kalavai-client 0.7.1__tar.gz → 0.7.3__tar.gz

{kalavai_client-0.7.1 → kalavai_client-0.7.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: kalavai-client
-Version: 0.7.1
+Version: 0.7.3
 Summary: Client app for kalavai platform
 License: Apache-2.0
 Keywords: LLM,platform

kalavai_client-0.7.3/kalavai_client/__init__.py

@@ -0,0 +1,2 @@
+
+__version__ = "0.7.3"

{kalavai_client-0.7.1 → kalavai_client-0.7.3}/kalavai_client/assets/apps.yaml

@@ -35,7 +35,7 @@ releases:
     namespace: kalavai
     chart: jetstacks/cert-manager
     version: v1.15.1
-    installed: true
+    installed: {{deploy_cert_manager|default("False", true)}}
     set:
     - name: crds.enabled
       value: true
@@ -44,16 +44,11 @@ releases:
     - kalavai/certificates
     namespace: kalavai
     chart: rocm/gpu-operator-charts
-    installed: true
-  - name: datashim
-    namespace: dlf
-    chart: kalavai/kalavai-datashim
-    version: "0.4.0"
-    installed: false
+    installed: {{deploy_rocm|default("false", true)}}
   - name: lago
     namespace: kalavai
     chart: kalavai/lago
-    installed: {{deploy_lago}}
+    installed: {{deploy_lago|default("false", true)}}
     set:
     - name: external.api.nodePort
       value: 32000
@@ -68,7 +63,7 @@ releases:
     - kalavai/longhorn
     namespace: minio
     chart: minio/minio
-    installed: false
+    installed: {{deploy_minio|default("false", true)}}
     set:
     - name: replicas
       value: {{minio_replicas}}
@@ -104,7 +99,7 @@ releases:
     namespace: kalavai
     chart: kalavai/kalavai-helios
     version: "0.1.11"
-    installed: false
+    installed: {{deploy_helios|default("false", true)}}
     set:
     - name: deployment.watcher_endpoint
       value: "http://{{watcher_service}}"
@@ -119,33 +114,37 @@ releases:
   - name: opencost
     namespace: opencost
     chart: opencost-charts/opencost
-    installed: {{deploy_opencost}}
+    installed: {{deploy_opencost|default("false", true)}}
   - name: prometheus
     namespace: prometheus-system
     chart: prometheus/prometheus
-    installed: {{deploy_prometheus}}
+    installed: {{deploy_prometheus|default("false", true)}}
     set:
     - name: prometheus-pushgateway.enabled
       value: false
     - name: alertmanager.enabled
       value: false
+    - name: server.retention
+      value: {{prometheus_server_retention}}
+    - name: server.persistentVolume.size
+      value: {{prometheus_disk_size}}
   - name: volcano-sh
     namespace: kalavai
     chart: volcano-sh/volcano
-    installed: true
+    installed: {{deploy_volcano|default("false", true)}}
   - name: kuberay
     namespace: kuberay
     chart: kuberay/kuberay-operator
-    installed: {{deploy_kuberay}}
+    installed: {{deploy_kuberay|default("false", true)}}
     version: "1.2.2"
   - name: kuberay-apiserver
     namespace: kuberay
     chart: kuberay/kuberay-apiserver
-    installed: false
+    installed: {{deploy_kuberay|default("false", true)}}
   - name: longhorn
     namespace: kalavai
     chart: longhorn/longhorn
-    installed: false
+    installed: {{deploy_longhorn|default("false", true)}}
     set:
     # security issue! enable for testing only
     - name: service.ui.type
@@ -162,22 +161,18 @@ releases:
       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: "0.3.8"
-    installed: true
+    installed: {{deploy_watcher|default("false", true)}}
     set:
     - name: namespace
       value: kalavai
     - name: replicas
       value: 1
     - name: image_tag
-      value: "v2025.08.3" #"v2025.07.34"
+      value: "v2025.09.1" #"v2025.07.34"
     - name: deployment.in_cluster
       value: "True"
     - name: deployment.kalavai_username_key
@@ -204,14 +199,10 @@ releases:
       value: {{longhorn_manager_endpoint}}
     - name: service.nodePort
       value: {{watcher_port}}
-  - name: nvidia-gpu-operator
-    namespace: kalavai
-    chart: kalavai/gpu
-    installed: false
   - name: hami-vgpu
     namespace: kalavai
     chart: hami-charts/hami
-    installed: true
+    installed: {{deploy_hami|default("false", true)}}
     set:
     - name: resourceCores
       value: "nvidia.com/gpucores"

kalavai_client-0.7.3/kalavai_client/assets/default_pool_config.yaml

@@ -0,0 +1,67 @@
+server:
+  platform: "amd64"
+  ip_address: null
+  location: null
+  name: "kalavai_cluster"
+
+core:
+  # Deploy systems
+  deploy:
+  - volcano
+  - cert_manager
+  - rocm
+  - watcher
+  - hami
+  #- lago
+  #- opencost
+  - prometheus
+  #- kuberay
+  #- minio
+  # "Kalavai API endpoint"
+  kalavai_api_endpoint: "https://platform.kalavai.net/_/api"
+  # "Opencost endpoint"
+  opencost_endpoint: "http://opencost.opencost.svc.cluster.local:9003"
+  # "Longhorn manager endpoint"
+  longhorn_manager_endpoint: "http://longhorn-backend.kalavai.svc.cluster.local:9500"
+  # "Helios harvest interval"
+  helios_harvest_interval: 120
+  # "Watcher is shared pool"
+  watcher_is_shared_pool: "True"
+  # "Prometheus endpoint"
+  prometheus_endpoint: "http://prometheus-server.prometheus-system.svc.cluster.local:80"
+  # "Prometheus server retention"
+  prometheus_server_retention: "30d"
+  # "Prometheus disk size"
+  prometheus_disk_size: "8Gi"
+  # "Longhorn UI port"
+  longhorn_ui_port: 30000
+  # "Longhorn manager port"
+  longhorn_manager_port: "" #31011
+  # "Longhorn minimal available percentage"
+  longhorn_minimal_available_percentage: 5
+  # "Longhorn label selector key"
+  longhorn_label_selector_key: "kalavai.storage.enabled"
+  # "Longhorn label selector"
+  longhorn_label_selector: "True"
+  # "Longhorn replicas"
+  longhorn_replicas: 2
+  # "MinIO replicas"
+  minio_replicas: 2
+  # "MinIO resources memory"
+  minio_resources_memory: "1Gi"
+  # "MinIO persistence storage class"
+  minio_persistence_storageClass: "longhorn"
+  # "MinIO persistence size"
+  minio_persistence_size: "10Gi"
+  # "MinIO service port"
+  minio_service_port: 32000
+  # "MinIO console port"
+  minio_console_port: 32001
+  # "MinIO root user"
+  minio_rootUser: "admin"
+  # "MinIO root password"
+  minio_rootPassword: "password"
+
+pool:
+  # Name of the storage class to use for the pool
+  storage_class_name: "longhorn"

{kalavai_client-0.7.1 → kalavai_client-0.7.3}/kalavai_client/bridge_api.py

@@ -102,10 +102,8 @@ def pool_create(request: CreatePoolRequest, api_key: str = Depends(verify_api_ke
     
     - **cluster_name**: Name of the cluster
     - **ip_address**: IP address for the pool
-    - **app_values**: Application configuration values
     - **num_gpus**: Number of GPUs to allocate
     - **node_name**: Name of the node
-    - **only_registered_users**: Whether to restrict to registered users
     - **location**: Location of the pool
     - **description**: Pool description
     - **token_mode**: Token type for authentication
@@ -113,7 +111,6 @@ def pool_create(request: CreatePoolRequest, api_key: str = Depends(verify_api_ke
     result = create_pool(
         cluster_name=request.cluster_name,
         ip_address=request.ip_address,
-        app_values=request.app_values,
         num_gpus=request.num_gpus,
         node_name=request.node_name,
         location=request.location,

{kalavai_client-0.7.1 → kalavai_client-0.7.3}/kalavai_client/bridge_models.py

@@ -10,7 +10,6 @@ class InvitesRequest(BaseModel):
 class CreatePoolRequest(BaseModel):
     cluster_name: str = Field(description="Name of the cluster to create")
     ip_address: str = Field(description="IP address for the pool")
-    app_values: dict = Field(None, description="Application configuration values")
     num_gpus: int = Field(None, description="Number of GPUs to allocate")
     node_name: str = Field(None, description="Name of the node")
     location: str = Field(None, description="Geographic location of the pool")

{kalavai_client-0.7.1 → kalavai_client-0.7.3}/kalavai_client/cli.py

@@ -396,7 +396,8 @@ def pool__list(*others, user_only=False):
 
 
 @arguably.command
-def pool__start(cluster_name, *others,  platform="amd64", ip_address: str=None, location: str=None, app_values: str=None, pool_config_values: str=None, non_interactive: bool=False):
+def pool__start(*others,  pool_config_file=None, apps: list=None, platform="amd64", ip_address: str=None, location: str=None, app_values: str=None, pool_config_values: str=None, non_interactive: bool=False):
+
     """
     Start Kalavai pool and start/resume sharing resources.
 
@@ -430,16 +431,15 @@ def pool__start(cluster_name, *others,  platform="amd64", ip_address: str=None,
     
     console.log(f"Using {ip_address} address for server")
 
-    console.log(f"[green]Creating {cluster_name} pool, this may take a few minutes...")
+    console.log(f"[green]Creating pool, this may take a few minutes...")
 
     result = create_pool(
-        target_platform=platform,
-        cluster_name=cluster_name,
         ip_address=ip_address,
-        app_values=app_values,
-        pool_config_values=pool_config_values,
-        num_gpus=input_gpus(non_interactive=non_interactive),
-        location=location
+        location=location,
+        target_platform=platform,
+        pool_config_file=pool_config_file,
+        apps=apps,
+        num_gpus=input_gpus(non_interactive=non_interactive)
     )
 
     if "warning" in result:

{kalavai_client-0.7.1 → kalavai_client-0.7.3}/kalavai_client/core.py

@@ -1,4 +1,5 @@
 import os
+import yaml
 import time
 from collections import defaultdict
 import math
@@ -71,7 +72,8 @@ from kalavai_client.env import (
     HELM_APPS_VALUES,
     POOL_CONFIG_DEFAULT_VALUES,
     POOL_CONFIG_TEMPLATE,
-    FORBIDEDEN_IPS
+    FORBIDEDEN_IPS,
+    DEFAULT_POOL_CONFIG_TEMPLATE
 )
 
 class Job(BaseModel):
@@ -505,7 +507,7 @@ def fetch_gpus(available=False):
                 status = gpu["ready"] if "ready" in gpu else True
                 if available and not status:
                     continue
-                row_gpus.append( (f"{gpu['model']} ({gpu['memory']}) vRAM)", str(status)))
+                row_gpus.append( (f"{gpu['model']} ({gpu['memory']} vRAM)", str(status)))
             if len(row_gpus) > 0:
                 models, statuses = zip(*row_gpus)
                 #rows.append([node, "\n".join(statuses), "\n".join(models), str(gpus["available"]), str(gpus["capacity"])])
@@ -752,47 +754,56 @@ def join_pool(
     return cluster_name
 
 def create_pool(
-        cluster_name: str,
-        ip_address: str,
+        cluster_name: str=None,
+        ip_address: str=None,
+        location: str=None,
+        target_platform: str="amd64",
+        pool_config_file: str=None,
         description: str="",
         token_mode: TokenType=TokenType.USER,
-        app_values: str=None,
-        pool_config_values: str=None,
-        num_gpus: int=0,
+        num_gpus: int=-1,
         node_name: str=None,
-        location: str=None,
-        target_platform: str="amd64"
+        apps: list=[]
     ):
 
     if not check_seed_compatibility():
         return {"error": "Requirements failed"}
     
-    if app_values is None:
-        app_values = HELM_APPS_VALUES
-    
-    if pool_config_values is None:
-        pool_config_values = POOL_CONFIG_DEFAULT_VALUES
+    if pool_config_file is None:
+        pool_config_file = DEFAULT_POOL_CONFIG_TEMPLATE
+
+    if node_name is None:
+        node_name = socket.gethostname()
 
-    node_name = socket.gethostname()
     user_id = load_user_id()
     
     node_labels = {
         STORAGE_CLASS_LABEL: is_storage_compatible(),
         NODE_ROLE_LABEL: "server"
     }
-
-    # if location is not None:
-    #     try:
-    #         vpn = get_vpn_details(
-    #             location=location,
-    #             user_cookie=USER_COOKIE)
-    #         node_labels[USER_NODE_LABEL] = user["username"]
-    #     except Exception as e:
-    #         return {"error": f"[red]Error when joining network: {str(e)}"}
         
-    if num_gpus is None:
+    if num_gpus < 0:
         num_gpus = get_max_gpus()
-    
+
+    # load values from pool config
+    with open(pool_config_file, "r") as f:
+        config_values = yaml.safe_load(f)
+    # use default values if not provided
+    try:
+        cluster_name = config_values["server"]["name"] if cluster_name is None else cluster_name
+        ip_address = config_values["server"]["ip_address"] if ip_address is None else ip_address
+        location = config_values["server"]["location"] if location is None else location
+        target_platform = config_values["server"]["platform"] if target_platform is None else target_platform
+        app_values = config_values["core"]
+        post_config_values = config_values["pool"]
+        deploy_apps = {
+            f"deploy_{app}": True for app in config_values["core"]["deploy"]
+        }
+        for app in apps:
+            deploy_apps[f"deploy_{app}"] = True
+    except Exception as e:
+        return {"error": f"Error when loading pool config. Missing format? {str(e)}"}
+
     # Generate docker compose recipe
     generate_compose_config(
         target_platform=target_platform,
@@ -818,13 +829,12 @@ def create_pool(
             time.sleep(10)
 
     # populate local cred files
-    
     auth_key = user_id if user_id is not None else str(uuid.uuid4())
     write_auth_key = str(uuid.uuid4())
     readonly_auth_key = str(uuid.uuid4())
     watcher_service = f"{ip_address}:{DEFAULT_WATCHER_PORT}"
     values = {
-        CLUSTER_NAME_KEY: cluster_name,
+        #CLUSTER_NAME_KEY: cluster_name,
         CLUSTER_IP_KEY: ip_address,
         USER_ID_KEY: user_id if user_id is not None else "",
         AUTH_KEY: auth_key,
@@ -851,8 +861,8 @@ def create_pool(
     # Generate helmfile recipe
     helm_yaml = load_template(
         template_path=HELM_APPS_FILE,
-        values=values,
-        default_values_path=app_values,
+        values={**values, **deploy_apps, **app_values},
+        #default_values_path=app_values,
         force_defaults=True)
     with open(USER_HELM_APPS_FILE, "w") as f:
         f.write(helm_yaml)
@@ -864,11 +874,6 @@ def create_pool(
         )
     except Exception as e:
         return {"error": f"Error when updating dependencies: {str(e)}"}
-
-    if location is not None:
-        # TODO: register with kalavai if it's a public cluster
-        pass
-        #pool__publish()
     
     # wait until the server is ready to create objects
     while True:
@@ -876,7 +881,7 @@ def create_pool(
         if is_watcher_alive(server_creds=USER_LOCAL_SERVER_FILE, user_cookie=USER_COOKIE):
             break
 
-    result = pool_init(pool_config_values_path=pool_config_values)
+    result = pool_init(config_values=post_config_values)
     if "error" in result or ("failed" in result and len(result['failed']) > 0):
         return {"error": f"Error when initialising pool: {result}"}
     # init default namespace
@@ -953,16 +958,15 @@ def get_pool_token(mode: TokenType):
     except Exception as e:
         return {"error": f"Error when generating token: {str(e)}"}
 
-def pool_init(pool_config_values_path=None):
+def pool_init(config_values=None):
     """Deploy configured objects to initialise pool"""
-    if pool_config_values_path is None:
+    if config_values is None:
         return
     
     # load template config and populate with values
     sidecar_template_yaml = load_template(
         template_path=POOL_CONFIG_TEMPLATE,
-        values={},
-        default_values_path=pool_config_values_path)
+        values=config_values)
 
     try:
         result = request_to_server(

{kalavai_client-0.7.1 → kalavai_client-0.7.3}/kalavai_client/env.py

@@ -51,6 +51,7 @@ HELM_APPS_FILE = resource_path("kalavai_client/assets/apps.yaml")
 HELM_APPS_VALUES = resource_path("kalavai_client/assets/apps_values.yaml")
 DOCKER_COMPOSE_TEMPLATE = resource_path("kalavai_client/assets/docker-compose-template.yaml")
 DOCKER_COMPOSE_GUI = resource_path("kalavai_client/assets/docker-compose-gui.yaml")
+DEFAULT_POOL_CONFIG_TEMPLATE = resource_path("kalavai_client/assets/default_pool_config.yaml")
 USER_WORKSPACE_TEMPLATE = resource_path("kalavai_client/assets/user_workspace.yaml")
 DEFAULT_USER_WORKSPACE_VALUES = resource_path("kalavai_client/assets/user_workspace_values.yaml")
 POOL_CONFIG_TEMPLATE = resource_path("kalavai_client/assets/pool_config_template.yaml")

{kalavai_client-0.7.1 → kalavai_client-0.7.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name            = "kalavai-client"
-version         = "0.7.1"
+version         = "0.7.3"
 authors = [
   {name = "Carlos Fernandez Musoles", email = "carlos@kalavai.net"}
 ]

kalavai_client-0.7.1/kalavai_client/__init__.py

@@ -1,2 +0,0 @@
-
-__version__ = "0.7.1"

kalavai_client-0.7.1/kalavai_client/assets/apps_values.yaml

@@ -1,106 +0,0 @@
-### APS ###
-- name: deploy_lago
-  default: "False"
-  description: "Deploy Lago payment system"
-
-- name: deploy_opencost
-  default: "False"
-  description: "Deploy Opencost cost monitoring system"
-
-- name: deploy_prometheus
-  default: "False"
-  description: "Deploy Prometheus system monitoring system"
-
-- name: deploy_langfuse
-  default: "False"
-  description: "Deploy Langfuse LLM tracing system"
-
-- name: deploy_kuberay
-  default: "False"
-  description: "Deploy Langfuse LLM tracing system"
-
-######
-
-### VARIABLES ###
-
-- 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-0.7.1/kalavai_client/assets/pool_config_values.yaml

@@ -1,4 +0,0 @@
-- name: storage_class_name
-  default: "longhorn"
-  description: ""
-