enterprise_ai.llm.reasoning

Source: sdks/python/src/enterprise_ai/llm/reasoning/__init__.py. Back to the Python SDK reference index.

note

This page is generated by tools/docs/gen_sdk_reference.py from the SDK sources and the OpenAPI text folders. Do not edit it by hand: change the source and regenerate.

Enumeration Effort

Normalized internal reasoning effort, ordered from none to max.

str, Enum

Values:

Member	Value	Description
NONE	'none'	No reasoning (none).
LOW	'low'	Low reasoning effort (low).
MEDIUM	'medium'	Medium reasoning effort (medium).
HIGH	'high'	High reasoning effort (high).
MAX	'max'	Highest reasoning effort (max).

Attributes:

Name	Type	Required	Description
rank	int	yes	Position of the effort in the order none (0) to max (4).

Effort.parse

parse(value: str | Effort | None) -> Effort

Parse a string (case insensitive). None means none.

Name	Type	Default	Required	Description
value	str | Effort | None		yes	Effort as a string, an Effort (returned as is) or None; an unknown value raises EffortNotSupported.

Returns: Effort. A Effort: Normalized internal reasoning effort, ordered from none to max.

Raises: EffortNotSupported.

Model EffortResolution

Result of :func:resolve_effort.

Fields:

Name	Type	Required	Description
requested	Effort	yes	Normalized internal reasoning effort, ordered from none to max.
effective	Effort	yes	Normalized internal reasoning effort, ordered from none to max.
native_parameter	str	no	Name of the provider reasoning parameter from spec.reasoning.native_parameter (default none).
native	dict[str, Any]	no	Provider parameters for the effective effort, such as {"reasoning_effort": "high"} or a thinking budget.
adjusted	bool	no	True when the requested effort was mapped to another level by the model policy.
reason	str | None	no	Why the action is taken. Written to the audit trail with the actor and the time.
reasoning_tokens	int	no	Explicit reasoning tokens, at least 0; overrides the reserve of reasoning_effort.

Functions

supported_efforts

supported_efforts(model: dict[str, Any]) -> list[Effort]

Efforts listed in reasoning_modes (none only when absent).

Name	Type	Default	Required	Description
model	dict[str, Any]		yes	Model Manifest or registry entry.

Returns: list[Effort]. A Effort: Normalized internal reasoning effort, ordered from none to max.

resolve_effort

resolve_effort(model: dict[str, Any], effort: str | Effort | None) -> EffortResolution

Check effort against a model manifest (or registry entry) and translate it.

Name	Type	Default	Required	Description
model	dict[str, Any]		yes	Model Manifest or registry entry whose reasoning_modes and reasoning policy apply.
effort	str | Effort | None		yes	Requested reasoning effort; None takes spec.reasoning.default_effort.

Returns: EffortResolution. A EffortResolution: Result of :func:resolve_effort.

Raises: EffortNotSupported.

reasoning_reserve

reasoning_reserve(model: dict[str, Any], effort: str | Effort | None) -> int

Tokens to reserve for reasoning when planning a context (0 for none).

Name	Type	Default	Required	Description
model	dict[str, Any]		yes	Model Manifest or registry entry.
effort	str | Effort | None		yes	Requested reasoning effort; None takes the model default effort.

Returns: int. The model budget_tokens for the effective effort, otherwise 1024 (low), 4096 (medium), 16384 (high) or 32768 (max).