yera 0.11.1

yera.models.reasoning_lookup

Reasoning capability lookup for LLM models by provider and model ID.

Provides pattern-based matching to determine reasoning support, default levels, and budgets for different models across providers like Anthropic, OpenAI, Mistral, and AWS Bedrock.

Symbols

def look_up_reasoning_spec — Look up the reasoning specification for a given provider/author/model.
class ReasoningSpec — Specification of reasoning capabilities for a model.

look_up_reasoning_spec

look_up_reasoning_spec(
    provider: str,
    author: str,
    model_id: str,
) → ReasoningSpec

Look up the reasoning specification for a given provider/author/model.

Parameters

provider
type: str

The LLM provider (e.g., "anthropic", "openai", "aws").

author
type: str

The model author or vendor (e.g., "anthropic", "amazon").

model_id
type: str

The exact model identifier to match against patterns.

Returns

type: ReasoningSpec

The matching reasoning spec, or a default none spec if no match found.

ReasoningSpec

Specification of reasoning capabilities for a model.

Fields

kind
type: Literal['effort', 'budget', 'none', 'always'] = 'none'
supported_levels
type: tuple[typing.Literal['off', 'minimal', 'low', 'medium', 'high', 'xhigh', 'max', 'auto'], ...] = ()
default_level
type: Optional[Literal['off', 'minimal', 'low', 'medium', 'high', 'xhigh', 'max', 'auto']] = None
default_budget
type: int | None = None
mandatory
type: bool = False

Methods


ReasoningSpec.__init__

__init__(
    kind: ReasoningKind = 'none',
    supported_levels: tuple[ReasoningLevel, ...] = (),
    default_level: ReasoningLevel | None = None,
    default_budget: int | None = None,
    mandatory: bool = False,
) → None