm/templates/ollama-statefulset.yaml

@@ -19,15 +19,32 @@ spec:
         image: {{ .Values.ollama.image }}
         ports:
         - containerPort: {{ .Values.ollama.servicePort }}
-        resources:
-          limits:
-            cpu: {{ .Values.ollama.resources.limits.cpu }}
-            memory: {{ .Values.ollama.resources.limits.memory }}
-            nvidia.com/gpu: {{ .Values.ollama.resources.limits.gpu }}
+        env:
+        {{- if .Values.ollama.gpu.enabled }}
+          - name: PATH
+            value: /usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+          - name: LD_LIBRARY_PATH
+            value: /usr/local/nvidia/lib:/usr/local/nvidia/lib64
+          - name: NVIDIA_DRIVER_CAPABILITIES
+            value: compute,utility
+        {{- end}}
+        {{- if .Values.ollama.resources }}
+        resources: {{- toYaml .Values.ollama.resources | nindent 10 }}
+        {{- end }}
         volumeMounts:
         - name: ollama-volume
           mountPath: /root/.ollama
         tty: true
+      {{- with .Values.ollama.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      tolerations:
+        {{- if .Values.ollama.gpu.enabled }}
+        - key: nvidia.com/gpu
+          operator: Exists
+          effect: NoSchedule
+        {{- end }}
   volumeClaimTemplates:
   - metadata:
       name: ollama-volume
@@ -35,4 +52,4 @@ spec:
       accessModes: [ "ReadWriteOnce" ]
       resources:
         requests:
-          storage: 1Gi
\ No newline at end of file
+          storage: {{ .Values.ollama.volumeSize }}
\ No newline at end of file

kubernetes/helm/templates/webui-deployment.yaml

@@ -15,14 +15,24 @@ spec:
     spec:
       containers:
       - name: ollama-webui
-        image: ghcr.io/ollama-webui/ollama-webui:main
+        image: {{ .Values.webui.image }}
         ports:
         - containerPort: 8080
-        resources:
-          limits:
-            cpu: "500m"
-            memory: "500Mi"
+        {{- if .Values.webui.resources }}
+        resources: {{- toYaml .Values.webui.resources | nindent 10 }}
+        {{- end }}
+        volumeMounts:
+        - name: webui-volume
+          mountPath: /app/backend/data
         env:
         - name: OLLAMA_API_BASE_URL
           value: "http://ollama-service.{{ .Values.namespace }}.svc.cluster.local:{{ .Values.ollama.servicePort }}/api"
         tty: true
+      {{- with .Values.webui.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      volumes:
+      - name: webui-volume
+        persistentVolumeClaim:
+          claimName: ollama-webui-pvc
\ No newline at end of file

kubernetes/helm/templates/webui-ingress.yaml

+{{- if .Values.webui.ingress.enabled }}
 apiVersion: networking.k8s.io/v1
 kind: Ingress
 metadata:
   name: ollama-webui-ingress
   namespace: {{ .Values.namespace }}
-  #annotations:
-    # Use appropriate annotations for your Ingress controller, e.g., for NGINX:
-    # nginx.ingress.kubernetes.io/rewrite-target: /
+{{- if .Values.webui.ingress.annotations }}
+  annotations:
+{{ toYaml .Values.webui.ingress.annotations | trimSuffix "\n" | indent 4 }}
+{{- end }}
 spec:
   rules:
   - host: {{ .Values.webui.ingress.host }}
@@ -18,3 +20,4 @@ spec:
             name: ollama-webui-service
             port:
               number: {{ .Values.webui.servicePort }}
+{{- end }}

kubernetes/helm/templates/webui-pvc.yaml

+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  labels:
+    app: ollama-webui
+  name: ollama-webui-pvc
+  namespace: {{ .Values.namespace }}  
+spec:
+  accessModes: [ "ReadWriteOnce" ]
+  resources:
+    requests:
+      storage: {{ .Values.webui.volumeSize }}
\ No newline at end of file

kubernetes/helm/templates/webui-service.yaml

@@ -4,7 +4,7 @@ metadata:
   name: ollama-webui-service
   namespace: {{ .Values.namespace }}
 spec:
-  type: NodePort  # Use LoadBalancer if you're on a cloud that supports it
+  type: {{ .Values.webui.service.type }} # Default: NodePort  # Use LoadBalancer if you're on a cloud that supports it
   selector:
     app: ollama-webui
   ports:

kubernetes/helm/values.yaml

@@ -10,6 +10,12 @@ ollama:
       memory: "2Gi"
       nvidia.com/gpu: "0"
   volumeSize: 1Gi
+  nodeSelector: {}
+  tolerations: []
+  service:
+    type: ClusterIP
+  gpu:
+    enabled: false
 
 webui:
   replicaCount: 1
@@ -20,4 +26,13 @@ webui:
       cpu: "500m"
       memory: "500Mi"
   ingress:
+    enabled: true
+    annotations:
+    # Use appropriate annotations for your Ingress controller, e.g., for NGINX:
+      # nginx.ingress.kubernetes.io/rewrite-target: /
     host: ollama.minikube.local
+  volumeSize: 1Gi
+  nodeSelector: {}
+  tolerations: []
+  service:
+    type: NodePort
\ No newline at end of file

run-compose.sh

@@ -80,12 +80,12 @@ usage() {
     echo "  -h, --help                 Show this help message."
     echo ""
     echo "Examples:"
-    echo "  ./$0 --drop"
-    echo "  ./$0 --enable-gpu[count=1]"
-    echo "  ./$0 --enable-api[port=11435]"
-    echo "  ./$0 --enable-gpu[count=1] --enable-api[port=12345] --webui[port=3000]"
-    echo "  ./$0 --enable-gpu[count=1] --enable-api[port=12345] --webui[port=3000] --data[folder=./ollama-data]"
-    echo "  ./$0 --enable-gpu[count=1] --enable-api[port=12345] --webui[port=3000] --data[folder=./ollama-data] --build"
+    echo "  $0 --drop"
+    echo "  $0 --enable-gpu[count=1]"
+    echo "  $0 --enable-api[port=11435]"
+    echo "  $0 --enable-gpu[count=1] --enable-api[port=12345] --webui[port=3000]"
+    echo "  $0 --enable-gpu[count=1] --enable-api[port=12345] --webui[port=3000] --data[folder=./ollama-data]"
+    echo "  $0 --enable-gpu[count=1] --enable-api[port=12345] --webui[port=3000] --data[folder=./ollama-data] --build"
     echo ""
     echo "This script configures and runs a docker-compose setup with optional GPU support, API exposure, and web UI configuration."
     echo "About the gpu to use, the script automatically detects it using the "lspci" command."