frogml-core 0.0.1__py3-none-any.whl → 0.0.3__py3-none-any.whl

frogml_core/automations/build_and_deploy_action.py

@@ -0,0 +1,840 @@
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Dict, List
+
+from google.protobuf.wrappers_pb2 import BoolValue
+
+from frogml_proto.qwak.automation.v1.action_pb2 import Action as ActionProto
+from frogml_proto.qwak.automation.v1.action_pb2 import (
+    AdvancedDeploymentOptions,
+    BuildAndDeployAction,
+    BuildMetricCondition,
+    BuildSpec,
+)
+from frogml_proto.qwak.automation.v1.action_pb2 import (
+    CpuResources as CpuResourcesProto,
+)
+from frogml_proto.qwak.automation.v1.action_pb2 import (
+    DeployedBuildMetricCondition as DeployedBuildMetricConditionProto,
+)
+from frogml_proto.qwak.automation.v1.action_pb2 import (
+    DeploymentCondition as DeploymentConditionProto,
+)
+from frogml_proto.qwak.automation.v1.action_pb2 import (
+    DeploymentSize,
+    DeploymentSpec,
+    GitModelSource,
+)
+from frogml_proto.qwak.automation.v1.action_pb2 import GpuResources as GpuResourceProto
+from frogml_proto.qwak.automation.v1.action_pb2 import GpuType
+from frogml_proto.qwak.automation.v1.action_pb2 import (
+    PurchaseOption as PurchaseOptionProto,
+)
+from frogml_proto.qwak.automation.v1.action_pb2 import Resources as ResourceProto
+from frogml_proto.qwak.automation.v1.auto_scaling_pb2 import (
+    AGGREGATION_TYPE_AVERAGE,
+    AGGREGATION_TYPE_MAX,
+    AGGREGATION_TYPE_MIN,
+    AGGREGATION_TYPE_P50,
+    AGGREGATION_TYPE_P90,
+    AGGREGATION_TYPE_P95,
+    AGGREGATION_TYPE_P99,
+    AGGREGATION_TYPE_SUM,
+    METRIC_TYPE_CPU,
+    METRIC_TYPE_LATENCY,
+    METRIC_TYPE_MEMORY,
+)
+from frogml_proto.qwak.automation.v1.auto_scaling_pb2 import (
+    AutoScalingConfig as AutoScalingConfigProto,
+)
+from frogml_proto.qwak.automation.v1.auto_scaling_pb2 import (
+    AutoScalingPrometheusTrigger as AutoScalingPrometheusTriggerProto,
+)
+from frogml_proto.qwak.automation.v1.auto_scaling_pb2 import QuerySpec
+from frogml_proto.qwak.automation.v1.auto_scaling_pb2 import (
+    ScaleTrigger as ScaleTriggerProto,
+)
+from frogml_proto.qwak.automation.v1.auto_scaling_pb2 import Triggers as TriggersProto
+from frogml_proto.qwak.user_application.common.v0.resources_pb2 import (
+    ClientPodComputeResources,
+)
+from frogml_proto.qwak.user_application.common.v0.resources_pb2 import (
+    CpuResources as CommonCpuResourcesProto,
+)
+from frogml_proto.qwak.user_application.common.v0.resources_pb2 import (
+    GpuResources as CommonGpuResourcesProto,
+)
+from frogml_proto.qwak.user_application.common.v0.resources_pb2 import (
+    GpuType as CommonGpuTypeProto,
+)
+from frogml_proto.qwak.user_application.common.v0.resources_pb2 import (
+    PodComputeResourceTemplateSpec,
+)
+from frogml_core.automations.common import (
+    Action,
+    ThresholdDirection,
+    get_memory_amount,
+    map_common_memory_units,
+    map_common_memory_units_proto,
+    map_memory_units,
+    map_memory_units_proto,
+    map_proto_threshold_to_direction,
+    map_threshold_direction_to_proto,
+)
+
+
+@dataclass
+class DeploymentCondition(ABC):
+    @abstractmethod
+    def to_proto(self):
+        # abstract method
+        pass
+
+    @staticmethod
+    @abstractmethod
+    def from_proto(message: DeploymentConditionProto):
+        # abstract method
+        pass
+
+
+@dataclass
+class AutoScaleTrigger(ABC):
+    @abstractmethod
+    def to_proto(self):
+        # abstract method
+        pass
+
+    @staticmethod
+    @abstractmethod
+    def from_proto(message: ScaleTriggerProto):
+        # abstract method
+        pass
+
+
+@dataclass
+class DeployedBuildMetric(DeploymentCondition):
+    metric_name: str = field(default="")
+    direction: ThresholdDirection = field(default=ThresholdDirection.ABOVE)
+    variation: str = field(default="default")
+
+    def to_proto(self):
+        return DeploymentConditionProto(
+            deployed_build_metric=DeployedBuildMetricConditionProto(
+                metric_name=self.metric_name,
+                threshold_direction=map_threshold_direction_to_proto(self.direction),
+                variation=self.variation,
+            )
+        )
+
+    @staticmethod
+    def from_proto(message: DeploymentConditionProto):
+        return DeployedBuildMetric(
+            metric_name=message.deployed_build_metric.metric_name,
+            direction=map_proto_threshold_to_direction(
+                message.deployed_build_metric.threshold_direction
+            ),
+            variation=message.deployed_build_metric.variation,
+        )
+
+    def __str__(self):
+        return f"Metric Name: \tDirection: {self.direction.name}"
+
+
+@dataclass
+class BuildMetric(DeploymentCondition):
+    metric_name: str = field(default="")
+    direction: ThresholdDirection = field(default=ThresholdDirection.ABOVE)
+    threshold: str = field(default="")
+
+    def to_proto(self):
+        return DeploymentConditionProto(
+            build_metric=BuildMetricCondition(
+                metric_name=self.metric_name,
+                threshold=self.threshold,
+                threshold_direction=map_threshold_direction_to_proto(self.direction),
+            )
+        )
+
+    @staticmethod
+    def from_proto(message: DeploymentConditionProto):
+        return BuildMetric(
+            metric_name=message.build_metric.metric_name,
+            threshold=message.build_metric.threshold,
+            direction=map_proto_threshold_to_direction(
+                message.build_metric.threshold_direction
+            ),
+        )
+
+    def __str__(self):
+        return f"Metric Name: {self.metric_name}\tThreshold: {self.threshold}\tDirection: {self.direction.name}"
+
+
+@dataclass
+class Resources(ABC):
+    @abstractmethod
+    def to_proto(self):
+        # abstract method
+        pass
+
+    @staticmethod
+    @abstractmethod
+    def from_proto(message: ResourceProto):
+        # abstract method
+        pass
+
+
+@dataclass
+class CpuResources(Resources):
+    cpu_fraction: float = field(default=2)
+    memory: str = field(default="2Gi")
+
+    def to_proto(self):
+        return ResourceProto(
+            cpu_resources=CpuResourcesProto(
+                cpu=self.cpu_fraction,
+                memory_units=map_memory_units(memory=self.memory),
+                memory_amount=get_memory_amount(self.memory),
+            )
+        )
+
+    def to_common_proto(self):
+        return CommonCpuResourcesProto(
+            cpu=self.cpu_fraction,
+            memory_units=map_common_memory_units(memory=self.memory),
+            memory_amount=get_memory_amount(self.memory),
+        )
+
+    @staticmethod
+    def from_proto(message: ResourceProto):
+        return CpuResources(
+            cpu_fraction=message.cpu_resources.cpu,
+            memory=str(message.cpu_resources.memory_amount)
+            + map_memory_units_proto(message.cpu_resources.memory_units),
+        )
+
+    @staticmethod
+    def from_common_cpu_proto(message: CommonCpuResourcesProto):
+        return CpuResources(
+            cpu_fraction=message.cpu,
+            memory=str(message.memory_amount)
+            + map_common_memory_units_proto(message.memory_units),
+        )
+
+    def __str__(self):
+        return f"CPU: {self.cpu_fraction}, Memory: {self.memory}"
+
+
+@dataclass
+class GpuResources(Resources):
+    gpu_type: str = field(default=None)
+    gpu_amount: int = field(default=None)
+
+    def to_proto(self):
+        return ResourceProto(
+            gpu_resources=GpuResourceProto(
+                gpu_type=self.gpu_type,
+                gpu_amount=self.gpu_amount,
+            )
+        )
+
+    def to_gpu_proto(self):
+        return GpuResourceProto(
+            gpu_type=self.gpu_type,
+            gpu_amount=self.gpu_amount,
+        )
+
+    def to_common_gpu_proto(self):
+        return CommonGpuResourcesProto(
+            gpu_amount=self.gpu_amount, gpu_type=self.gpu_type
+        )
+
+    @staticmethod
+    def from_proto(message: ResourceProto):
+        return GpuResources(
+            gpu_type=GpuType.Name(message.gpu_resources.gpu_type),
+            gpu_amount=message.gpu_resources.gpu_amount,
+        )
+
+    @staticmethod
+    def from_gpu_proto(message: GpuResourceProto):
+        return GpuResources(
+            gpu_type=GpuType.Name(message.gpu_type),
+            gpu_amount=message.gpu_amount,
+        )
+
+    @staticmethod
+    def from_common_gpu_proto(message: CommonGpuResourcesProto):
+        return GpuResources(
+            gpu_type=CommonGpuTypeProto.Name(message.gpu_type),
+            gpu_amount=message.gpu_amount,
+        )
+
+    def __str__(self):
+        return f"GPU Type: {self.gpu_type}, GPU Amount: {self.gpu_amount}"
+
+
+@dataclass
+class ClientResources(Resources):
+    instance: str = field(default="")
+    gpu_resources: GpuResources = field(default=None)
+    cpu_resources: CpuResources = field(default=None)
+
+    def to_proto(self):
+        if self.instance:
+            client_pod_compute_resources = ClientPodComputeResources(
+                template_spec=PodComputeResourceTemplateSpec(template_id=self.instance)
+            )
+        elif self.gpu_resources:
+            client_pod_compute_resources = ClientPodComputeResources(
+                gpu_resources=self.gpu_resources.to_common_gpu_proto()
+            )
+        elif self.cpu_resources:
+            client_pod_compute_resources = ClientPodComputeResources(
+                cpu_resources=self.cpu_resources.to_common_proto()
+            )
+        else:
+            client_pod_compute_resources = ClientPodComputeResources()
+
+        return ResourceProto(client_pod_compute_resources=client_pod_compute_resources)
+
+    @staticmethod
+    def from_proto(message: ResourceProto):
+        resources_type = message.client_pod_compute_resources.WhichOneof("resources")
+        if resources_type == "template_spec":
+            return ClientResources(
+                instance=message.client_pod_compute_resources.template_spec.template_id
+            )
+        elif resources_type == "gpu_resources":
+            return ClientResources(
+                gpu_resources=GpuResources.from_common_gpu_proto(
+                    message.client_pod_compute_resources.gpu_resources
+                )
+            )
+        else:
+            return ClientResources(
+                cpu_resources=CpuResources.from_common_cpu_proto(
+                    message.client_pod_compute_resources.cpu_resources
+                )
+            )
+
+
+@dataclass
+class BuildSpecifications:
+    parameters: Dict[str, str] = field(default_factory=dict)
+    git_uri: str = field(default=None)
+    tags: List[str] = field(default_factory=list)
+    git_access_token_secret: str = field(default=None)
+    git_branch: str = field(default="main")
+    main_dir: str = field(default="main")
+    base_image: str = field(default="")
+    assumed_iam_role: str = field(default="")
+    resources: Resources = field(default_factory=CpuResources)
+    dependency_file_path: str = field(default=None)
+    env_vars: List[str] = field(default_factory=list)
+    git_ssh_token_secret: str = field(default=None)
+    gpu_compatible: bool = field(default=False)
+    push_image: bool = field(default=True)
+    purchase_option: str = field(default="spot")
+    provision_instance_timeout: int = field(default=120)
+    service_account_key_secret_name: str = field(default="")
+
+    def to_proto(self):
+        return BuildSpec(
+            parameters=self.parameters,
+            git_model_source=GitModelSource(
+                git_uri=self.git_uri,
+                git_credentials_secret_name=self.git_access_token_secret,
+                git_branch=self.git_branch,
+                git_ssh_secret_name=self.git_ssh_token_secret,
+            ),
+            main_dir=self.main_dir,
+            tags=self.tags,
+            resource=self.resources.to_proto(),
+            base_image=self.base_image,
+            assumed_iam_role=self.assumed_iam_role,
+            dependency_file_path=self.dependency_file_path,
+            env_vars=self.env_vars,
+            gpu_compatible=self.gpu_compatible,
+            push_image=BoolValue(value=self.push_image),
+            purchase_option=self.__purchase_option_to_proto(),
+            provision_instance_timeout=self.provision_instance_timeout,
+            service_account_key_secret_name=self.service_account_key_secret_name,
+        )
+
+    def __purchase_option_to_proto(self):
+        purchase_option = self.purchase_option.lower()
+        if purchase_option == "spot":
+            return PurchaseOptionProto.SPOT
+        elif purchase_option == "ondemand":
+            return PurchaseOptionProto.ON_DEMAND
+        else:
+            raise ValueError(
+                f"Invalid purchase option {self.purchase_option}. Purchase options can be either spot or ondemand"
+            )
+
+    @staticmethod
+    def from_proto(build_spec: BuildSpec):
+        resources = map_resources_name_to_class(
+            build_spec.resource.WhichOneof("resource")
+        )
+        push_image = True
+        if build_spec.HasField("push_image"):
+            push_image = build_spec.push_image.value
+
+        return BuildSpecifications(
+            parameters=build_spec.parameters,
+            git_uri=build_spec.git_model_source.git_uri,
+            git_access_token_secret=build_spec.git_model_source.git_credentials_secret_name,
+            git_branch=build_spec.git_model_source.git_branch,
+            tags=build_spec.tags,
+            main_dir=build_spec.main_dir,
+            base_image=build_spec.base_image,
+            resources=resources.from_proto(build_spec.resource) if resources else None,
+            dependency_file_path=build_spec.dependency_file_path,
+            assumed_iam_role=build_spec.assumed_iam_role,
+            env_vars=build_spec.env_vars,
+            git_ssh_token_secret=build_spec.git_model_source.git_ssh_secret_name,
+            gpu_compatible=build_spec.gpu_compatible,
+            push_image=push_image,
+            purchase_option=BuildSpecifications.__purchase_option_from_proto(
+                build_spec.purchase_option
+            ),
+            service_account_key_secret_name=build_spec.service_account_key_secret_name,
+            provision_instance_timeout=build_spec.provision_instance_timeout,
+        )
+
+    @staticmethod
+    def __purchase_option_from_proto(purchase_option: PurchaseOptionProto):
+        if purchase_option == PurchaseOptionProto.SPOT:
+            return "spot"
+        elif purchase_option == PurchaseOptionProto.ON_DEMAND:
+            return "ondemand"
+        else:
+            raise ValueError(
+                f"Invalid purchase option {purchase_option}. Purchase options can be either spot or ondemand"
+            )
+
+    def __str__(self):
+        result = f"Git Uri:\t{self.git_uri}\n"
+        result += f"Git Branch:\t{self.git_branch}\n"
+        result += (
+            f"Git Access Token Secret:\t{self.git_access_token_secret}\n"
+            if self.git_access_token_secret
+            else ""
+        )
+        result += (
+            f"Git Ssh Secret:\t{self.git_ssh_token_secret}\n"
+            if self.git_ssh_token_secret
+            else ""
+        )
+        result += f"Main Dir:\t{self.main_dir}\n" if self.main_dir != "main" else ""
+        result += f"Gpu Compatible:\t{self.gpu_compatible}\n"
+        result += f"Purchase Option:\t{self.purchase_option}\n"
+        result += f"Provision Instance Timeout:\t{self.provision_instance_timeout}\n"
+        result += f"Parameters:\t{self.parameters}\n" if self.parameters else ""
+        result += f"Tags:\t{self.tags}\n" if self.tags else ""
+        result += (
+            f"IAM Role:\t{self.assumed_iam_role}\n" if self.assumed_iam_role else ""
+        )
+        result += (
+            f"Service Account Key Secret Name:\t{self.service_account_key_secret_name}\n"
+            if self.service_account_key_secret_name
+            else ""
+        )
+        result += f"Base Image:\t{self.base_image}\n" if self.base_image else ""
+        result += f"Resources:\n{self.resources}\n" if self.resources else ""
+        result += f"Environment Variables:\n{self.env_vars}\n" if self.env_vars else ""
+        return result
+
+
+class MetricType(Enum):
+    cpu = METRIC_TYPE_CPU
+    latency = METRIC_TYPE_LATENCY
+    memory = METRIC_TYPE_MEMORY
+
+
+class AggregationType(Enum):
+    avg = AGGREGATION_TYPE_AVERAGE
+    max = AGGREGATION_TYPE_MAX
+    min = AGGREGATION_TYPE_MIN
+    sum = AGGREGATION_TYPE_SUM
+    p50 = AGGREGATION_TYPE_P50
+    p90 = AGGREGATION_TYPE_P90
+    p95 = AGGREGATION_TYPE_P95
+    p99 = AGGREGATION_TYPE_P99
+
+
+@dataclass
+class AutoScalingConfig:
+    min_replica_count: int = field(default=None)
+    max_replica_count: int = field(default=None)
+    polling_interval: int = field(default=None)
+    cool_down_period: int = field(default=None)
+    triggers: List[AutoScaleTrigger] = field(default_factory=list)
+
+    def to_proto(self):
+        return AutoScalingConfigProto(
+            min_replica_count=self.min_replica_count,
+            max_replica_count=self.max_replica_count,
+            polling_interval=self.polling_interval,
+            cool_down_period=self.cool_down_period,
+            triggers=TriggersProto(
+                triggers=[trigger.to_proto() for trigger in self.triggers]
+            ),
+        )
+
+    @staticmethod
+    def from_proto(message: AutoScalingConfigProto):
+        triggers = [
+            map_autoscaling_trigger_name_to_class(
+                trigger.WhichOneof("trigger_type")
+            ).from_proto(trigger)
+            for trigger in message.triggers.triggers
+        ]
+        return AutoScalingConfig(
+            min_replica_count=message.min_replica_count,
+            max_replica_count=message.max_replica_count,
+            polling_interval=message.polling_interval,
+            cool_down_period=message.cool_down_period,
+            triggers=triggers,
+        )
+
+    def __str__(self):
+        result = f"min replica count: {self.min_replica_count} \t max replica count: {self.max_replica_count} \t polling interval: {self.polling_interval} \t cool down period: {self.cool_down_period} \n"
+        result += "Scale triggers: "
+        for trigger in self.triggers:
+            result += f" {trigger}"
+        return result
+
+
+@dataclass
+class DeploymentSpecifications:
+    number_of_http_server_workers: int = field(default=2)
+    http_request_timeout_ms: int = field(default=5000)
+    max_batch_size: int = field(default=1)
+    daemon_mode: bool = field(default=False)
+    variation_name: str = field(default="default")
+    custom_iam_role_arn: str = field(default="")
+    number_of_pods: int = field(default=2)
+    cpu_fraction: float = field(default=2)
+    memory: str = field(default="2Gi")
+    gpu_resources: GpuResources = field(default=None)
+    auto_scale_config: AutoScalingConfig = field(default_factory=AutoScalingConfig)
+    environments: List[str] = field(default_factory=list)
+    env_vars: List[str] = field(default_factory=list)
+    deployment_timeout: int = field(default=0)
+    instance: str = field(default="")
+    service_account_key_secret_name: str = field(default="")
+
+    def to_proto(self):
+        deployment_size = self.__get_deployment_size_proto()
+        return DeploymentSpec(
+            selected_variation_name=self.variation_name,
+            deployment_size=deployment_size,
+            advanced_options=AdvancedDeploymentOptions(
+                number_of_http_server_workers=self.number_of_http_server_workers,
+                http_request_timeout_ms=self.http_request_timeout_ms,
+                max_batch_size=self.max_batch_size,
+                daemon_mode=self.daemon_mode,
+                custom_iam_role_arn=self.custom_iam_role_arn,
+                auto_scaling_config=self.auto_scale_config.to_proto(),
+                deployment_process_timeout_limit=self.deployment_timeout,
+                service_account_key_secret_name=self.service_account_key_secret_name,
+            ),
+            environments=self.environments,
+            env_vars=self.env_vars,
+        )
+
+    @staticmethod
+    def from_proto(message: DeploymentSpec):
+        if (
+            (
+                message.deployment_size.client_pod_compute_resources.cpu_resources
+                and message.deployment_size.client_pod_compute_resources.cpu_resources.cpu
+                > 0
+            )
+            or (
+                message.deployment_size.client_pod_compute_resources.gpu_resources
+                and message.deployment_size.client_pod_compute_resources.gpu_resources.gpu_amount
+                > 0
+            )
+            or (
+                message.deployment_size.client_pod_compute_resources.template_spec
+                and message.deployment_size.client_pod_compute_resources.template_spec.template_id
+            )
+        ):
+            return DeploymentSpecifications.__from_proto_common_resources(message)
+        else:
+            return DeploymentSpecifications.__from_proto_deprecated_resources(message)
+
+    def __str__(self):
+        result = self.__get_deployment_resources_output()
+        result += (
+            f"Number of HTTP Workers: {self.number_of_http_server_workers}\t"
+            f"HTTP Request Timeout: {self.http_request_timeout_ms}\t"
+            f"Daemon Mode: {self.daemon_mode}\t"
+            f"Max Batch Size: {self.max_batch_size}\n"
+        )
+        result += f"Environment Variables: {self.env_vars}\n"
+        result += (
+            f"Custom IAM Role: {self.custom_iam_role_arn}"
+            if self.custom_iam_role_arn
+            else ""
+        )
+        result += (
+            f"Custom Service Account Key Secret Name: {self.service_account_key_secret_name}\n"
+            if self.service_account_key_secret_name
+            else ""
+        )
+        result += f"Auto scaling config: {self.auto_scale_config}"
+        return result
+
+    def __get_deployment_resources_output(self) -> str:
+        result = f"Number of Pods: {self.number_of_pods}\t"
+        if self.instance:
+            result += f"Instance: {self.instance}\n"
+        elif self.gpu_resources and self.gpu_resources.gpu_amount > 0:
+            result += (
+                f"GPU: {self.gpu_resources.gpu_amount}\t"
+                f"Type: {self.gpu_resources.gpu_type}\n"
+            )
+        else:
+            result += f"CPU: {self.cpu_fraction}\tMemory: {self.memory}\n"
+
+        return result
+
+    def __get_deployment_size_proto(self) -> DeploymentSize:
+        if self.instance:
+            client_pod_compute_resources = ClientPodComputeResources(
+                template_spec=PodComputeResourceTemplateSpec(template_id=self.instance)
+            )
+        elif self.gpu_resources and self.gpu_resources.gpu_amount > 0:
+            client_pod_compute_resources = ClientPodComputeResources(
+                gpu_resources=CommonGpuResourcesProto(
+                    gpu_amount=self.gpu_resources.gpu_amount,
+                    gpu_type=self.gpu_resources.gpu_type,
+                )
+            )
+        else:
+            client_pod_compute_resources = ClientPodComputeResources(
+                cpu_resources=CommonCpuResourcesProto(
+                    cpu=self.cpu_fraction,
+                    memory_amount=get_memory_amount(self.memory),
+                    memory_units=map_common_memory_units(self.memory),
+                )
+            )
+        return DeploymentSize(
+            number_of_pods=self.number_of_pods,
+            client_pod_compute_resources=client_pod_compute_resources,
+            cpu=self.cpu_fraction,
+            memory_units=map_memory_units(memory=self.memory),
+            memory_amount=get_memory_amount(self.memory),
+            gpu_resources=(
+                self.gpu_resources.to_gpu_proto() if self.gpu_resources else None
+            ),
+        )
+
+    @staticmethod
+    def __from_proto_common_resources(message):
+        return DeploymentSpecifications(
+            variation_name=message.selected_variation_name,
+            number_of_http_server_workers=message.advanced_options.number_of_http_server_workers,
+            http_request_timeout_ms=message.advanced_options.http_request_timeout_ms,
+            max_batch_size=message.advanced_options.max_batch_size,
+            daemon_mode=message.advanced_options.daemon_mode,
+            custom_iam_role_arn=message.advanced_options.custom_iam_role_arn,
+            number_of_pods=message.deployment_size.number_of_pods,
+            cpu_fraction=message.deployment_size.client_pod_compute_resources.cpu_resources.cpu,
+            memory=str(
+                message.deployment_size.client_pod_compute_resources.cpu_resources.memory_amount
+            )
+            + map_common_memory_units_proto(
+                message.deployment_size.client_pod_compute_resources.cpu_resources.memory_units
+            ),
+            gpu_resources=(
+                GpuResources.from_common_gpu_proto(
+                    message.deployment_size.client_pod_compute_resources.gpu_resources
+                )
+                if (
+                    message.deployment_size.client_pod_compute_resources.gpu_resources
+                    and message.deployment_size.client_pod_compute_resources.gpu_resources.gpu_amount
+                    > 0
+                )
+                else None
+            ),
+            instance=(
+                message.deployment_size.client_pod_compute_resources.template_spec.template_id
+                if message.deployment_size.client_pod_compute_resources.template_spec.template_id
+                else ""
+            ),
+            auto_scale_config=AutoScalingConfig.from_proto(
+                message.advanced_options.auto_scaling_config
+            ),
+            environments=message.environments,
+            env_vars=message.env_vars,
+            deployment_timeout=message.advanced_options.deployment_process_timeout_limit,
+            service_account_key_secret_name=message.advanced_options.service_account_key_secret_name,
+        )
+
+    @staticmethod
+    def __from_proto_deprecated_resources(message):
+        return DeploymentSpecifications(
+            variation_name=message.selected_variation_name,
+            number_of_http_server_workers=message.advanced_options.number_of_http_server_workers,
+            http_request_timeout_ms=message.advanced_options.http_request_timeout_ms,
+            max_batch_size=message.advanced_options.max_batch_size,
+            daemon_mode=message.advanced_options.daemon_mode,
+            custom_iam_role_arn=message.advanced_options.custom_iam_role_arn,
+            number_of_pods=message.deployment_size.number_of_pods,
+            cpu_fraction=message.deployment_size.cpu,
+            memory=str(message.deployment_size.memory_amount)
+            + map_memory_units_proto(message.deployment_size.memory_units),
+            gpu_resources=(
+                GpuResources.from_gpu_proto(message.deployment_size.gpu_resources)
+                if message.deployment_size.gpu_resources
+                else None
+            ),
+            auto_scale_config=AutoScalingConfig.from_proto(
+                message.advanced_options.auto_scaling_config
+            ),
+            environments=message.environments,
+            env_vars=message.env_vars,
+            deployment_timeout=message.advanced_options.deployment_process_timeout_limit,
+            service_account_key_secret_name=message.advanced_options.service_account_key_secret_name,
+        )
+
+
+@dataclass
+class FrogmlBuildDeploy(Action):
+    build_spec: BuildSpecifications
+    deployment_condition: DeploymentCondition
+    deployment_spec: DeploymentSpecifications
+
+    def to_proto(self):
+        return ActionProto(
+            build_deploy=BuildAndDeployAction(
+                build_spec=self.build_spec.to_proto(),
+                deployment_spec=self.deployment_spec.to_proto(),
+                deployment_condition=self.deployment_condition.to_proto(),
+            )
+        )
+
+    @staticmethod
+    def from_proto(message: ActionProto):
+        deployment_condition = map_deployment_condition_name_to_class(
+            message.build_deploy.deployment_condition.WhichOneof("condition")
+        )
+        return FrogmlBuildDeploy(
+            build_spec=BuildSpecifications.from_proto(message.build_deploy.build_spec),
+            deployment_spec=DeploymentSpecifications.from_proto(
+                message.build_deploy.deployment_spec
+            ),
+            deployment_condition=(
+                deployment_condition.from_proto(
+                    message.build_deploy.deployment_condition
+                )
+                if deployment_condition
+                else None
+            ),
+        )
+
+    def __str__(self):
+        return f"Build Specifications:\n{self.build_spec}\nDeployment Specification:\n{self.deployment_spec}\nDeployment Condition:\n{self.deployment_condition}"
+
+
+@dataclass
+class AutoScaleQuerySpec:
+    metric_type: str = field(default=None)
+    aggregation_type: str = field(default=None)
+    time_period: int = field(default=None)
+
+    def to_proto(self):
+        return QuerySpec(
+            time_period=self.time_period,
+            metric_type=MetricType[self.metric_type.lower()].value,
+            aggregation_type=AggregationType[self.aggregation_type.lower()].value,
+        )
+
+    @staticmethod
+    def from_proto(message: QuerySpec):
+        return AutoScaleQuerySpec(
+            metric_type=map_auto_scaling_metric_type_proto_to_name(message.metric_type),
+            aggregation_type=map_aggregation_type_proto_to_name(
+                message.aggregation_type
+            ),
+            time_period=message.time_period,
+        )
+
+    def __str__(self):
+        return f"metric type: {self.metric_type}\taggregation type: {self.aggregation_type}\ttime period: {self.time_period}"
+
+
+@dataclass
+class AutoScalingPrometheusTrigger(AutoScaleTrigger):
+    query_spec: AutoScaleQuerySpec = field(default=QuerySpec)
+    threshold: int = field(default=None)
+
+    def to_proto(self):
+        return ScaleTriggerProto(
+            prometheus_trigger=AutoScalingPrometheusTriggerProto(
+                query_spec=self.query_spec.to_proto(), threshold=self.threshold
+            )
+        )
+
+    @staticmethod
+    def from_proto(message: ScaleTriggerProto):
+        return AutoScalingPrometheusTrigger(
+            threshold=message.prometheus_trigger.threshold,
+            query_spec=AutoScaleQuerySpec.from_proto(
+                message.prometheus_trigger.query_spec
+            ),
+        )
+
+    def __str__(self):
+        return f"threshold: {self.threshold} \t {self.query_spec}"
+
+
+def map_resources_name_to_class(resource_name: str):
+    mapping = {
+        "cpu_resources": CpuResources,
+        "gpu_resources": GpuResources,
+        "client_pod_compute_resources": ClientResources,
+    }
+    return mapping.get(resource_name)
+
+
+def map_deployment_condition_name_to_class(deployment_condition_name: str):
+    mapping = {
+        "build_metric": BuildMetric,
+        "deployed_build_metric": DeployedBuildMetric,
+    }
+    return mapping.get(deployment_condition_name)
+
+
+def map_auto_scaling_metric_type_proto_to_name(metric_type):
+    mapping = {
+        METRIC_TYPE_CPU: "cpu",
+        METRIC_TYPE_LATENCY: "latency",
+        METRIC_TYPE_MEMORY: "memory",
+    }
+    return mapping.get(metric_type)
+
+
+def map_aggregation_type_proto_to_name(aggregation_type):
+    mapping = {
+        AGGREGATION_TYPE_AVERAGE: "avg",
+        AGGREGATION_TYPE_MAX: "max",
+        AGGREGATION_TYPE_MIN: "min",
+        AGGREGATION_TYPE_SUM: "sum",
+        AGGREGATION_TYPE_P50: "p50",
+        AGGREGATION_TYPE_P90: "p90",
+        AGGREGATION_TYPE_P95: "p95",
+        AGGREGATION_TYPE_P99: "p99",
+    }
+    return mapping.get(aggregation_type)
+
+
+def map_autoscaling_trigger_name_to_class(auto_scaling_trigger_name: str):
+    mapping = {"prometheus_trigger": AutoScalingPrometheusTrigger}
+    return mapping.get(auto_scaling_trigger_name)