feat(localai): runtimeclass nvidia

projects/ai/project.yaml

@@ -3,9 +3,12 @@ config:
 
 apps:
 - name: localai
-  repoURL: https://go-skynet.github.io/helm-charts
-  chart: local-ai
-  targetRevision: 3.1.0
+  # repoURL: https://go-skynet.github.io/helm-charts
+  # chart: local-ai
+  # targetRevision: 3.1.0
+  repoURL: https://github.com/nold360/localai-charts.git
+  path: charts/local-ai
+  targetRevision: feat/runtimeclass
 
 - name: anythingllm
   repo: bjw-s 

projects/ai/values/localai.yaml

@@ -1,9 +1,14 @@
 replicaCount: 1
 
 deployment:
-  image: quay.io/go-skynet/local-ai:master-ffmpeg-core
+  image: 
+    repository: quay.io/go-skynet/local-ai
+    tag: v2.8.0-cublas-cuda12
+  pullPolicy: Always
+  runtimeClassName: nvidia
+
   env:
-    threads: 16
+    # threads: 16
     context_size: 2048
     DEBUG: "true"
 
@@ -54,9 +59,6 @@ deployment:
   prompt_templates:
     # To use cloud provided (eg AWS) image, provide it like: 1234356789.dkr.ecr.us-REGION-X.amazonaws.com/busybox
     image: busybox
-  pullPolicy: Always
-  imagePullSecrets: []
-    # - name: secret-names
 
 resources:
   requests:
@@ -80,17 +82,25 @@ models:
 
   # The list of URLs to download models from
   # Note: the name of the file will be the name of the loaded model
-  list:
-    - url: "https://gpt4all.io/models/ggml-gpt4all-j.bin"
+  list: []
+    # - url: "https://gpt4all.io/models/ggml-gpt4all-j.bin"
       # basicAuth: base64EncodedCredentials
 
-  persistence:
-    pvc:
+persistence:
+  models: 
     enabled: true
+    annotations: {}
+    storageClass: ssd
+    accessModes: ReadWriteOnce
     size: 100Gi
-      accessModes:
-        - ReadWriteOnce
-      storageClass: "ssd"
+    globalMount: /models
+  output:
+    enabled: false
+    annotations: {}
+    storageClass: ssd
+    accessModes: ReadWriteOnce
+    size: 100Gi
+    globalMount: /tmp/generated
 
 service:
   type: ClusterIP
@@ -115,5 +125,3 @@ ingress:
       hosts:
         - ai.dc
 
-image:
-  pullPolicy: IfNotPresent