Module model_selector 

Intelligent model selection based on cost / quality / latency tradeoffs.

Key types

• ModelProfile — static description of a model’s characteristics.
• SelectionCriteria — caller-supplied hard constraints (max cost, min quality, …).
• SelectionStrategy — ranking algorithm applied to the candidate set.
• ModelSelector — registry + selection engine.
• ModelUsageTracker — records actual runtime usage to compare with predictions.

Example

use tokio_prompt_orchestrator::model_selector::{
    ModelProfile, ModelSelector, SelectionCriteria, SelectionStrategy,
};

let mut selector = ModelSelector::new(SelectionStrategy::CheapestFirst);
selector.register(ModelProfile {
    id: "fast-cheap".to_string(),
    cost_per_1k_tokens: 0.001,
    avg_latency_ms: 200.0,
    quality_score: 0.75,
    max_context_tokens: 8192,
    supports_streaming: true,
    supports_tools: false,
});
let criteria = SelectionCriteria::default();
let model = selector.select(&criteria).unwrap();
assert_eq!(model.id, "fast-cheap");

Structs

ModelProfile

Static description of a model’s cost, performance, and capability profile.

ModelSelector

Registry and selection engine for ModelProfiles.

ModelUsage

Accumulated runtime usage for a single model.

ModelUsageTracker

Thread-safe tracker for actual model usage.

SelectionCriteria

Hard constraints that a model must satisfy to be eligible for selection.

Enums

SelectionStrategy

Algorithm used to rank the eligible model candidates.