Struct ModelSelectionStrategy 

pub struct ModelSelectionStrategy { /* private fields */ }

Model Selection Strategy

This strategy evaluates different LLM models based on:

• Cost: Token cost and total daily cost
• Performance: Latency (P50, P95, P99)
• Quality: User ratings and success rates

It uses Pareto optimization to find models that are optimal for at least one objective while not being dominated by other models in all objectives.

Implementations

impl ModelSelectionStrategy

pub fn new(config: ModelSelectionConfig) -> Self

Create a new model selection strategy

Trait Implementations

impl Debug for ModelSelectionStrategy

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

Formats the value using the given formatter.

impl OptimizationStrategy for ModelSelectionStrategy

fn name(&self) -> &str

Get the name of this strategy

fn priority(&self) -> u32

Get the priority of this strategy (higher = more important)

fn is_applicable(&self, input: &DecisionInput) -> bool

Check if this strategy is applicable to the given input

fn evaluate<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, input: &'life1 DecisionInput, criteria: &'life2 DecisionCriteria, ) -> Pin<Box<dyn Future<Output = DecisionResult<Vec<Decision>>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Evaluate and generate decisions

fn validate<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, decision: &'life1 Decision, criteria: &'life2 DecisionCriteria, ) -> Pin<Box<dyn Future<Output = DecisionResult<()>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Validate a decision before execution

fn learn<'life0, 'life1, 'life2, 'async_trait>( &'life0 mut self, decision: &'life1 Decision, outcome: &'life2 DecisionOutcome, ) -> Pin<Box<dyn Future<Output = DecisionResult<()>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Learn from a decision outcome

fn get_stats(&self) -> Value

Get strategy-specific statistics