skypilot-nightly 1.0.0.dev20251011__py3-none-any.whl → 1.0.0.dev20251013__py3-none-any.whl

sky/server/requests/requests.py

@@ -292,6 +292,100 @@ class Request:
             raise
 
 
+def encode_requests(requests: List[Request]) -> List[payloads.RequestPayload]:
+    """Serialize the SkyPilot API request for display purposes.
+
+        This function should be called on the server side to serialize the
+        request body into human readable format, e.g., the entrypoint should
+        be a string, and the pid, error, or return value are not needed.
+
+        The returned value will then be displayed on the client side in request
+        table.
+
+        We do not use `encode` for display to avoid a large amount of data being
+        sent to the client side, especially for the request table could include
+        all the requests.
+        """
+    encoded_requests = []
+    all_users = global_user_state.get_all_users()
+    all_users_map = {user.id: user.name for user in all_users}
+    for request in requests:
+        if request.request_body is not None:
+            assert isinstance(request.request_body,
+                              payloads.RequestBody), (request.name,
+                                                      request.request_body)
+        user_name = all_users_map.get(request.user_id)
+        payload = payloads.RequestPayload(
+            request_id=request.request_id,
+            name=request.name,
+            entrypoint=request.entrypoint.__name__
+            if request.entrypoint is not None else '',
+            request_body=request.request_body.model_dump_json()
+            if request.request_body is not None else json.dumps(None),
+            status=request.status.value,
+            return_value=json.dumps(None),
+            error=json.dumps(None),
+            pid=None,
+            created_at=request.created_at,
+            schedule_type=request.schedule_type.value,
+            user_id=request.user_id,
+            user_name=user_name,
+            cluster_name=request.cluster_name,
+            status_msg=request.status_msg,
+            should_retry=request.should_retry,
+            finished_at=request.finished_at,
+        )
+        encoded_requests.append(payload)
+    return encoded_requests
+
+
+def _update_request_row_fields(
+        row: Tuple[Any, ...],
+        fields: Optional[List[str]] = None) -> Tuple[Any, ...]:
+    """Update the request row fields."""
+    if not fields:
+        return row
+
+    # Convert tuple to dictionary for easier manipulation
+    content = dict(zip(fields, row))
+
+    # Required fields in RequestPayload
+    if 'request_id' not in fields:
+        content['request_id'] = ''
+    if 'name' not in fields:
+        content['name'] = ''
+    if 'entrypoint' not in fields:
+        content['entrypoint'] = server_constants.EMPTY_PICKLED_VALUE
+    if 'request_body' not in fields:
+        content['request_body'] = server_constants.EMPTY_PICKLED_VALUE
+    if 'status' not in fields:
+        content['status'] = RequestStatus.PENDING.value
+    if 'created_at' not in fields:
+        content['created_at'] = 0
+    if 'user_id' not in fields:
+        content['user_id'] = ''
+    if 'return_value' not in fields:
+        content['return_value'] = json.dumps(None)
+    if 'error' not in fields:
+        content['error'] = json.dumps(None)
+    if 'schedule_type' not in fields:
+        content['schedule_type'] = ScheduleType.SHORT.value
+    # Optional fields in RequestPayload
+    if 'pid' not in fields:
+        content['pid'] = None
+    if 'cluster_name' not in fields:
+        content['cluster_name'] = None
+    if 'status_msg' not in fields:
+        content['status_msg'] = None
+    if 'should_retry' not in fields:
+        content['should_retry'] = False
+    if 'finished_at' not in fields:
+        content['finished_at'] = None
+
+    # Convert back to tuple in the same order as REQUEST_COLUMNS
+    return tuple(content[col] for col in REQUEST_COLUMNS)
+
+
 def kill_cluster_requests(cluster_name: str, exclude_request_name: str):
     """Kill all pending and running requests for a cluster.
 
@@ -634,6 +728,7 @@ class RequestTaskFilter:
             Mutually exclusive with exclude_request_names.
         finished_before: if provided, only include requests finished before this
             timestamp.
+        limit: the number of requests to show. If None, show all requests.
 
     Raises:
         ValueError: If both exclude_request_names and include_request_names are
@@ -645,6 +740,8 @@ class RequestTaskFilter:
     exclude_request_names: Optional[List[str]] = None
     include_request_names: Optional[List[str]] = None
     finished_before: Optional[float] = None
+    limit: Optional[int] = None
+    fields: Optional[List[str]] = None
 
     def __post_init__(self):
         if (self.exclude_request_names is not None and
@@ -687,8 +784,13 @@ class RequestTaskFilter:
         if filter_str:
             filter_str = f' WHERE {filter_str}'
         columns_str = ', '.join(REQUEST_COLUMNS)
-        return (f'SELECT {columns_str} FROM {REQUEST_TABLE}{filter_str} '
-                'ORDER BY created_at DESC'), filter_params
+        if self.fields:
+            columns_str = ', '.join(self.fields)
+        query_str = (f'SELECT {columns_str} FROM {REQUEST_TABLE}{filter_str} '
+                     'ORDER BY created_at DESC')
+        if self.limit is not None:
+            query_str += f' LIMIT {self.limit}'
+        return query_str, filter_params
 
 
 @init_db
@@ -722,6 +824,21 @@ async def get_request_tasks_async(
     return [Request.from_row(row) for row in rows]
 
 
+@init_db_async
+@metrics_lib.time_me_async
+async def get_request_tasks_with_fields_async(
+    req_filter: RequestTaskFilter,
+    fields: Optional[List[str]] = None,
+) -> List[Request]:
+    """Async version of get_request_tasks."""
+    assert _DB is not None
+    async with _DB.execute_fetchall_async(*req_filter.build_query()) as rows:
+        if not rows:
+            return []
+    rows = [_update_request_row_fields(row, fields) for row in rows]
+    return [Request.from_row(row) for row in rows]
+
+
 @init_db_async
 @metrics_lib.time_me_async
 async def get_api_request_ids_start_with(incomplete: str) -> List[str]:

sky/server/server.py

@@ -3,6 +3,7 @@
 import argparse
 import asyncio
 import base64
+from concurrent.futures import ThreadPoolExecutor
 import contextlib
 import datetime
 import hashlib
@@ -1629,6 +1630,10 @@ async def api_status(
         None, description='Request IDs to get status for.'),
     all_status: bool = fastapi.Query(
         False, description='Get finished requests as well.'),
+    limit: Optional[int] = fastapi.Query(
+        None, description='Number of requests to show.'),
+    fields: Optional[List[str]] = fastapi.Query(
+        None, description='Fields to get. If None, get all fields.'),
 ) -> List[payloads.RequestPayload]:
     """Gets the list of requests."""
     if request_ids is None:
@@ -1638,9 +1643,15 @@ async def api_status(
                 requests_lib.RequestStatus.PENDING,
                 requests_lib.RequestStatus.RUNNING,
             ]
-        request_tasks = await requests_lib.get_request_tasks_async(
-            req_filter=requests_lib.RequestTaskFilter(status=statuses))
-        return [r.readable_encode() for r in request_tasks]
+        request_tasks = await requests_lib.get_request_tasks_with_fields_async(
+            req_filter=requests_lib.RequestTaskFilter(
+                status=statuses,
+                limit=limit,
+                fields=fields,
+            ),
+            fields=fields,
+        )
+        return requests_lib.encode_requests(request_tasks)
     else:
         encoded_request_tasks = []
         for request_id in request_ids:
@@ -1721,9 +1732,9 @@ async def kubernetes_pod_ssh_proxy(websocket: fastapi.WebSocket,
     logger.info(f'WebSocket connection accepted for cluster: {cluster_name}')
 
     # Run core.status in another thread to avoid blocking the event loop.
-    cluster_records = await context_utils.to_thread(core.status,
-                                                    cluster_name,
-                                                    all_users=True)
+    with ThreadPoolExecutor(max_workers=1) as thread_pool_executor:
+        cluster_records = await context_utils.to_thread_with_executor(
+            thread_pool_executor, core.status, cluster_name, all_users=True)
     cluster_record = cluster_records[0]
     if cluster_record['status'] != status_lib.ClusterStatus.UP:
         raise fastapi.HTTPException(

sky/setup_files/dependencies.py

@@ -222,6 +222,7 @@ extras_require: Dict[str, List[str]] = {
     'hyperbolic': [],  # No dependencies needed for hyperbolic
     'seeweb': ['ecsapi>=0.2.0'],
     'server': server_dependencies,
+    'shadeform': [],  # No dependencies needed for shadeform
 }
 
 # Calculate which clouds should be included in the [all] installation.

sky/skylet/constants.py

@@ -471,7 +471,7 @@ CATALOG_DIR = '~/.sky/catalogs'
 ALL_CLOUDS = ('aws', 'azure', 'gcp', 'ibm', 'lambda', 'scp', 'oci',
               'kubernetes', 'runpod', 'vast', 'vsphere', 'cudo', 'fluidstack',
               'paperspace', 'primeintellect', 'do', 'nebius', 'ssh',
-              'hyperbolic', 'seeweb')
+              'hyperbolic', 'seeweb', 'shadeform')
 # END constants used for service catalog.
 
 # The user ID of the SkyPilot system.

sky/templates/shadeform-ray.yml.j2

@@ -0,0 +1,72 @@
+cluster_name: {{cluster_name_on_cloud}}
+
+# The maximum number of workers nodes to launch in addition to the head node.
+max_workers: {{num_nodes - 1}}
+upscaling_speed: {{num_nodes - 1}}
+idle_timeout_minutes: 60
+
+provider:
+  type: external
+  module: sky.provision.shadeform
+  region: "{{region}}"
+  disable_launch_config_check: true
+
+auth:
+  ssh_user: shadeform
+  ssh_private_key: {{ssh_private_key}}
+  ssh_key_id: {{ssh_key_id}}
+
+available_node_types:
+  ray_head_default:
+{%- if custom_resources %}
+    resources: {{custom_resources}}
+{%- else %}
+    resources: {}
+{%- endif %}
+    node_config:
+      InstanceType: {{instance_type}}
+      PublicKey: |-
+        skypilot:ssh_public_key_content
+
+head_node_type: ray_head_default
+
+# Format: `REMOTE_PATH : LOCAL_PATH`
+file_mounts: {
+  "{{sky_ray_yaml_remote_path}}": "{{sky_ray_yaml_local_path}}",
+  "{{sky_remote_path}}/{{sky_wheel_hash}}": "{{sky_local_path}}",
+{%- for remote_path, local_path in credentials.items() %}
+  "{{remote_path}}": "{{local_path}}",
+{%- endfor %}
+}
+
+rsync_exclude: []
+
+initialization_commands: []
+
+# List of shell commands to run to set up nodes.
+# NOTE: these are very performance-sensitive. Each new item opens/closes an SSH
+# connection, which is expensive. Try your best to co-locate commands into fewer
+# items!
+#
+# Increment the following for catching performance bugs easier:
+#   current num items (num SSH connections): 1
+setup_commands:
+  # Create ~/.ssh/config file in case the file does not exist in the image.
+  # Line 'rm ..': there is another installation of pip.
+  # Line 'sudo bash ..': set the ulimit as suggested by ray docs for performance. https://docs.ray.io/en/latest/cluster/vms/user-guides/large-cluster-best-practices.html#system-configuration
+  # Line 'sudo grep ..': set the number of threads per process to unlimited to avoid ray job submit stucking issue when the number of running ray jobs increase.
+  # Line 'mkdir -p ..': disable host key check
+  # Line 'python3 -c ..': patch the buggy ray files and enable `-o allow_other` option for `goofys`
+  - {%- for initial_setup_command in initial_setup_commands %}
+    {{ initial_setup_command }}
+    {%- endfor %}
+    mkdir -p ~/.ssh; touch ~/.ssh/config; which patch > /dev/null || sudo apt install -y patch;
+    {{ conda_installation_commands }}
+    {{ ray_skypilot_installation_commands }}
+    sudo bash -c 'rm -rf /etc/security/limits.d; echo "* soft nofile 1048576" >> /etc/security/limits.conf; echo "* hard nofile 1048576" >> /etc/security/limits.conf';
+    sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
+    (grep -Pzo -q "Host \*\n  StrictHostKeyChecking no" ~/.ssh/config) || printf "Host *\n  StrictHostKeyChecking no\n" >> ~/.ssh/config;
+    {{ ssh_max_sessions_config }}
+
+# Command to start ray clusters are now placed in `sky.provision.instance_setup`.
+# We do not need to list it here anymore.

sky/utils/context_utils.py

@@ -1,5 +1,6 @@
 """Utilities for SkyPilot context."""
 import asyncio
+import concurrent.futures
 import contextvars
 import functools
 import io
@@ -191,14 +192,17 @@ def to_thread(func: Callable[P, T], /, *args: P.args,
 
     This is same as asyncio.to_thread added in python 3.9
     """
+    return to_thread_with_executor(None, func, *args, **kwargs)
+
+
+def to_thread_with_executor(executor: Optional[concurrent.futures.Executor],
+                            func: Callable[P, T], /, *args: P.args,
+                            **kwargs: P.kwargs) -> 'asyncio.Future[T]':
+    """Asynchronously run function *func* in a separate thread with
+    a custom executor."""
+
     loop = asyncio.get_running_loop()
-    # This is critical to pass the current coroutine context to the new thread
     pyctx = contextvars.copy_context()
-    func_call: Callable[..., T] = functools.partial(
-        # partial deletes arguments type and thus can't figure out the return
-        # type of pyctx.run
-        pyctx.run,  # type: ignore
-        func,
-        *args,
-        **kwargs)
-    return loop.run_in_executor(None, func_call)
+    func_call: Callable[..., T] = functools.partial(pyctx.run, func, *args,
+                                                    **kwargs)
+    return loop.run_in_executor(executor, func_call)

{skypilot_nightly-1.0.0.dev20251011.dist-info → skypilot_nightly-1.0.0.dev20251013.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: skypilot-nightly
-Version: 1.0.0.dev20251011
+Version: 1.0.0.dev20251013
 Summary: SkyPilot: Run AI on Any Infra — Unified, Faster, Cheaper.
 Author: SkyPilot Team
 License: Apache 2.0
@@ -154,52 +154,53 @@ Requires-Dist: grpcio>=1.63.0; extra == "server"
 Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "server"
 Requires-Dist: aiosqlite; extra == "server"
 Requires-Dist: greenlet; extra == "server"
+Provides-Extra: shadeform
 Provides-Extra: all
-Requires-Dist: google-cloud-storage; extra == "all"
-Requires-Dist: ray[default]>=2.6.1; extra == "all"
-Requires-Dist: nebius>=0.2.47; extra == "all"
-Requires-Dist: awscli>=1.27.10; extra == "all"
-Requires-Dist: azure-mgmt-network>=27.0.0; extra == "all"
-Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "all"
-Requires-Dist: azure-core>=1.24.0; extra == "all"
-Requires-Dist: pyopenssl<24.3.0,>=23.2.0; extra == "all"
-Requires-Dist: azure-storage-blob>=12.23.1; extra == "all"
-Requires-Dist: websockets; extra == "all"
-Requires-Dist: passlib; extra == "all"
-Requires-Dist: pyvmomi==8.0.1.0.2; extra == "all"
-Requires-Dist: ibm-cloud-sdk-core; extra == "all"
-Requires-Dist: ibm-cos-sdk; extra == "all"
-Requires-Dist: azure-identity>=1.19.0; extra == "all"
-Requires-Dist: kubernetes!=32.0.0,>=20.0.0; extra == "all"
-Requires-Dist: pyjwt; extra == "all"
-Requires-Dist: pydo>=0.3.0; extra == "all"
-Requires-Dist: aiohttp; extra == "all"
+Requires-Dist: python-dateutil; extra == "all"
 Requires-Dist: ibm-platform-services>=0.48.0; extra == "all"
-Requires-Dist: colorama<0.4.5; extra == "all"
-Requires-Dist: azure-mgmt-compute>=33.0.0; extra == "all"
+Requires-Dist: tomli; python_version < "3.11" and extra == "all"
 Requires-Dist: anyio; extra == "all"
-Requires-Dist: oci; extra == "all"
-Requires-Dist: greenlet; extra == "all"
-Requires-Dist: ecsapi>=0.2.0; extra == "all"
-Requires-Dist: msrestazure; extra == "all"
-Requires-Dist: boto3>=1.26.1; extra == "all"
-Requires-Dist: msgraph-sdk; extra == "all"
+Requires-Dist: azure-storage-blob>=12.23.1; extra == "all"
+Requires-Dist: aiosqlite; extra == "all"
 Requires-Dist: casbin; extra == "all"
-Requires-Dist: ibm-vpc; extra == "all"
-Requires-Dist: python-dateutil; extra == "all"
-Requires-Dist: botocore>=1.29.10; extra == "all"
-Requires-Dist: vastai-sdk>=0.1.12; extra == "all"
-Requires-Dist: grpcio>=1.63.0; extra == "all"
+Requires-Dist: pyjwt; extra == "all"
 Requires-Dist: azure-cli>=2.65.0; extra == "all"
-Requires-Dist: tomli; python_version < "3.11" and extra == "all"
-Requires-Dist: aiosqlite; extra == "all"
-Requires-Dist: runpod>=1.6.1; extra == "all"
+Requires-Dist: pyopenssl<24.3.0,>=23.2.0; extra == "all"
+Requires-Dist: passlib; extra == "all"
+Requires-Dist: azure-core>=1.31.0; extra == "all"
+Requires-Dist: azure-mgmt-network>=27.0.0; extra == "all"
+Requires-Dist: vastai-sdk>=0.1.12; extra == "all"
+Requires-Dist: ibm-cloud-sdk-core; extra == "all"
 Requires-Dist: cudo-compute>=0.1.10; extra == "all"
+Requires-Dist: pyvmomi==8.0.1.0.2; extra == "all"
+Requires-Dist: grpcio>=1.63.0; extra == "all"
+Requires-Dist: boto3>=1.26.1; extra == "all"
+Requires-Dist: nebius>=0.2.47; extra == "all"
+Requires-Dist: greenlet; extra == "all"
 Requires-Dist: azure-common; extra == "all"
-Requires-Dist: sqlalchemy_adapter; extra == "all"
-Requires-Dist: docker; extra == "all"
+Requires-Dist: botocore>=1.29.10; extra == "all"
+Requires-Dist: azure-mgmt-compute>=33.0.0; extra == "all"
+Requires-Dist: ibm-vpc; extra == "all"
+Requires-Dist: oci; extra == "all"
+Requires-Dist: kubernetes!=32.0.0,>=20.0.0; extra == "all"
+Requires-Dist: azure-identity>=1.19.0; extra == "all"
+Requires-Dist: ibm-cos-sdk; extra == "all"
 Requires-Dist: google-api-python-client>=2.69.0; extra == "all"
-Requires-Dist: azure-core>=1.31.0; extra == "all"
+Requires-Dist: colorama<0.4.5; extra == "all"
+Requires-Dist: awscli>=1.27.10; extra == "all"
+Requires-Dist: docker; extra == "all"
+Requires-Dist: aiohttp; extra == "all"
+Requires-Dist: msrestazure; extra == "all"
+Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "all"
+Requires-Dist: ecsapi>=0.2.0; extra == "all"
+Requires-Dist: google-cloud-storage; extra == "all"
+Requires-Dist: azure-core>=1.24.0; extra == "all"
+Requires-Dist: sqlalchemy_adapter; extra == "all"
+Requires-Dist: runpod>=1.6.1; extra == "all"
+Requires-Dist: msgraph-sdk; extra == "all"
+Requires-Dist: ray[default]>=2.6.1; extra == "all"
+Requires-Dist: pydo>=0.3.0; extra == "all"
+Requires-Dist: websockets; extra == "all"
 Dynamic: author
 Dynamic: classifier
 Dynamic: description
@@ -249,10 +250,11 @@ Dynamic: summary
 ----
 
 :fire: *News* :fire:
+- [Oct 2025] Run large-scale **LLM training with TorchTitan** on any AI infra: [**example**](./llm/torchtitan/)
+- [Sep 2025] Scaling AI infrastructure at Abridge - **10x faster development** with SkyPilot: [**blog**](https://blog.skypilot.co/abridge/)
+- [Sep 2025] Network and Storage Benchmarks for LLM training on the cloud: [**blog**](https://maknee.github.io/blog/2025/Network-And-Storage-Training-Skypilot/)
 - [Aug 2025] Serve and finetune **OpenAI GPT-OSS models** (gpt-oss-120b, gpt-oss-20b) with one command on any infra: [**serve**](./llm/gpt-oss/) + [**LoRA and full finetuning**](./llm/gpt-oss-finetuning/)
-- [Jul 2025] Run large-scale **LLM training with TorchTitan** on any cloud: [**example**](./llm/torchtitan/)
 - [Jul 2025] Run distributed **RL training for LLMs** with Verl (PPO, GRPO) on any cloud: [**example**](./llm/verl/)
-- [Jul 2025] 🎉 SkyPilot v0.10.0 released! [**blog post**](https://blog.skypilot.co/announcing-skypilot-0.10.0/), [**release notes**](https://github.com/skypilot-org/skypilot/releases/tag/v0.10.0)
 - [Jul 2025] Finetune **Llama4** on any distributed cluster/cloud: [**example**](./llm/llama-4-finetuning/)
 - [Jul 2025] Two-part blog series, `The Evolution of AI Job Orchestration`: (1) [Running AI jobs on GPU Neoclouds](https://blog.skypilot.co/ai-job-orchestration-pt1-gpu-neoclouds/), (2) [The AI-Native Control Plane & Orchestration that Finally Works for ML](https://blog.skypilot.co/ai-job-orchestration-pt2-ai-control-plane/)
 - [Apr 2025] Spin up **Qwen3** on your cluster/cloud: [**example**](./llm/qwen/)