kalavai-client 0.7.5__py3-none-any.whl → 0.7.7__py3-none-any.whl

kalavai_client/__init__.py

@@ -1,2 +1,2 @@
 
-__version__ = "0.7.5"
+__version__ = "0.7.7"

kalavai_client/assets/apps.yaml

@@ -128,6 +128,8 @@ releases:
       value: {{prometheus_server_retention}}
     - name: server.persistentVolume.size
       value: {{prometheus_disk_size}}
+    - name: prometheus-node-exporter.hostRootFsMount.enabled
+      value: false
   - name: volcano-sh
     namespace: kalavai
     chart: volcano-sh/volcano
@@ -172,7 +174,7 @@ releases:
     - name: replicas
       value: 1
     - name: image_tag
-      value: "v2025.09.2" #"v2025.07.34"
+      value: "v2025.10.2" #"v2025.07.34"
     - name: deployment.in_cluster
       value: "True"
     - name: deployment.kalavai_username_key

kalavai_client/bridge_api.py

@@ -59,7 +59,6 @@ from kalavai_client.core import (
     add_node_labels,
     get_node_labels,
     generate_worker_package,
-    get_deployment_values,
     TokenType
 )
 from kalavai_client.utils import (
@@ -447,20 +446,6 @@ def job_deploy(request: DeployJobRequest, api_key: str = Depends(verify_api_key)
     )
     return result
 
-@app.get("/get_deployment_values",
-    operation_id="get_deployment_values",
-    summary="Get deployment template values for a given model",
-    description="Given a model id from Huggingface, return the deployment template values required to load the model instance in the pool, including number of workers, number of gpus and gpu backend.",
-    tags=["job_management"],
-    response_description="Deployment template values")
-def get_deployment_template_values(model_id: str, api_key: str = Depends(verify_api_key)):
-    """
-    Get the deployment template values for a given model id:
-    
-    - **model_id**: Model id from Huggingface type mode
-    """
-    return get_deployment_values(model_id=model_id)
-
 @app.post("/delete_job",
     operation_id="delete_job",
     summary="Terminate and remove a job from the pool",

kalavai_client/cli.py

@@ -6,6 +6,7 @@ import uuid
 import time
 import socket
 from pathlib import Path
+from typing import Annotated
 
 import yaml
 
@@ -62,8 +63,7 @@ from kalavai_client.core import (
     uncordon_nodes,
     TokenType,
     unregister_pool,
-    update_pool,
-    get_deployment_values
+    update_pool
 )
 from kalavai_client.utils import (
     check_gpu_drivers,
@@ -78,7 +78,8 @@ from kalavai_client.utils import (
     load_user_id,
     SERVER_IP_KEY,
     CLUSTER_NAME_KEY,
-    KALAVAI_AUTH
+    KALAVAI_AUTH,
+    parse_key_value_pairs
 )
 
 
@@ -397,7 +398,19 @@ def pool__list(*others, user_only=False):
 
 
 @arguably.command
-def pool__start(*others,  pool_config_file=None, apps: list=None, mtu: str=None, 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,
+    mtu: str=None,
+    platform="amd64",
+    ip_address: str=None,
+    location: str=None,
+    app_values: str=None,
+    pool_config_values: str=None,
+    non_interactive: bool=False,
+    node_labels: Annotated[dict, arguably.arg.handler(parse_key_value_pairs)] = {}
+):
 
     """
     Start Kalavai pool and start/resume sharing resources.
@@ -409,6 +422,9 @@ def pool__start(*others,  pool_config_file=None, apps: list=None, mtu: str=None,
     if CLUSTER.is_cluster_init():
         console.log(f"[white] You are already connected to {load_server_info(data_key=CLUSTER_NAME_KEY, file=USER_LOCAL_SERVER_FILE)}. Enter [yellow]kalavai pool stop[white] to exit and join another one.")
         return
+
+    if node_labels:
+        console.log(f"[blue]Configuration received: {node_labels}")
     
     # User acknowledgement
     if not non_interactive:
@@ -441,7 +457,8 @@ def pool__start(*others,  pool_config_file=None, apps: list=None, mtu: str=None,
         pool_config_file=pool_config_file,
         apps=apps,
         num_gpus=input_gpus(non_interactive=non_interactive),
-        mtu=mtu
+        mtu=mtu,
+        node_labels=node_labels
     )
 
     if "warning" in result:
@@ -499,14 +516,32 @@ def pool__check_token(token, *others, public=False, verbose=False):
     return True
 
 @arguably.command
-def pool__join(token, *others, mtu=None, platform="amd64", node_name=None, non_interactive=False):
+def pool__join(
+    token,
+    *others,
+    mtu=None,
+    platform="amd64",
+    node_name=None,
+    non_interactive=False,
+    node_labels: Annotated[dict, arguably.arg.handler(parse_key_value_pairs)] = {}
+):
     """
     Join Kalavai pool and start/resume sharing resources.
 
     Args:
+        token: Pool join token
         *others: all the other positional arguments go here
+        mtu: Maximum transmission unit
+        platform: Target platform (default: amd64)
+        node_name: Name for this node
+        non_interactive: Run in non-interactive mode
+        node_labels: Node labels as key=value pairs (e.g., "key1=value1,key2=value2")
     """
     
+    # Process node labels if provided
+    if node_labels:
+        console.log(f"[blue]Configuration received: {node_labels}")
+    
     # check that k3s is not running already in the host
     # k3s service running or preinstalled
     if CLUSTER.is_agent_running():
@@ -554,7 +589,8 @@ def pool__join(token, *others, mtu=None, platform="amd64", node_name=None, non_i
         node_name=node_name,
         num_gpus=num_gpus,
         ip_address=ip_address,
-        mtu=mtu
+        mtu=mtu,
+        node_labels=node_labels
     )
     if "error" in result:
         console.log(f"[red]Error when connecting: {result}")
@@ -1170,12 +1206,6 @@ def job__delete(name, *others, force_namespace: str=None):
         console.log(f"{result}")
 
 
-@arguably.command
-def job__model_requirements(model_id: str, *others):
-    values = get_deployment_values(model_id=model_id)
-    console.log(values)
-
-
 @arguably.command
 def job__estimate(
     *others, 

kalavai_client/core.py

@@ -104,70 +104,6 @@ class TokenType(Enum):
     WORKER = 2
 
 
-def get_deployment_values(model_id: str):
-    """
-    Given a model ID and the resources in the pool, identify key
-    computing values required to deploy the model.
-    - GPU_BACKEND: rocm or cuda
-    - WORKERS: number of nodes to use
-    - 
-    """
-    # get hardcoded deployment values (per model)
-    with open(MODEL_DEPLOYMENT_VALUES_MAPPING, "r") as f:
-        mapping = yaml.safe_load(f)
-
-    def _parse_memory_str(memory: str):
-        memory = memory.replace("G", "")
-        return int(memory)
-    
-    def _get_num_workers(memory_values: list[int], size):
-        workers = 0
-        available_memory = 0
-        for gpu_mem in memory_values:
-            available_memory += gpu_mem
-            workers += 1
-            if available_memory >= size:
-                break
-        return workers
-    
-    # get resources
-    if model_id in mapping:
-        model_size = mapping[model_id]["size"]
-        # get gpus and extract available memory
-        nvidia_gpu_mems = []
-        amd_gpu_mems = []
-        backends = set()
-        for node_name, gpus in load_gpu_models():
-            for gpu in gpus["gpus"]:
-                if "nvidia" in gpu["model"].lower():
-                    nvidia_gpu_mems.append(_parse_memory_str(gpu["memory"]))
-                    backends.add("cuda")
-                else:
-                    amd_gpu_mems.append(_parse_memory_str(gpu["memory"]))
-                    backends.add("rocm")
-        nvidia_gpu_mems = sorted(nvidia_gpu_mems, reverse=False)
-        amd_gpu_mems = sorted(amd_gpu_mems, reverse=False)
-        # calculate num workers required
-        if sum(nvidia_gpu_mems) >= model_size and sum(amd_gpu_mems) < model_size:
-            gpu_backend = "cuda"
-            num_workers = _get_num_workers(memory_values=nvidia_gpu_mems, size=model_size)
-        elif sum(amd_gpu_mems) >= model_size and sum(nvidia_gpu_mems) < model_size:
-            gpu_backend = "rocm"
-            num_workers = _get_num_workers(memory_values=amd_gpu_mems, size=model_size)
-        else:
-            gpu_backend = random.choice(list(backends))
-            num_workers = _get_num_workers(
-                memory_values=amd_gpu_mems if gpu_backend == "rocm" else nvidia_gpu_mems,
-                size=model_size
-            )
-        # populate selected template
-        mapping[model_id][gpu_backend]["values"]["workers"] = num_workers
-        mapping[model_id][gpu_backend]["values"]["pipeline_parallel_size"] = num_workers
-
-        return mapping[model_id][gpu_backend]
-    return None
-
-
 def set_schedulable(schedulable, node_names):
     """
     Delete job in the cluster
@@ -735,7 +671,8 @@ def join_pool(
         node_name=None,
         ip_address=None,
         target_platform="amd64",
-        mtu="1420"
+        mtu="1420",
+        node_labels={}
 ):
     compatibility = check_worker_compatibility()
     if len(compatibility["issues"]) > 0:
@@ -746,6 +683,9 @@ def join_pool(
 
     if node_name is None:
         node_name = socket.gethostname()
+
+    if mtu is None:
+        mtu = "1420"
     
     # check token
     valid = check_token(token=token)
@@ -765,6 +705,7 @@ def join_pool(
     
     # join private network if provided
     node_labels = {
+        **node_labels,
         STORAGE_CLASS_LABEL: is_storage_compatible(),
         NODE_ROLE_LABEL: "worker"
     }  
@@ -829,7 +770,8 @@ def create_pool(
         num_gpus: int=-1,
         node_name: str=None,
         mtu: str=None,
-        apps: list=[]
+        apps: list=[],
+        node_labels: dict={}
     ):
 
     if not check_seed_compatibility():
@@ -844,6 +786,7 @@ def create_pool(
     user_id = load_user_id()
     
     node_labels = {
+        **node_labels,
         STORAGE_CLASS_LABEL: is_storage_compatible(),
         NODE_ROLE_LABEL: "server"
     }

kalavai_client/utils.py

@@ -106,6 +106,32 @@ def is_storage_compatible():
         return False
 ################
 
+def parse_key_value_pairs(input_str: str) -> dict:
+    """Parse key=value pairs from a string into a dictionary.
+    
+    Args:
+        input_str: String containing key=value pairs separated by commas
+        
+    Returns:
+        Dictionary with parsed key-value pairs
+        
+    Raises:
+        ValueError: If any pair is not in key=value format
+    """
+    if not input_str.strip():
+        return {}
+    
+    result = {}
+    for pair in input_str.split(','):
+        pair = pair.strip()
+        if not pair:
+            continue
+        if '=' not in pair:
+            raise ValueError(f"Invalid key=value pair: '{pair}'. Expected format: key=value")
+        key, value = pair.split('=', 1)
+        result[key.strip()] = value.strip()
+    return result
+
 def extract_auth_token(headers):
     """
     Extract auth token. Valid headers:

{kalavai_client-0.7.5.dist-info → kalavai_client-0.7.7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kalavai-client
-Version: 0.7.5
+Version: 0.7.7
 Summary: Client app for kalavai platform
 License-Expression: Apache-2.0
 License-File: LICENSE

{kalavai_client-0.7.5.dist-info → kalavai_client-0.7.7.dist-info}/RECORD

@@ -1,7 +1,7 @@
-kalavai_client/__init__.py,sha256=2yNV33mxa4mw0NRZdX1rF9R0X_uQaQl59WtFaMqS_ss,22
+kalavai_client/__init__.py,sha256=_he-4_uHDJNaqLXyzYdBG_rgSv3K0P5DCywjUCh7v5c,22
 kalavai_client/__main__.py,sha256=WQUfxvRsBJH5gsCJg8pLz95QnZIj7Ol8psTO77m0QE0,73
 kalavai_client/assets/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-kalavai_client/assets/apps.yaml,sha256=z5lLrBbR8KVP5Ou3pbD6cELYWATs9VJuRu-PPe1GMNk,7007
+kalavai_client/assets/apps.yaml,sha256=juO6kIl8LjykWI-BPsALSdADc9cDE73t8wP2W6ll2co,7087
 kalavai_client/assets/default_pool_config.yaml,sha256=FLfT0i-xPgEZhleTOZmfyFYrB9fAEMOmZ0YGLyJr0OA,1879
 kalavai_client/assets/docker-compose-gui.yaml,sha256=OAVO0ohaCpDB9FGeih0yAbVNwUfDtaCzssZ25uiuJyA,787
 kalavai_client/assets/docker-compose-template.yaml,sha256=8giwNUVbVGxF_YppNMIsS7evWZ8qb4n2enK22u-x6Pk,1920
@@ -11,15 +11,15 @@ kalavai_client/assets/pool_config_template.yaml,sha256=MhBZQsEMKrBgbUVSKgIGmXWhy
 kalavai_client/assets/user_workspace.yaml,sha256=wDvlMYknOPABAEo0dsQwU7bac8iubjAG9tdkFbJZ5Go,476
 kalavai_client/assets/user_workspace_values.yaml,sha256=G0HOzQUxrDMCwuW9kbWUZaKMzDDPVwDwzBHCL2Xi2ZM,542
 kalavai_client/auth.py,sha256=EB3PMvKUn5_KAQkezkEHEt-OMZXyfkZguIQlUFkEHcA,3243
-kalavai_client/bridge_api.py,sha256=6NGRIbu3CtWyT75XCFKkLSfllF54vGJmYBjnX9qiWX8,27657
+kalavai_client/bridge_api.py,sha256=0TvAGgsyfMkbcVqoPKsjhXQLo06WetBH93mZH-pOM7U,26921
 kalavai_client/bridge_models.py,sha256=bq6vQNTI1py7e_1YgnBZhorFsAKoBqBVN7nRukCuQRE,2960
-kalavai_client/cli.py,sha256=7EdEHjJxoS0lgmWFyT72k7rl3-qJ7ODz9ju_wZGYB1I,48312
+kalavai_client/cli.py,sha256=yGCOS-eI01E_EdxWFG9IgUZnSQyjaEV8NqrX2M-YcZw,49010
 kalavai_client/cluster.py,sha256=Z2PIXbZuSAv9xmw-MyZP1M41BpVMpirLzG51bqGA-zc,13548
-kalavai_client/core.py,sha256=BUlCOeQnXTAqwjD4rxWIshk9Sf_s-QJy9UB6cUbcECo,38906
+kalavai_client/core.py,sha256=ElDeF0AjzoZaECUg1RbwolJItuX6b71i03Gr9gIXLfY,36529
 kalavai_client/env.py,sha256=0L5gfEo5KY8gflrW-rSADx10ffDa-8gXmmrGWztKUd8,3099
-kalavai_client/utils.py,sha256=L0FQJT8EucabPAbxGwzMqeawJdj4wTZqmmZ0OdsyVwM,13470
-kalavai_client-0.7.5.dist-info/METADATA,sha256=u04xcVwDMzNCp-xU_6TE791B5xQlSc2jnqnzRB587vI,13175
-kalavai_client-0.7.5.dist-info/WHEEL,sha256=zp0Cn7JsFoX2ATtOhtaFYIiE2rmFAD4OcMhtUki8W3U,88
-kalavai_client-0.7.5.dist-info/entry_points.txt,sha256=9T6D45gxwzfVbglMm1r6XPdXuuZdHfy_7fCeu2jUphc,50
-kalavai_client-0.7.5.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-kalavai_client-0.7.5.dist-info/RECORD,,
+kalavai_client/utils.py,sha256=swMB-3elI20GihzZgqSFD5sQfQ-Nh2_TGbEnKNHx7EU,14230
+kalavai_client-0.7.7.dist-info/METADATA,sha256=o6ywphe6vy6cDQI7RuRhBFjcW5-KZOYV_xMCFrPmpMc,13175
+kalavai_client-0.7.7.dist-info/WHEEL,sha256=zp0Cn7JsFoX2ATtOhtaFYIiE2rmFAD4OcMhtUki8W3U,88
+kalavai_client-0.7.7.dist-info/entry_points.txt,sha256=9T6D45gxwzfVbglMm1r6XPdXuuZdHfy_7fCeu2jUphc,50
+kalavai_client-0.7.7.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+kalavai_client-0.7.7.dist-info/RECORD,,