dstack 0.19.9__py3-none-any.whl → 0.19.11__py3-none-any.whl

dstack/_internal/server/testing/common.py

@@ -262,6 +262,7 @@ async def create_run(
     run_spec: Optional[RunSpec] = None,
     run_id: Optional[UUID] = None,
     deleted: bool = False,
+    priority: int = 0,
 ) -> RunModel:
     if run_spec is None:
         run_spec = get_run_spec(
@@ -282,6 +283,7 @@ async def create_run(
         run_spec=run_spec.json(),
         last_processed_at=submitted_at,
         jobs=[],
+        priority=priority,
     )
     session.add(run)
     await session.commit()

dstack/_internal/server/utils/routers.py

@@ -93,13 +93,10 @@ def get_server_client_error_details(error: ServerClientError) -> List[Dict]:
     return details
 
 
-def request_size_exceeded(request: Request, limit: int) -> bool:
+def get_request_size(request: Request) -> int:
     if "content-length" not in request.headers:
-        return True
-    content_length = int(request.headers["content-length"])
-    if content_length > limit:
-        return True
-    return False
+        return 0
+    return int(request.headers["content-length"])
 
 
 def check_client_server_compatibility(

dstack/_internal/settings.py

@@ -12,6 +12,10 @@ DSTACK_RELEASE = os.getenv("DSTACK_RELEASE") is not None or version.__is_release
 DSTACK_USE_LATEST_FROM_BRANCH = os.getenv("DSTACK_USE_LATEST_FROM_BRANCH") is not None
 
 
+DSTACK_BASE_IMAGE = os.getenv("DSTACK_BASE_IMAGE", "dstackai/base")
+DSTACK_BASE_IMAGE_VERSION = os.getenv("DSTACK_BASE_IMAGE_VERSION", version.base_image)
+
+
 class FeatureFlags:
     """
     dstack feature flags. Feature flags are temporary and can be used when developing

dstack/api/_public/runs.py

@@ -691,29 +691,32 @@ class RunCollection:
         logger.warning("The exec_plan() method is deprecated in favor of apply_plan().")
         return self.apply_plan(run_plan=run_plan, repo=repo, reserve_ports=reserve_ports)
 
-    def list(self, all: bool = False) -> List[Run]:
+    def list(self, all: bool = False, limit: Optional[int] = None) -> List[Run]:
         """
         List runs.
 
         Args:
             all: Show all runs (active and finished) if `True`.
+            limit: Limit the number of runs to return. Must be less than 100.
 
         Returns:
             List of runs.
         """
         # Return only one page of latest runs (<=100). Returning all the pages may be costly.
         # TODO: Consider introducing `since` filter with a reasonable default.
-        only_active = not all
+        only_active = not all and limit is None
         runs = self._api_client.runs.list(
             project_name=self._project,
             repo_id=None,
             only_active=only_active,
+            limit=limit or 100,
         )
         if only_active and len(runs) == 0:
             runs = self._api_client.runs.list(
                 project_name=self._project,
                 repo_id=None,
-            )[:1]
+                limit=1,
+            )
         return [self._model_to_run(run) for run in runs]
 
     def get(self, run_name: str) -> Optional[Run]:

dstack/api/server/_runs.py

@@ -186,6 +186,8 @@ def _get_run_spec_excludes(run_spec: RunSpec) -> Optional[Dict]:
         configuration_excludes["rate_limits"] = True
     if configuration.shell is None:
         configuration_excludes["shell"] = True
+    if configuration.priority is None:
+        configuration_excludes["priority"] = True
 
     if configuration_excludes:
         spec_excludes["configuration"] = configuration_excludes

dstack/plugins/builtin/rest_plugin/__init__.py

@@ -0,0 +1,18 @@
+# ruff: noqa: F401
+from dstack.plugins.builtin.rest_plugin._models import (
+    FleetSpecRequest,
+    FleetSpecResponse,
+    GatewaySpecRequest,
+    GatewaySpecResponse,
+    RunSpecRequest,
+    RunSpecResponse,
+    SpecApplyRequest,
+    SpecApplyResponse,
+    VolumeSpecRequest,
+    VolumeSpecResponse,
+)
+from dstack.plugins.builtin.rest_plugin._plugin import (
+    PLUGIN_SERVICE_URI_ENV_VAR_NAME,
+    CustomApplyPolicy,
+    RESTPlugin,
+)

dstack/plugins/builtin/rest_plugin/_models.py

@@ -0,0 +1,48 @@
+from typing import Generic, Optional, TypeVar
+
+from pydantic import BaseModel, Field
+from typing_extensions import Annotated
+
+from dstack._internal.core.models.fleets import FleetSpec
+from dstack._internal.core.models.gateways import GatewaySpec
+from dstack._internal.core.models.runs import RunSpec
+from dstack._internal.core.models.volumes import VolumeSpec
+
+SpecType = TypeVar("SpecType", RunSpec, FleetSpec, VolumeSpec, GatewaySpec)
+
+
+class SpecApplyRequest(BaseModel, Generic[SpecType]):
+    user: Annotated[str, Field(description="The name of the user making the apply request")]
+    project: Annotated[str, Field(description="The name of the project the request is for")]
+    spec: Annotated[SpecType, Field(description="The spec to be applied")]
+
+    # Override dict() to remove __orig_class__ attribute and avoid "TypeError: Object of type _GenericAlias is not JSON serializable"
+    # error. This issue doesn't happen though when running the code in pytest, only when running the server.
+    def dict(self, *args, **kwargs):
+        d = super().dict(*args, **kwargs)
+        d.pop("__orig_class__", None)
+        return d
+
+
+RunSpecRequest = SpecApplyRequest[RunSpec]
+FleetSpecRequest = SpecApplyRequest[FleetSpec]
+VolumeSpecRequest = SpecApplyRequest[VolumeSpec]
+GatewaySpecRequest = SpecApplyRequest[GatewaySpec]
+
+
+class SpecApplyResponse(BaseModel, Generic[SpecType]):
+    spec: Annotated[
+        SpecType,
+        Field(
+            description="The spec to apply, original spec if error otherwise original or mutated by plugin service if approved"
+        ),
+    ]
+    error: Annotated[
+        Optional[str], Field(description="Error message if request is rejected", min_length=1)
+    ] = None
+
+
+RunSpecResponse = SpecApplyResponse[RunSpec]
+FleetSpecResponse = SpecApplyResponse[FleetSpec]
+VolumeSpecResponse = SpecApplyResponse[VolumeSpec]
+GatewaySpecResponse = SpecApplyResponse[GatewaySpec]

dstack/plugins/builtin/rest_plugin/_plugin.py

@@ -0,0 +1,127 @@
+import json
+import os
+from typing import Type
+
+import requests
+from pydantic import ValidationError
+
+from dstack._internal.core.errors import ServerClientError
+from dstack._internal.core.models.fleets import FleetSpec
+from dstack._internal.core.models.gateways import GatewaySpec
+from dstack._internal.core.models.volumes import VolumeSpec
+from dstack.plugins import ApplyPolicy, ApplySpec, Plugin, RunSpec, get_plugin_logger
+from dstack.plugins.builtin.rest_plugin import (
+    FleetSpecRequest,
+    FleetSpecResponse,
+    GatewaySpecRequest,
+    GatewaySpecResponse,
+    RunSpecRequest,
+    RunSpecResponse,
+    SpecApplyRequest,
+    SpecApplyResponse,
+    VolumeSpecRequest,
+    VolumeSpecResponse,
+)
+
+logger = get_plugin_logger(__name__)
+
+PLUGIN_SERVICE_URI_ENV_VAR_NAME = "DSTACK_PLUGIN_SERVICE_URI"
+PLUGIN_REQUEST_TIMEOUT_SEC = 8
+
+
+class CustomApplyPolicy(ApplyPolicy):
+    def __init__(self):
+        self._plugin_service_uri = os.getenv(PLUGIN_SERVICE_URI_ENV_VAR_NAME)
+        logger.info(f"Found plugin service at {self._plugin_service_uri}")
+        if not self._plugin_service_uri:
+            logger.error(
+                f"Cannot create policy because {PLUGIN_SERVICE_URI_ENV_VAR_NAME} is not set"
+            )
+            raise ServerClientError(f"{PLUGIN_SERVICE_URI_ENV_VAR_NAME} is not set")
+
+    def _check_request_rejected(self, response: SpecApplyResponse):
+        if response.error is not None:
+            logger.error(f"Plugin service rejected apply request: {response.error}")
+            raise ServerClientError(f"Apply request rejected: {response.error}")
+
+    def _call_plugin_service(self, spec_request: SpecApplyRequest, endpoint: str) -> ApplySpec:
+        response = None
+        try:
+            response = requests.post(
+                f"{self._plugin_service_uri}{endpoint}",
+                json=spec_request.dict(),
+                headers={"accept": "application/json", "Content-Type": "application/json"},
+                timeout=PLUGIN_REQUEST_TIMEOUT_SEC,
+            )
+            response.raise_for_status()
+            spec_json = json.loads(response.text)
+            return spec_json
+        except requests.exceptions.ConnectionError as e:
+            logger.error(
+                f"Could not connect to plugin service at {self._plugin_service_uri}: %s", e
+            )
+            raise ServerClientError(
+                f"Could not connect to plugin service at {self._plugin_service_uri}"
+            )
+        except requests.RequestException as e:
+            logger.error("Request to the plugin service failed: %s", e)
+            raise ServerClientError("Request to the plugin service failed")
+
+    def _on_apply(
+        self,
+        request_cls: Type[SpecApplyRequest],
+        response_cls: Type[SpecApplyResponse],
+        endpoint: str,
+        user: str,
+        project: str,
+        spec: ApplySpec,
+    ) -> ApplySpec:
+        try:
+            spec_request = request_cls(user=user, project=project, spec=spec)
+            spec_json = self._call_plugin_service(spec_request, endpoint)
+            response = response_cls(**spec_json)
+            self._check_request_rejected(response)
+            return response.spec
+        except ValidationError:
+            logger.error(f"Plugin service returned invalid response:\n{spec_json}")
+            raise ServerClientError("Plugin service returned an invalid response")
+
+    def on_run_apply(self, user: str, project: str, spec: RunSpec) -> RunSpec:
+        return self._on_apply(
+            RunSpecRequest, RunSpecResponse, "/apply_policies/on_run_apply", user, project, spec
+        )
+
+    def on_fleet_apply(self, user: str, project: str, spec: FleetSpec) -> FleetSpec:
+        return self._on_apply(
+            FleetSpecRequest,
+            FleetSpecResponse,
+            "/apply_policies/on_fleet_apply",
+            user,
+            project,
+            spec,
+        )
+
+    def on_volume_apply(self, user: str, project: str, spec: VolumeSpec) -> VolumeSpec:
+        return self._on_apply(
+            VolumeSpecRequest,
+            VolumeSpecResponse,
+            "/apply_policies/on_volume_apply",
+            user,
+            project,
+            spec,
+        )
+
+    def on_gateway_apply(self, user: str, project: str, spec: GatewaySpec) -> GatewaySpec:
+        return self._on_apply(
+            GatewaySpecRequest,
+            GatewaySpecResponse,
+            "/apply_policies/on_gateway_apply",
+            user,
+            project,
+            spec,
+        )
+
+
+class RESTPlugin(Plugin):
+    def get_apply_policies(self) -> list[ApplyPolicy]:
+        return [CustomApplyPolicy()]

dstack/version.py

@@ -1,3 +1,3 @@
-__version__ = "0.19.9"
+__version__ = "0.19.11"
 __is_release__ = True
-base_image = "0.7"
+base_image = "0.8"

{dstack-0.19.9.dist-info → dstack-0.19.11.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dstack
-Version: 0.19.9
+Version: 0.19.11
 Summary: dstack is an open-source orchestration engine for running AI workloads on any cloud or on-premises.
 Project-URL: Homepage, https://dstack.ai
 Project-URL: Source, https://github.com/dstackai/dstack
@@ -21,7 +21,7 @@ Requires-Dist: cryptography
 Requires-Dist: cursor
 Requires-Dist: filelock
 Requires-Dist: gitpython
-Requires-Dist: gpuhunt==0.1.5
+Requires-Dist: gpuhunt==0.1.6
 Requires-Dist: jsonschema
 Requires-Dist: packaging
 Requires-Dist: paramiko>=3.2.0
@@ -56,6 +56,7 @@ Requires-Dist: azure-mgmt-subscription>=3.1.1; extra == 'all'
 Requires-Dist: backports-entry-points-selectable; extra == 'all'
 Requires-Dist: boto3>=1.38.13; extra == 'all'
 Requires-Dist: botocore; extra == 'all'
+Requires-Dist: cryptography>=44.0.3; extra == 'all'
 Requires-Dist: datacrunch; extra == 'all'
 Requires-Dist: docker>=6.0.0; extra == 'all'
 Requires-Dist: fastapi; extra == 'all'
@@ -71,8 +72,9 @@ Requires-Dist: httpx; extra == 'all'
 Requires-Dist: jinja2; extra == 'all'
 Requires-Dist: kubernetes; extra == 'all'
 Requires-Dist: nebius<0.3,>=0.2.19; (python_version >= '3.10') and extra == 'all'
-Requires-Dist: oci; extra == 'all'
+Requires-Dist: oci>=2.150.0; extra == 'all'
 Requires-Dist: prometheus-client; extra == 'all'
+Requires-Dist: pyopenssl>=23.2.0; extra == 'all'
 Requires-Dist: python-dxf==12.1.0; extra == 'all'
 Requires-Dist: python-json-logger>=3.1.0; extra == 'all'
 Requires-Dist: sentry-sdk[fastapi]; extra == 'all'
@@ -280,13 +282,15 @@ Requires-Dist: alembic>=1.10.2; extra == 'oci'
 Requires-Dist: apscheduler<4; extra == 'oci'
 Requires-Dist: asyncpg; extra == 'oci'
 Requires-Dist: backports-entry-points-selectable; extra == 'oci'
+Requires-Dist: cryptography>=44.0.3; extra == 'oci'
 Requires-Dist: docker>=6.0.0; extra == 'oci'
 Requires-Dist: fastapi; extra == 'oci'
 Requires-Dist: grpcio>=1.50; extra == 'oci'
 Requires-Dist: httpx; extra == 'oci'
 Requires-Dist: jinja2; extra == 'oci'
-Requires-Dist: oci; extra == 'oci'
+Requires-Dist: oci>=2.150.0; extra == 'oci'
 Requires-Dist: prometheus-client; extra == 'oci'
+Requires-Dist: pyopenssl>=23.2.0; extra == 'oci'
 Requires-Dist: python-dxf==12.1.0; extra == 'oci'
 Requires-Dist: python-json-logger>=3.1.0; extra == 'oci'
 Requires-Dist: sentry-sdk[fastapi]; extra == 'oci'
@@ -423,9 +427,9 @@ To point the CLI to the `dstack` server, configure it
 with the server address, user token, and project name:
 
 ```shell
-$ dstack config \
+$ dstack project add \
+    --name main \
     --url http://127.0.0.1:3000 \
-    --project main \
     --token bbae0f28-d3dd-4820-bf61-8f4bb40815da
     
 Configuration is updated at ~/.dstack/config.yml