Struct GenerationOptions

Source

pub struct GenerationOptions { /* private fields */ }

Expand description

Tuning parameters for a single generation request.

Values are optional; None uses the model’s built-in default. Numeric settings are stored as Temperature and MaxTokens so validated generation semantics cannot be bypassed after construction.

Implementations§

Source §

impl GenerationOptions

Source

pub fn new() -> Self

Creates options using model defaults.

Source

pub fn temperature(self, temperature: Temperature) -> Self

Sets the generation temperature.

Range: 0.0 (fully deterministic) to 2.0 (very creative).

Source

pub fn with_temperature(self, temperature: Temperature) -> Self

Alias for GenerationOptions::temperature.

Source

pub fn try_temperature(self, temperature: f64) -> Result<Self, Error>

Parses and sets a generation temperature from a raw boundary value.

Prefer GenerationOptions::temperature when your code already has a Temperature. Use this at IO boundaries such as CLI, JSON, or UI input.

§Errors

Returns Error::InvalidTemperature when temperature is outside Apple Intelligence’s supported range.

Source

pub fn max_tokens(self, max_tokens: MaxTokens) -> Self

Sets the maximum number of response tokens.

The model’s session has a combined context window of 4 096 tokens (instructions + all prompts + all responses). Leaving this unset lets the model decide.

Source

pub fn with_max_tokens(self, max_tokens: MaxTokens) -> Self

Alias for GenerationOptions::max_tokens.

Source

pub fn try_max_tokens(self, max_tokens: usize) -> Result<Self, Error>

Parses and sets a maximum token count from a raw boundary value.

Prefer GenerationOptions::max_tokens when your code already has a MaxTokens. Use this at IO boundaries such as CLI, JSON, or UI input.

§Errors

Returns Error::InvalidMaxTokens when the value cannot be represented by the Swift bridge.

Source

pub fn temperature_value(&self) -> Option<Temperature>

Returns the configured typed temperature, if any.

Source

pub fn max_tokens_value(&self) -> Option<MaxTokens>

Returns the configured typed maximum response token count, if any.

Source

pub fn validate(&self) -> Result<(), Error>

Validates all configured option values.

Values constructed through this type are already validated. This method is kept so generic setup code can verify options before storing them.

§Examples

use aimx::{GenerationOptions, MaxTokens, Temperature};

let options = GenerationOptions::new()
    .temperature(Temperature::new(0.4)?)
    .max_tokens(MaxTokens::new(128)?);
options.validate()?;