SAMGuardrail

SAM model of the Guardrail kind, generated from the Smarter platform’s JSON Schema.

Do not edit: run scripts/generate_sam_models.py. See smarter.models.sam.

class smarter.models.sam.guardrail.SAMDependency(*, kind: str, name: str, **extra_data: Any)

Bases: SAMModel

A resource that depends on the resource that a manifest describes.

kind: str

Kind[String]: The SAM kind of the resource that depends on this resource, e.g. LLMClient.

Read only.

name: str

Name[String]: The name of the resource that depends on this resource.

Read only.

model_config = {'extra': 'allow', 'populate_by_name': True, 'protected_namespaces': (), 'regex_engine': 'python-re', 'use_attribute_docstrings': True}

Configuration for the model, should be a dictionary conforming to [`ConfigDict`][pydantic.config.ConfigDict].

class smarter.models.sam.guardrail.SAMGuardrailMetadata(*, annotations: None | list[dict[str, str | int | float | bool | date | datetime | UUID | bytes | list | dict[str, Any]]] = None, description: str | None, name: str, tags: list[str] | None = None, version: str | None, **extra_data: Any)

Bases: SAMModel

Smarter API Guardrail Manifest - Metadata class.

annotations: None | list[dict[str, str | int | float | bool | date | datetime | UUID | bytes | list | dict[str, Any]]]

The manifest annotations.

Used for storing arbitrary metadata as

key-value pairs. Example: [{‘smarter.sh/test-manifest/project-name’: ‘Scooby dooby do’}]. The key should be a valid url-friendly string. The value accepts multi-line string values (YAML block scalars) and various scalar types including str, int, float, bool, datetime.date, datetime.datetime, decimal.Decimal, uuid.UUID, bytes, list, dict.

description: str | None

The description for this resource.

Be brief. Keep it under 255 characters.

name: str

The camelCase name of the manifest resource.

tags: list[str] | None

The tags of the manifest.

Used for generic resource categorization and search. Example: [‘tag1’, ‘tag2’]

version: str | None

The semantic version of the manifest.

Example: 0.1.0

model_config = {'extra': 'allow', 'populate_by_name': True, 'protected_namespaces': (), 'regex_engine': 'python-re', 'use_attribute_docstrings': True}

Configuration for the model, should be a dictionary conforming to [`ConfigDict`][pydantic.config.ConfigDict].

class smarter.models.sam.guardrail.Action(*values)

Bases: Enum

Guardrail.spec.config.action[str]: what the guardrail does when it triggers.

One of: [‘log’, ‘flag’, ‘redact’, ‘transform’, ‘block’, ‘escalate’].

class smarter.models.sam.guardrail.Category(*values)

Bases: Enum

Guardrail.spec.config.category[str]: the risk the guardrail addresses.

One of: [‘pii’, ‘secrets’, ‘prompt_injection’, ‘jailbreak’, ‘toxicity’, ‘self_harm’, ‘hallucination’, ‘off_topic’, ‘compliance’, ‘formatting’, ‘custom’].

class smarter.models.sam.guardrail.Mode(*values)

Bases: Enum

Guardrail.spec.config.mode[str]: enforce, to take the action, or monitor, to only record what the action would have been.

class smarter.models.sam.guardrail.Stage(*values)

Bases: Enum

Guardrail.spec.config.stage[str]: when the guardrail runs.

One of: [‘input’, ‘output’, ‘both’].

class smarter.models.sam.guardrail.Strategy(*values)

Bases: Enum

Guardrail.spec.config.strategy[str]: how the guardrail detects.

One of: [‘regex’, ‘keyword’, ‘detector’, ‘semantic’, ‘moderation’, ‘llm_judge’].

class smarter.models.sam.guardrail.SAMGuardrailSpecConfig(*, action: Action, caseSensitive: bool | None = False, categories: list[str] | None = None, category: Category | None = 'custom', detectors: list[str] | None = None, failClosed: bool | None = False, flags: list[str] | None = None, isActive: bool | None = True, judgePrompt: str | None = None, keywords: list[str] | None = None, message: str | None = None, mode: Mode | None = 'enforce', model: str | None = None, pattern: str | None = None, priority: Annotated[int, None, Interval(gt=None, ge=0, lt=None, le=None), None] | None = 100, provider: str | None = None, referenceTexts: list[str] | None = None, replacement: str | None = None, severity: Annotated[int, None, Interval(gt=None, ge=1, lt=None, le=5), None] | None = 3, stage: Stage, strategy: Strategy, threshold: Annotated[float, None, Interval(gt=None, ge=0.0, lt=None, le=1.0), None, None] | None = None, wholeWord: bool | None = True, **extra_data: Any)

Bases: SAMModel

Smarter API Guardrail Manifest Guardrail.spec.config.

action: Action

Guardrail.spec.config.action[str]: what the guardrail does when it triggers.

One of: [‘log’, ‘flag’, ‘redact’, ‘transform’, ‘block’, ‘escalate’].

caseSensitive: bool | None

Guardrail.spec.config.caseSensitive[bool]: whether the keyword strategy is case sensitive.

categories: list[str] | None

Guardrail.spec.config.categories[list]: the moderation strategy’s categories.

Empty means all. Any of: [‘harassment’, ‘harassment/threatening’, ‘hate’, ‘hate/threatening’, ‘illicit’, ‘illicit/violent’, ‘self-harm’, ‘self-harm/intent’, ‘self-harm/instructions’, ‘sexual’, ‘sexual/minors’, ‘violence’, ‘violence/graphic’].

category: Category | None

Guardrail.spec.config.category[str]: the risk the guardrail addresses.

One of: [‘pii’, ‘secrets’, ‘prompt_injection’, ‘jailbreak’, ‘toxicity’, ‘self_harm’, ‘hallucination’, ‘off_topic’, ‘compliance’, ‘formatting’, ‘custom’].

detectors: list[str] | None

Guardrail.spec.config.detectors[list]: the detector strategy’s detectors.

Any of: [‘credit_card’, ‘us_ssn’, ‘email’, ‘phone_number’, ‘ip_address’, ‘iban’, ‘aws_access_key’, ‘private_key’, ‘api_key’, ‘jwt’, ‘password_assignment’].

failClosed: bool | None

Guardrail.spec.config.failClosed[bool]: if true, block when the check itself fails, e.g. because the LLM provider is unavailable.

If false, the failure is recorded, and the prompt continues.

flags: list[str] | None

Guardrail.spec.config.flags[list]: the regex strategy’s flags.

Any of: [‘IGNORECASE’, ‘MULTILINE’, ‘DOTALL’].

isActive: bool | None

Guardrail.spec.config.isActive[bool]: whether the guardrail runs.

judgePrompt: str | None

Guardrail.spec.config.judgePrompt[str]: the llm_judge strategy’s prompt, with a {text} placeholder for the text to judge.

Any other braces must be doubled, e.g. {{ and }}.

keywords: list[str] | None

Guardrail.spec.config.keywords[list]: the keyword strategy’s words and phrases.

message: str | None

Guardrail.spec.config.message[str]: what block returns to the user, instead of the LLM’s reply.

mode: Mode | None

Guardrail.spec.config.mode[str]: enforce, to take the action, or monitor, to only record what the action would have been.

model: str | None

Guardrail.spec.config.model[str]: the model of the semantic, moderation and llm_judge strategies.

pattern: str | None

Guardrail.spec.config.pattern[str]: the regex strategy’s regular expression.

priority: conint(ge=0) | None

Guardrail.spec.config.priority[int]: the order in which guardrails run.

Lower runs first.

provider: str | None

Guardrail.spec.config.provider[str]: the name of the Smarter LLM Provider that the semantic, moderation and llm_judge strategies call.

Defaults to openai.

referenceTexts: list[str] | None

Guardrail.spec.config.referenceTexts[list]: the semantic strategy’s examples of what to detect.

model_config = {'extra': 'allow', 'populate_by_name': True, 'protected_namespaces': (), 'regex_engine': 'python-re', 'use_attribute_docstrings': True}

Configuration for the model, should be a dictionary conforming to [`ConfigDict`][pydantic.config.ConfigDict].

replacement: str | None

Guardrail.spec.config.replacement[str]: what redact and transform replace each match with.

Defaults to [REDACTED], which an empty replacement also means. {label} is replaced with what matched, e.g. credit_card. For the regex strategy, transform may refer to groups, e.g.
.

severity: conint(ge=1, le=5) | None

Guardrail.spec.config.severity[int]: from 1 (low) to 5 (critical), for reporting and review.

stage: Stage

Guardrail.spec.config.stage[str]: when the guardrail runs.

One of: [‘input’, ‘output’, ‘both’].

strategy: Strategy

Guardrail.spec.config.strategy[str]: how the guardrail detects.

One of: [‘regex’, ‘keyword’, ‘detector’, ‘semantic’, ‘moderation’, ‘llm_judge’].

threshold: confloat(ge=0.0, le=1.0) | None

Guardrail.spec.config.threshold[float]: the minimum confidence, from 0 to 1, at which the semantic, moderation and llm_judge strategies trigger.

Defaults to 0.8.

wholeWord: bool | None

Guardrail.spec.config.wholeWord[bool]: whether the keyword strategy matches whole words only.

class smarter.models.sam.guardrail.SAMGuardrailStatus(*, accountNumber: str, blocked: int | None = None, created: datetime, dependencies: list[SAMDependency] | None = None, errors: int | None = None, lastTriggered: datetime | None = None, modified: datetime, recordLocator: str, triggered: int | None = None, username: str, **extra_data: Any)

Bases: SAMModel

Smarter API Guardrail Manifest - Status class.

Read only. Besides ownership, it reports the guardrail’s events of the last 30 days. The LLMClients that use it are reported in dependencies.

model_config = {'extra': 'allow', 'populate_by_name': True, 'protected_namespaces': (), 'regex_engine': 'python-re', 'use_attribute_docstrings': True}

Configuration for the model, should be a dictionary conforming to [`ConfigDict`][pydantic.config.ConfigDict].

accountNumber: str

Guardrail.status.account_number: The account owner of this Guardrail.

Read only.

blocked: int | None

Guardrail.status.blocked: how many prompts this Guardrail blocked in the last 30 days.

Read only.

created: datetime

The date in which this resource was created.

Read only.

dependencies: list[SAMDependency] | None

Dependencies[List]: The resources that depend on this resource, and that must be deleted, or stop referring to it, before it can be deleted.

Lists only the resources that you can read. Read only.

errors: int | None

Guardrail.status.errors: how many times this Guardrail failed to run in the last 30 days.

Read only.

lastTriggered: datetime | None

Guardrail.status.lastTriggered: when this Guardrail last triggered.

Read only.

modified: datetime

The date in which this resource was most recently changed.

Read only.

recordLocator: str

RecordLocator[String]: An optional identifier used to locate the resource record associated with this manifest.

Read only.

triggered: int | None

Guardrail.status.triggered: how many times this Guardrail triggered in the last 30 days.

Read only.

username: str

Guardrail.status.username: The Smarter user who created this Guardrail.

Read only.

class smarter.models.sam.guardrail.SAMGuardrailSpec(*, config: SAMGuardrailSpecConfig, **extra_data: Any)

Bases: SAMModel

Smarter API Guardrail Manifest Guardrail.spec.

model_config = {'extra': 'allow', 'populate_by_name': True, 'protected_namespaces': (), 'regex_engine': 'python-re', 'use_attribute_docstrings': True}

Configuration for the model, should be a dictionary conforming to [`ConfigDict`][pydantic.config.ConfigDict].

config: SAMGuardrailSpecConfig

Guardrail.spec.config[object].

The configuration for the Guardrail.

class smarter.models.sam.guardrail.SAMGuardrail(*, apiVersion: str, kind: str, metadata: SAMGuardrailMetadata, spec: SAMGuardrailSpec, status: SAMGuardrailStatus | None = None, **extra_data: Any)

Bases: SAMModel

Smarter API Manifest - Guardrail.

model_config = {'extra': 'allow', 'populate_by_name': True, 'protected_namespaces': (), 'regex_engine': 'python-re', 'use_attribute_docstrings': True}

Configuration for the model, should be a dictionary conforming to [`ConfigDict`][pydantic.config.ConfigDict].

apiVersion: str

ApiVersion[String]: Required.

The API version of the AbstractSAMBase.

kind: str

Kind[String]: Required.

The kind of resource described by the manifest.

metadata: SAMGuardrailMetadata

Guardrail.metadata[obj]: Required, the Guardrail metadata.

spec: SAMGuardrailSpec

Guardrail.spec[obj]: Required, the Guardrail specification.

status: SAMGuardrailStatus | None

Guardrail.status[obj]: Optional, Read-only.

Stateful status information about the Guardrail.