Enum ReasoningCapability 

pub enum ReasoningCapability {
    Unsupported,
    Binary,
    Levels(Vec<ReasoningLevel>),
    Budget {
        min: usize,
        max: usize,
    },
}

Describes the reasoning controls a model exposes.

Adapters declare this via ModelCapabilities::supports_reasoning. The user-facing ReasoningLevel is mapped onto whatever shape the model actually accepts via nearest_effort().

Variants

Unsupported

Model has no reasoning controls. Any ReasoningLevel is silently ignored by the adapter.

Binary

Model has on/off reasoning only (most Ollama-hosted models like deepseek-r1, qwen3, kimi-k2-thinking — think: true/false). ReasoningLevel::None → off; anything else → on.

Levels(Vec<ReasoningLevel>)

Model exposes a discrete enum of supported levels (gpt-oss, OpenAI o-series + GPT-5, Anthropic adaptive thinking, Groq, Fireworks, vLLM-on-gpt-oss). nearest_effort() maps the requested level onto this set.

Budget

Model expects a numeric token budget (legacy Anthropic budget_tokens, legacy Gemini 2.5 thinking_budget). Adapters translate ReasoningLevel onto a value in [min, max].

Fields

min: usize

max: usize

Trait Implementations

impl Clone for ReasoningCapability

fn clone(&self) -> ReasoningCapability

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ReasoningCapability

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for ReasoningCapability

impl PartialEq for ReasoningCapability

fn eq(&self, other: &ReasoningCapability) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for ReasoningCapability