Django ORM<a class="headerlink" href="#module-smarter.apps.llmhost.models" title="Link to this heading">

Annotations. Key-value pairs for annotating this resource.

Type:: Type

api_format

Api format

Choices:

openai_compatible
huggingface
ollama_native
custom

Type:: Type

api_key_secret

ForeignKey to Secret

Api key secret. The Smarter Secret with the API key that clients send as a Bearer token. (related name: llmhosts)

Type:: Type

api_key_secret_id: Internal field, use api_key_secret instead.

compute

ForeignKey to LLMHostCompute

Compute. The kind of node, and node group, that the LLMHost runs on: spec.compute. (related name: llmhosts)

Type:: Type

compute_id: Internal field, use compute instead.

context_window

Context window

Type:: Type

cost_per_hour

DecimalField

Cost per hour. The cost per hour of one replica: its share of its compute’s node.

Type:: Type

created_at

Created at

Timestamp indicating when the model instance was created.

This field is automatically set to the current date and time when the instance is first created. It is indexed in the database for efficient querying.

Type:: Type

deployed_at

Deployed at. When the LLMHost was last launched.

Type:: Type

description

Description. A brief description of this resource. Be verbose, but not too verbose.

Type:: Type

embedding_dimensions

Embedding dimensions. Set only for embedding models.

Type:: Type

endpoint_url

URLField

Endpoint url. The base URL inside the cluster.

Type:: Type

property estimated_cost: Decimal | None

uptime times replicas times cost_per_hour.

Type:: The cost since the LLMHost was launched

events

Reverse ForeignKey from LLMHostEvent

All events of this llm host (related name of llmhost)

Type:: Type

get_api_format_display(*, field=<django.db.models.CharField: api_format>): Shows the label of the api_format. See get_FOO_display() for more information.

get_inference_engine_display(*, field=<django.db.models.CharField: inference_engine>): Shows the label of the inference_engine. See get_FOO_display() for more information.

get_model_source_display(*, field=<django.db.models.CharField: model_source>): Shows the label of the model_source. See get_FOO_display() for more information.

get_model_task_display(*, field=<django.db.models.CharField: model_task>): Shows the label of the model_task. See get_FOO_display() for more information.

get_quantization_display(*, field=<django.db.models.CharField: quantization>): Shows the label of the quantization. See get_FOO_display() for more information.

get_status_display(*, field=<django.db.models.CharField: status>): Shows the label of the status. See get_FOO_display() for more information.

gpu_count

Gpu count. GPUs per replica.

Type:: Type

gpu_type

Gpu type. e.g. ‘A10G’

Type:: Type

health_check_url

URLField

Health check url

Type:: Type

id

BigAutoField

Primary key: ID

Type:: Type

inference_engine

Inference engine

Choices:

vllm
tgi
sglang
llama_cpp
ollama
tei
custom

Type:: Type

is_active

Is active

Type:: Type

property is_deployed: bool: Whether the LLMHost’s Kubernetes resources exist, as of the last status check.

last_health_check_at

Last health check at

Type:: Type

last_health_ok

Last health ok

Type:: Type

license

License

Type:: Type

model_architecture

Model architecture. e.g. ‘llama’, ‘mistral’, ‘qwen3’

Type:: Type

model_repository

Model repository. e.g. ‘meta-llama/Llama-3.1-8B-Instruct’, ‘llama3.1:8b’ or ‘s3://bucket/prefix’

Type:: Type

model_revision

Model revision. Commit SHA, branch or tag pinned for reproducible deploys.

Type:: Type

model_source

Model source

Choices:

huggingface
ollama
s3
url
pvc

Type:: Type

model_task

Model task

Choices:

text-generation
embedding

Type:: Type

name

Name. Name in camelCase, e.g., ‘apiKey’, no special characters.

Type:: Type

parameter_count

BigIntegerField

Parameter count

Type:: Type

public_url

URLField

Public url. The base URL of the Ingress, if any.

Type:: Type

quantization

Quantization

Choices:

none
fp32
fp16
bf16
fp8
mxfp4
int8
int4
gguf
awq
gptq

Type:: Type

ready_replicas

Ready replicas

Type:: Type

record_event(event_type, message='', details=None)[source]

Record a lifecycle event, with the LLMHost’s current status.

Return type:: LLMHostEvent

replicas

Replicas

Type:: Type

served_model_name

Served model name. The model name that clients send in their requests.

Type:: Type

spec

Spec. The manifest’s spec, in camelCase. The source of truth for launching the LLMHost.

Type:: Type

status

Status

Choices:

provisioning
pending
downloading
deploying
active
degraded
inactive
error

Type:: Type

status_message

Status message

Type:: Type

supports_function_calling

Supports function calling

Type:: Type

supports_reasoning

Supports reasoning

Type:: Type

supports_streaming

Supports streaming

Type:: Type

supports_vision

Supports vision

Type:: Type

tagged_items

Reverse GenericRelation from LLMHost

All + of this tagged item (related name of tagged_items)

Type:: Type

tags = <taggit.managers._TaggableManager object>

updated_at

Updated at

Timestamp indicating when the model instance was last updated.

This field is automatically updated to the current date and time whenever the instance is saved. It is indexed in the database for efficient querying.

Type:: Type

property uptime: timedelta | None: The time since the LLMHost was launched, if it is deployed.

property uptime_hours: float | None

user_profile

ForeignKey to UserProfile

User profile (related name: llmhost)

Type:: Type

user_profile_id: Internal field, use user_profile instead.

version

Version. Semantic version in the format MAJOR.MINOR.PATCH, e.g., ‘1.0.0’.

Type:: Type

vram_required_gb

Vram required gb

Type:: Type

class smarter.apps.llmhost.models.LLMHostCompute(*args, **kwargs)[source]

Bases: MetaDataWithOwnershipModel

A kind of node, e.g. g6.2xlarge with one NVIDIA L4, and its node group.

The node’s capacity is what the instance type provides. What pods can request, its allocatable capacity, is less: the kubelet, the operating system and the cluster’s DaemonSets reserve some of it. See allocatable_cpu_millicores and allocatable_memory_mib.

Parameters:

id (Unknown) – Primary key: ID
created_at (Unknown) – Created at
updated_at (Unknown) – Updated at
name (Unknown) – Name. Name in camelCase, e.g., ‘apiKey’, no special characters.
description (Unknown) – Description. A brief description of this resource. Be verbose, but not too verbose.
version (Unknown) – Version. Semantic version in the format MAJOR.MINOR.PATCH, e.g., ‘1.0.0’.
annotations (Unknown) – Annotations. Key-value pairs for annotating this resource.
spec (Unknown) – Spec. The manifest’s spec, in camelCase. The source of truth for the node group.
instance_type (Unknown) – Instance type. e.g. ‘g6.2xlarge’.
cpu (Unknown) – Cpu. vCPUs.
memory_gb (Unknown) – Memory gb. GiB.
gpu_type (Unknown) – Gpu type. e.g. ‘L4’.
gpu_count (Unknown) – Gpu count
gpu_memory_gb (Unknown) – Gpu memory gb. Per GPU.
max_nodes (Unknown) – Max nodes. The most nodes that Smarter adds.
price_per_hour (Unknown) – Price per hour. Of one node, e.g. its on-demand price.
nodegroup_status (Unknown) – Nodegroup status
desired_nodes (Unknown) – Desired nodes
ready_nodes (Unknown) – Ready nodes
status_message (Unknown) – Status message
last_reconciled_at (Unknown) – Last reconciled at

Relationship fields:

Parameters:

user_profile (Unknown) – User profile (related name: llmhostcompute)
tags (Unknown) – Tags. Tags for categorizing and organizing this resource. (related name: llmhostcompute)
tagged_items (Unknown) – Tagged items (related name: +)

Reverse relationships:

Parameters:: llmhosts (Unknown) – All llmhosts of this LLMHost Compute (related name of compute)

exception DoesNotExist: Bases: ObjectDoesNotExist

exception MultipleObjectsReturned: Bases: MultipleObjectsReturned

exception NotUpdated: Bases: ObjectNotUpdated, DatabaseError

property allocatable_cpu_millicores: int

An estimate of the CPU that pods can request on one node.

The kubelet’s reservation, and the cluster’s DaemonSets, e.g. the CNI, kube-proxy and the NVIDIA device plugin: half a core plus 2%.

property allocatable_memory_mib: int

less 1.5 GiB plus 3%.

Type:: An estimate of the memory that pods can request on one node

property ami_type: str

annotations

Annotations. Key-value pairs for annotating this resource.

Type:: Type

property capacity_type: str

clone(new_name=None, new_version=None, user_profile=None)[source]

Clone the LLMHostCompute’s spec, but not its node group: the clone is a new kind of node, whose node group Smarter creates when an LLMHost first needs one of its nodes.

Return type:: LLMHostCompute

cpu

Cpu. vCPUs.

Type:: Type

created_at

Created at

Timestamp indicating when the model instance was created.

This field is automatically set to the current date and time when the instance is first created. It is indexed in the database for efficient querying.

Type:: Type

description

Description. A brief description of this resource. Be verbose, but not too verbose.

Type:: Type

desired_nodes

Desired nodes

Type:: Type

property disk_size_gb: int

gpu_count

Gpu count

Type:: Type

gpu_memory_gb

Gpu memory gb. Per GPU.

Type:: Type

property gpu_resource: str

gpu_type

Gpu type. e.g. ‘L4’.

Type:: Type

property has_gpu: bool

id

BigAutoField

Primary key: ID

Type:: Type

instance_type

Instance type. e.g. ‘g6.2xlarge’.

Type:: Type

last_reconciled_at

Last reconciled at

Type:: Type

llmhosts

Reverse ForeignKey from LLMHost

All llmhosts of this LLMHost Compute (related name of compute)

Type:: Type

max_nodes

Max nodes. The most nodes that Smarter adds.

Type:: Type

memory_gb

Memory gb. GiB.

Type:: Type

name

Name. Name in camelCase, e.g., ‘apiKey’, no special characters.

Type:: Type

property node_spec: dict[str, Any]: Spec.node, as stored.

property nodegroup_name: str: The node group’s name, unique to the environment, e.g. smarter-alpha-12-gpu-l4-1x.

property nodegroup_spec: dict[str, Any]: Spec.nodeGroup, as stored.

nodegroup_status

Nodegroup status

Type:: Type

price_per_hour

DecimalField

Price per hour. Of one node, e.g. its on-demand price.

Type:: Type

ready_nodes

Ready nodes

Type:: Type

rename(new_name)[source]

Rename the LLMHostCompute.

Refused while its node group exists, whose name includes the LLMHostCompute’s, or while LLMHosts use it, whose spec.compute names it.

Raises:: ValueError – if it cannot be renamed.
Return type:: LLMHostCompute

spec

Spec. The manifest’s spec, in camelCase. The source of truth for the node group.

Type:: Type

status_message

Status message

Type:: Type

property subnet_ids: list[str]

tagged_items

Reverse GenericRelation from LLMHostCompute

All + of this tagged item (related name of tagged_items)

Type:: Type

tags = <taggit.managers._TaggableManager object>

property taint: dict[str, str]: The node group’s taint, which LLMHost pods tolerate.

updated_at

Updated at

Timestamp indicating when the model instance was last updated.

This field is automatically updated to the current date and time whenever the instance is saved. It is indexed in the database for efficient querying.

Type:: Type

user_profile

ForeignKey to UserProfile

User profile (related name: llmhostcompute)

Type:: Type

user_profile_id: Internal field, use user_profile instead.

version

Version. Semantic version in the format MAJOR.MINOR.PATCH, e.g., ‘1.0.0’.

Type:: Type

class smarter.apps.llmhost.models.LLMHostEvent(*args, **kwargs)[source]

Bases: TimestampedModel

A lifecycle event of an LLMHost: launched, status changed, destroyed, or an error.

Events are the LLMHost’s audit trail, and the basis of uptime and cost reporting.

Parameters:

id (Unknown) – Primary key: ID
created_at (Unknown) – Created at
updated_at (Unknown) – Updated at
llmhost_name (Unknown) – Llmhost name. The LLMHost’s name, in case it is deleted.
event_type (Unknown) – Event type
status (Unknown) – Status
message (Unknown) – Message
details (Unknown) – Details

Relationship fields:

Parameters:: llmhost (Unknown) – Llmhost (related name: events)

exception DoesNotExist: Bases: ObjectDoesNotExist

exception MultipleObjectsReturned: Bases: MultipleObjectsReturned

exception NotUpdated: Bases: ObjectNotUpdated, DatabaseError

created_at

Created at

Timestamp indicating when the model instance was created.

This field is automatically set to the current date and time when the instance is first created. It is indexed in the database for efficient querying.

Type:: Type

details

Details

Type:: Type

event_type

Event type

Choices:

launched
status_changed
destroyed
nodes_scaled
error

Type:: Type

get_event_type_display(*, field=<django.db.models.CharField: event_type>): Shows the label of the event_type. See get_FOO_display() for more information.

get_status_display(*, field=<django.db.models.CharField: status>): Shows the label of the status. See get_FOO_display() for more information.

id

BigAutoField

Primary key: ID

Type:: Type

llmhost

ForeignKey to LLMHost

Llmhost (related name: events)

Type:: Type

llmhost_id: Internal field, use llmhost instead.

llmhost_name

Llmhost name. The LLMHost’s name, in case it is deleted.

Type:: Type

message

Message

Type:: Type

objects = <django.db.models.Manager object>

status

Status

Choices:

provisioning
pending
downloading
deploying
active
degraded
inactive
error

Type:: Type

updated_at