kalavai-client 0.5.0__py2.py3-none-any.whl

kalavai_client/__init__.py

@@ -0,0 +1,2 @@
+
+__version__ = "0.5.0"

kalavai_client/__main__.py

@@ -0,0 +1,5 @@
+from kalavai_client.cli import app
+
+
+if __name__ == "__main__":
+    app()

kalavai_client/assets/apps.yaml

@@ -0,0 +1,201 @@
+helmDefaults:
+  timeout: 1200
+
+repositories:
+  - name: kuberay
+    url: https://ray-project.github.io/kuberay-helm/
+  - name: nvidia
+    url: https://helm.ngc.nvidia.com/nvidia
+  - name: kalavai
+    url: https://kalavai-net.github.io/helm-charts/
+  - name: longhorn
+    url: https://charts.longhorn.io
+  - name: volcano-sh
+    url: https://volcano-sh.github.io/helm-charts
+  - name: prometheus
+    url: https://prometheus-community.github.io/helm-charts
+  - name: opencost-charts
+    url: https://opencost.github.io/opencost-helm-chart
+  - name: minio
+    url: https://charts.min.io/
+
+releases:
+  - name: datashim
+    namespace: dlf
+    chart: kalavai/kalavai-datashim
+    version: "0.4.0"
+    installed: false
+  - name: minio
+    needs: 
+    - kalavai/longhorn
+    namespace: minio
+    chart: minio/minio
+    installed: false
+    set:
+    - name: replicas
+      value: {{minio_replicas}}
+    - name: resources.requests.memory
+      value: "{{minio_resources_memory}}"
+    - name: persistence.enabled
+      value: true
+    - name: persistence.storageClass
+      value: {{minio_persistence_storageClass}}
+    - name: persistence.size
+      value: "{{minio_persistence_size}}"
+    - name: persistence.accessMode
+      value: "ReadWriteMany"
+    - name: rootUser
+      value: {{minio_rootUser}}
+    - name: rootPassword
+      value: {{minio_rootPassword}}
+    - name: service.type
+      value: "NodePort"
+    - name: service.nodePort
+      value: {{minio_service_port}}
+    - name: consoleService.type
+      value: "NodePort"
+    - name: consoleService.nodePort
+      value: {{minio_console_port}}
+    - name: buckets[0].name
+      value: "llm-storage"
+    - name: buckets[0].policy
+      value: "public"
+    - name: buckets[0].purge
+      value: false
+  - name: helios
+    namespace: kalavai
+    chart: kalavai/kalavai-helios
+    version: "0.1.9"
+    installed: {{not watcher_allow_unregistered_user}}
+    set:
+    - name: deployment.watcher_endpoint
+      value: "http://{{watcher_service}}"
+    - name: deployment.watcher_auth_key
+      value: "{{watcher_readonly_key}}"
+    - name: deployment.kalavai_api_endpoint
+      value: {{kalavai_api_endpoint}}
+    - name: deployment.user_node_label
+      value: "{{user_node_label}}"
+    - name: deployment.sleep_interval
+      value: "{{helios_harvest_interval}}"
+  - name: opencost
+    namespace: opencost
+    chart: opencost-charts/opencost
+    installed: {{not watcher_allow_unregistered_user}}
+  - name: prometheus
+    namespace: prometheus-system
+    chart: prometheus/prometheus
+    installed: {{not watcher_allow_unregistered_user}}
+    set:
+    - name: prometheus-pushgateway.enabled
+      value: false
+    - name: alertmanager.enabled
+      value: false
+  - name: volcano-sh
+    namespace: kalavai
+    chart: volcano-sh/volcano
+    installed: true
+  - name: kuberay
+    namespace: kuberay
+    chart: kuberay/kuberay-operator
+    installed: true
+    version: "1.2.2"
+  - name: kuberay-apiserver
+    namespace: kuberay
+    chart: kuberay/kuberay-apiserver
+    installed: false
+  - name: longhorn
+    namespace: kalavai
+    chart: longhorn/longhorn
+    installed: false
+    set:
+    # security issue! enable for testing only
+    - name: service.ui.type
+      value: NodePort
+    - name: service.ui.nodePort
+      value: {{longhorn_ui_port}}
+    - name: service.manager.type
+      value: NodePort
+    - name: service.manager.nodePort
+      value: {{longhorn_manager_port}}
+    - name: persistence.defaultClassReplicaCount
+      value: {{longhorn_replicas}}
+    - name: global.nodeSelector.{{longhorn_label_selector_key}}
+      value: "{{longhorn_label_selector_value}}"
+    - name: defaultSettings.storageMinimalAvailablePercentage
+      value: {{longhorn_minimal_available_percentage}}
+  - name: lws
+    namespace: kalavai
+    chart: kalavai/lws
+    installed: false
+  - name: kalavai-watcher
+    namespace: kalavai
+    chart: kalavai/kalavai-watcher
+    version: "0.3.6"
+    installed: true
+    set:
+    - name: namespace
+      value: kalavai
+    - name: replicas
+      value: 2
+    - name: image_tag
+      value: "v2025.01"
+    - name: deployment.in_cluster
+      value: "True"
+    - name: deployment.use_auth_key
+      value: "True"
+    - name: deployment.admin_key
+      value: "{{watcher_admin_key}}"
+    - name: deployment.write_key
+      value: "{{watcher_write_key}}"
+    - name: deployment.readonly_key
+      value: "{{watcher_readonly_key}}"
+    - name: deployment.is_shared_pool
+      value: {{watcher_is_shared_pool}}
+    - name: deployment.allow_unregistered_user
+      value: "{{watcher_allow_unregistered_user}}"
+    - name: deployment.kalavai_api_endpoint
+      value: {{kalavai_api_endpoint}}
+    - name: deployment.prometheus_endpoint
+      value: {{prometheus_endpoint}}
+    - name: deployment.opencost_endpoint
+      value: {{opencost_endpoint}}
+    - name: deployment.longhorn_manager_endpoint
+      value: {{longhorn_manager_endpoint}}
+    - name: service.nodePort
+      value: {{watcher_port}}
+  - name: nvidia-gpu-operator
+    namespace: kalavai
+    chart: kalavai/gpu
+    installed: true    
+  - name: hami-vgpu
+    namespace: kalavai
+    chart: kalavai/hami
+    installed: true
+    set:
+    - name: resourceCores
+      value: "nvidia.com/gpucores"
+    - name: devicePlugin.runtimeClassName
+      value: "nvidia"
+    - name: scheduler.defaultSchedulerPolicy.nodeSchedulerPolicy
+      value: "binpack"
+    - name: scheduler.defaultSchedulerPolicy.gpuSchedulerPolicy
+      value: "binpack"
+    - name: scheduler.defaultCores
+      value: "100"
+    - name: scheduler.kubeScheduler.imageTag
+      value: v1.31.1
+    - name: devicePlugin.deviceMemoryScaling
+      value: "1"
+    - name: devicePlugin.deviceSplitCount
+      value: "1"
+    - name: scheduler.customWebhook.port
+      value: "30498"
+    - name: scheduler.service.schedulerPort
+      value: "30498"
+    - name: scheduler.service.monitorPort
+      value: "30493"
+    - name: devicePlugin.service.httpPort
+      value: "30492"
+    
+

kalavai_client/assets/apps_values.yaml

@@ -0,0 +1,83 @@
+# TODO: add helm versions here
+
+- name: kalavai_api_endpoint
+  default: https://platform.kalavai.net/_/api
+  description: ""
+
+- name: prometheus_endpoint
+  default: http://prometheus-server.prometheus-system.svc.cluster.local:80
+  description: ""
+
+- name: opencost_endpoint
+  default: http://opencost.opencost.svc.cluster.local:9003
+  description: ""
+
+- name: longhorn_manager_endpoint
+  default: http://longhorn-backend.kalavai.svc.cluster.local:9500
+  description: ""
+
+- name: helios_harvest_interval
+  default: 120
+  description: "Interval (minutes) at which to report resource usage in public pools"
+
+- name: watcher_is_shared_pool
+  default: "True"
+  description: "All users see each other's work"
+
+## Longhorn
+- name: longhorn_ui_port
+  default: 30000
+  description: ""
+
+- name: longhorn_manager_port
+  default: "" #31011
+  description: ""
+
+- name: longhorn_minimal_available_percentage
+  default: 5
+  description: ""
+
+- name: longhorn_label_selector_key
+  default: "kalavai.storage.enabled"
+  description: ""
+
+- name: longhorn_label_selector
+  default: "True"
+  description: ""
+
+- name: longhorn_replicas
+  default: 2
+  description: ""
+
+## MinIO
+- name: minio_replicas
+  default: 2
+  description: ""
+
+- name: minio_resources_memory
+  default: "1Gi"
+  description: ""
+
+- name: minio_persistence_storageClass
+  default: "longhorn"
+  description: ""
+
+- name: minio_persistence_size
+  default: "10Gi"
+  description: ""
+
+- name: minio_service_port
+  default: 32000
+  description: ""
+
+- name: minio_console_port
+  default: 32001
+  description: ""
+
+- name: minio_rootUser
+  default: "admin"
+  description: ""
+
+- name: minio_rootPassword
+  default: "password"
+  description: ""

kalavai_client/assets/docker-compose-template.yaml

@@ -0,0 +1,55 @@
+services:
+   {{service_name}}-{{command}}:
+      image: bundenth/kalavai-runner:gpu-latest
+      container_name: {{service_name}}
+      hostname: {{hostname}}
+      privileged: true
+      restart: unless-stopped
+      ports:
+      - "6443:6443" # kube server
+      - "8472:8472" # flannel vxlan
+      - "51820:51820" # flannel wireguard
+{% if command == "server" %}
+      - "30000-30500:30000-30500"
+{% endif %}
+      networks:
+      - custom-network
+      command: >
+        {{command}}
+{% if command == "server" %}
+        --flannel-backend wireguard-native
+        --service-node-port-range "30000-30500"
+{% else %}
+        --server {{pool_ip}}
+        --token {{token}}
+{% endif %}
+        --node-label role={{command}}
+{% if node_labels %}
+        {{node_labels}}
+{% endif %}
+{% if num_gpus and num_gpus > 0 %}
+        --node-label gpu=on
+{% else %}
+        --node-label gpu=off
+{% endif %}
+        --node-ip {{ip_address}}
+        --node-external-ip {{ip_address}}
+{% if flannel_iface %}
+        --flannel-iface {{flannel_iface}}
+{% endif %}
+      volumes:
+      - {{k3s_path}}:/var/lib/rancher/k3s # Persist data
+      - {{etc_path}}:/etc/rancher/k3s # Config files
+{% if num_gpus and num_gpus > 0 %}
+      deploy:
+        resources:
+          reservations:
+            devices:
+            - driver: nvidia
+              count: {{num_gpus}}
+              capabilities: [gpu]
+{% endif %}
+
+networks:
+  custom-network:
+    driver: bridge

kalavai_client/assets/pool_config_template.yaml

@@ -0,0 +1,19 @@
+# # # storage class for model weights (temporary)
+# kind: StorageClass
+# apiVersion: storage.k8s.io/v1
+# metadata:
+#   name: {{storage_class_name}}
+# provisioner: driver.longhorn.io
+# allowVolumeExpansion: true
+# reclaimPolicy: Delete
+# volumeBindingMode: Immediate
+# parameters:
+#   shareManagerNodeSelector: {{storage_label_selector}}
+#   numberOfReplicas: "{{storage_replicas}}"
+#   staleReplicaTimeout: "10"
+#   fromBackup: ""
+#   fsType: "ext4"
+#   dataLocality: "disabled"
+#   unmapMarkSnapChainRemoved: "ignored"
+#   disableRevisionCounter: "true"
+#   dataEngine: "v1"

kalavai_client/assets/pool_config_values.yaml

@@ -0,0 +1,12 @@
+# STORAGE #
+- name: storage_label_selector
+  default: "kalavai.storage.enabled:True"
+  description: ""
+
+- name: storage_class_name
+  default: "longhorn-model"
+  description: ""
+
+- name: storage_replicas
+  default: 1
+  description: ""

kalavai_client/assets/user_workspace.yaml

@@ -0,0 +1,19 @@
+# apiVersion: v1
+# kind: Secret
+# metadata:
+#   name: {{dataset_secret_name}}
+# stringData:
+#   accessKeyID: "{{dataset_access_key_id}}"
+#   secretAccessKey: "{{dataset_secret_access_key}}"
+# ---
+# apiVersion: datashim.io/v1alpha1
+# kind: Dataset
+# metadata:
+#   name: {{dataset_name}}
+# spec:
+#   local:
+#     type: "COS"
+#     secret-name: {{dataset_secret_name}}
+#     endpoint: "{{dataset_endpoint}}"
+#     bucket: {{dataset_bucket}}
+#     readonly: "{{dataset_readonly}}"

kalavai_client/assets/user_workspace_values.yaml

@@ -0,0 +1,29 @@
+- name: dataset_secret_name
+  default: "kalavai-s3-secret"
+  description: ""
+
+- name: dataset_access_key_id
+  default: "admin"
+  description: ""
+
+# must match minio password
+- name: dataset_secret_access_key
+  default: "password"
+  description: ""
+
+- name: dataset_name
+  default: "models-hub"
+  description: ""
+
+- name: dataset_endpoint
+  default: "http://minio.minio.svc.cluster.local:9000"
+  description: ""
+
+- name: dataset_bucket
+  default: "llm-storage"
+  description: ""
+
+- name: dataset_readonly
+  default: "false"
+  description: ""
+

kalavai_client/assets/vpn-template.yaml

@@ -0,0 +1,13 @@
+# https://docs.netmaker.io/docs/netclient#docker
+services:
+    {{service_name}}:
+        network_mode: host
+        privileged: true
+        restart: always
+        environment:
+        - TOKEN={{vpn_token}}
+        - IFACE_NAME={{flannel_iface}}
+        volumes:
+        - '{{etc_path}}/netclient:/etc/netclient'
+        container_name: {{service_name}}
+        image: 'gravitl/netclient:latest'

kalavai_client/auth.py

@@ -0,0 +1,68 @@
+import os
+import pickle
+
+import anvil.server
+import anvil.users
+
+
+AUTH_UPLINK_KEY = os.getenv("ANVIL_UPLINK_KEY", "client_AOPKTWK227ZV3R4ENTMOQRIY-ADMVMYW5OIRPH75P")
+
+class KalavaiAuthClient:
+    def __init__(self, user_cookie_file=None):
+        anvil.server.connect(AUTH_UPLINK_KEY, quiet=True)
+        self.user_cookie_file = user_cookie_file
+        user = self.load_user_session()
+
+    def login(self, username, password):
+        try:
+            user = anvil.users.login_with_email(username, password, remember=True)
+        except:
+            return None
+        
+        if self.user_cookie_file:
+            with open(self.user_cookie_file, "wb") as f:
+                    pickle.dump(user, f)  
+        return user
+
+    def logout(self):
+        anvil.users.logout()
+        try:
+            os.remove(self.user_cookie_file)
+        except:
+            pass
+    
+    def is_logged_in(self):
+        user = self.load_user_session()
+        return user is not None
+
+    def load_user_session(self):
+        user = anvil.users.get_user()
+        if user:
+            return user
+        try:
+            with open(self.user_cookie_file, "rb") as f:
+                user = pickle.load(f)
+                return user
+        except:
+            return None
+        
+    def call_function(self, fn, *args):
+        return anvil.server.call(
+            fn,
+            *args
+        )
+
+    
+
+if __name__ == "__main__":
+    auth = KalavaiAuthClient(
+        user_cookie_file="here.pickle"
+    )
+    user = auth.load_user_session()
+    if not user:
+        user = auth.login(username="carlos@kalavai.net", password="wrong_pass")
+
+    if user is None:
+        print("Failed to login")
+    else:
+        print(user)