kalavai-client 0.6.20__tar.gz → 0.6.22__tar.gz

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: kalavai-client
-Version: 0.6.20
+Version: 0.6.22
 Summary: Client app for kalavai platform
 License: Apache-2.0
 Keywords: LLM,platform

kalavai_client-0.6.22/kalavai_client/__init__.py

@@ -0,0 +1,2 @@
+
+__version__ = "0.6.22"

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/kalavai_client/assets/apps.yaml

@@ -20,6 +20,8 @@ repositories:
     url: https://charts.min.io/
   - name: langfuse
     url: https://langfuse.github.io/langfuse-k8s
+  - name: hami-charts
+    url: https://project-hami.github.io/HAMi
 
 releases:
   - name: datashim
@@ -154,7 +156,7 @@ releases:
     - name: replicas
       value: 1
     - name: image_tag
-      value: "v2025.07.33"
+      value: "v2025.07.34"
     - name: deployment.in_cluster
       value: "True"
     - name: deployment.kalavai_username_key
@@ -185,9 +187,30 @@ releases:
     namespace: kalavai
     chart: kalavai/gpu
     installed: false    
+  # - 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: hami-vgpu
     namespace: kalavai
-    chart: kalavai/hami
+    chart: hami-charts/hami
     installed: true
     set:
     - name: resourceCores
@@ -206,13 +229,5 @@ releases:
       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-0.6.20 → kalavai_client-0.6.22}/kalavai_client/assets/docker-compose-template.yaml

@@ -3,7 +3,7 @@ services:
   {{vpn_name}}:
     image: gravitl/netclient:v0.90.0
     container_name: {{vpn_name}}
-    platform: linux/amd64
+    platform: linux/{{target_platform}}
     cap_add:
       - NET_ADMIN
       - SYS_MODULE
@@ -21,6 +21,7 @@ services:
     image: docker.io/bundenth/kalavai-runner:{{target_platform}}-latest
     pull_policy: always
     container_name: {{service_name}}
+    platform: linux/{{target_platform}}
   {% if vpn %}
     depends_on:
     - {{vpn_name}}

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/kalavai_client/bridge_api.py

@@ -9,6 +9,10 @@ from starlette.requests import Request
 import uvicorn
 
 from kalavai_client.core import Job
+from kalavai_client.env import (
+    KALAVAI_SERVICE_LABEL,
+    KALAVAI_SERVICE_LABEL_VALUE
+)
 from kalavai_client.bridge_models import (
     CreatePoolRequest,
     InvitesRequest,
@@ -18,7 +22,8 @@ from kalavai_client.bridge_models import (
     DeleteJobRequest,
     JobDetailsRequest,
     NodesActionRequest,
-    NodeLabelsRequest
+    NodeLabelsRequest,
+    WorkerConfigRequest
 )
 from kalavai_client.core import (
     create_pool,
@@ -34,6 +39,7 @@ from kalavai_client.core import (
     fetch_job_logs,
     fetch_job_templates,
     fetch_job_defaults,
+    fetch_pod_logs,
     deploy_job,
     delete_job,
     authenticate_user,
@@ -52,6 +58,7 @@ from kalavai_client.core import (
     uncordon_nodes,
     add_node_labels,
     get_node_labels,
+    generate_worker_package,
     TokenType
 )
 from kalavai_client.utils import (
@@ -241,6 +248,21 @@ def get_token(mode: int, api_key: str = Depends(verify_api_key)):
     """
     return get_pool_token(mode=TokenType(mode))
 
+@app.post("/generate_worker_config",
+    operation_id="generate_worker_config",
+    summary="Generate a config file for a remote worker to connect to the pool",
+    description="Generate a config file for a remote worker to connect to the pool. Different token types provide different levels of access - join tokens allow nodes to contribute resources, while attach tokens allow management access.",
+    tags=["pool_management"],
+    response_description="Worker config file")
+def generate_worker_config(request: WorkerConfigRequest, api_key: str = Depends(verify_api_key)):
+    return generate_worker_package(
+        node_name=request.node_name,
+        mode=TokenType(request.mode),
+        target_platform=request.target_platform,
+        num_gpus=request.num_gpus,
+        ip_address=request.ip_address,
+        storage_compatible=request.storage_compatible)
+
 @app.get("/fetch_devices",
     operation_id="fetch_devices",
     summary="Get list of all compute devices in the pool",
@@ -251,6 +273,15 @@ def get_devices(api_key: str = Depends(verify_api_key)):
     """Get list of available devices"""
     return fetch_devices()
 
+@app.get("/fetch_service_logs",
+    operation_id="fetch_service_logs",
+    summary="Get logs for the kalavai API service",
+    description="Get logs for the kalavai API service, including internal logs, debugging messages and status of the service.",
+    tags=["info"],
+    response_description="Logs")
+def get_service_logs(api_key: str = Depends(verify_api_key)):
+    return fetch_pod_logs(label_key=KALAVAI_SERVICE_LABEL, label_value=KALAVAI_SERVICE_LABEL_VALUE, force_namespace="kalavai")
+
 @app.post("/send_pool_invites",
     operation_id="send_pool_invites",
     summary="Send invitations to join the pool",

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/kalavai_client/bridge_models.py

@@ -17,6 +17,15 @@ class CreatePoolRequest(BaseModel):
     token_mode: TokenType = Field(TokenType.USER, description="Token type for authentication")
     description: str = Field("", description="Description of the pool")
 
+class WorkerConfigRequest(BaseModel):
+    node_name: str = Field(None, description="Name for the worker node")
+    mode: int = Field(2, description="Access mode for the worker (admin, worker or user)")
+    target_platform: str = Field("amd64", description="Target platform architecture for the worker (amd64 or arm64)")
+    num_gpus: int = Field(0, description="Number of GPUs to use on the worker node")
+    ip_address: str = Field("0.0.0.0", description="IP address of the worker node")
+    storage_compatible: bool = Field(True, description="Whether to use the node's storage capacity for volumes")
+
+
 class NodesActionRequest(BaseModel):
     nodes: list[str] = Field(None, description="List of node names to perform the action on, defaults to None")
 

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/kalavai_client/cli.py

@@ -27,6 +27,8 @@ from kalavai_client.env import (
     USER_TEMPLATES_FOLDER,
     DOCKER_COMPOSE_GUI,
     USER_GUI_COMPOSE_FILE,
+    KALAVAI_SERVICE_LABEL,
+    KALAVAI_SERVICE_LABEL_VALUE,
     user_path,
     resource_path,
 )
@@ -37,6 +39,7 @@ from kalavai_client.core import (
     fetch_job_details,
     fetch_devices,
     fetch_job_logs,
+    fetch_pod_logs,
     fetch_gpus,
     generate_worker_package,
     load_gpu_models,
@@ -689,6 +692,25 @@ def pool__update(*others):
     else:
         console.log(f"[green]{result}")
 
+@arguably.command
+def pool__logs(*others):
+    """
+    Get the logs for the Kalavai API
+    """
+    logs = []
+
+    logs.append("Getting Kalavai API logs...")
+
+    logs = fetch_pod_logs(
+        label_key=KALAVAI_SERVICE_LABEL,
+        label_value=KALAVAI_SERVICE_LABEL_VALUE,
+        force_namespace="kalavai"
+    )  
+    for name, log in logs.items():
+        console.log(f"[yellow]LOGS for service: {name}")
+        for key, value in log.items():
+            console.log(f"[yellow]{key}")
+            console.log(json.dumps(value, indent=2))
 
 @arguably.command
 def pool__status(*others, log_file=None):

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/kalavai_client/core.py

@@ -313,7 +313,7 @@ def fetch_job_details(jobs: list[Job]):
                         restart_counts = sum([c["restart_count"] for c in values["conditions"]])
                     workers_status[values["status"]] += 1
                     # get nodes involved in deployment (needs kubewatcher)
-                    if "node_name" in values:
+                    if "node_name" in values and values["node_name"] is not None:
                         host_nodes.add(values["node_name"])
 
             workers = "\n".join([f"{k}: {v}" for k, v in workers_status.items()])
@@ -454,9 +454,18 @@ def fetch_devices():
         return {"error": str(e)}
 
 def fetch_job_logs(job_name, force_namespace=None, pod_name=None, tail=100):
+    return fetch_pod_logs(
+        label_key=TEMPLATE_LABEL,
+        label_value=job_name,
+        pod_name=pod_name,
+        force_namespace=force_namespace,
+        tail=tail
+    )
+
+def fetch_pod_logs(label_key, label_value, force_namespace=None, pod_name=None, tail=100):
     data = {
-        "label": TEMPLATE_LABEL,
-        "value": job_name,
+        "label": label_key,
+        "value": label_value,
         "tail": tail
     }
     if force_namespace is not None:
@@ -609,9 +618,16 @@ def attach_to_pool(token, node_name=None):
 
     return cluster_name
 
-def generate_worker_package(target_platform="amd64", num_gpus=0, node_name=None, ip_address="0.0.0.0", storage_compatible=True):
+def generate_worker_package(
+        target_platform="amd64",
+        num_gpus=0,
+        node_name=None,
+        ip_address="0.0.0.0",
+        storage_compatible=True,
+        mode=TokenType.WORKER
+):
     # get pool data from token  
-    token = get_pool_token(mode=TokenType.WORKER)
+    token = get_pool_token(mode=mode)
     if "error" in token:
         return {"error": f"[red]Error when getting pool token: {token['error']}"}
     

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/kalavai_client/env.py

@@ -32,6 +32,8 @@ def resource_path(relative_path: str):
 
 
 TEMPLATE_LABEL = "kalavai.job.name"
+KALAVAI_SERVICE_LABEL = "app"
+KALAVAI_SERVICE_LABEL_VALUE = "kube-watcher-api"
 STORAGE_CLASS_LABEL = "kalavai.storage.enabled"
 USER_NODE_LABEL = "kalavai.cluster.user"
 SERVER_IP_KEY = "server_ip"

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/kalavai_client/utils.py

@@ -175,7 +175,6 @@ def is_watcher_alive(server_creds, user_cookie, timeout=30):
             timeout=timeout
         )
     except Exception as e:
-        print(str(e))
         return False
     return True
 

{kalavai_client-0.6.20 → kalavai_client-0.6.22}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name            = "kalavai-client"
-version         = "0.6.20"
+version         = "0.6.22"
 authors = [
   {name = "Carlos Fernandez Musoles", email = "carlos@kalavai.net"}
 ]

kalavai_client-0.6.20/kalavai_client/__init__.py

@@ -1,2 +0,0 @@
-
-__version__ = "0.6.20"