Struct ModelRole 

pub struct ModelRole {
    pub model: ModelId,
    pub max_output_tokens: Option<u32>,
    pub max_input_tokens: Option<u32>,
    pub reasoning_effort: Option<ReasoningEffort>,
    pub pricing: Option<Pricing>,
}

One model role, as a declaration resolves it: which model, and the ceilings the reviewer put beside it.

A manifest declares max_tokens and reasoning_effort per role because the model is only half the decision — the role that reads hostile content is the one whose output most needs a ceiling somebody reviewed. Passing the id alone is how those two fields get parsed into the digest and then dropped, so the runtime’s seams carry the whole role and applied_to puts it on a call in one motion.

Fields

model: ModelId

max_output_tokens: Option<u32>

Cap on generated tokens, when the declaration set one.

max_input_tokens: Option<u32>

Cap on the tokens one call may send, when the declaration set one.

reasoning_effort: Option<ReasoningEffort>

Requested reasoning depth, when the declaration set one.

pricing: Option<Pricing>

What this role’s tokens cost, when the declaration stated it.

Implementations

impl ModelRole

pub fn new(model: ModelId) -> Self

A role with no declared ceilings — the driver defaults apply.

pub fn applied_to(&self, call: ModelCall) -> ModelCall

Put this role’s declared ceilings on a call.

The model itself is not applied here — a ModelCall is constructed with its model, so applying it afterwards would be a second place the choice is made.

Trait Implementations

impl Clone for ModelRole

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelRole

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

Formats the value using the given formatter.

impl Eq for ModelRole

impl PartialEq for ModelRole

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for ModelRole