Struct PromptOptimizer 

pub struct PromptOptimizer { /* private fields */ }

Available on crate feature eval only.

Iteratively improves an agent’s system instructions using an optimizer LLM and an evaluation set.

The optimizer runs a loop of evaluate → propose improvements → apply best → repeat, logging progress via tracing at each iteration.

Implementations

impl PromptOptimizer

pub fn new( optimizer_llm: Arc<dyn Llm>, evaluator: Evaluator, config: OptimizerConfig, ) -> PromptOptimizer

Create a new prompt optimizer.

Arguments

• optimizer_llm - The LLM used to propose instruction improvements (separate from the agent’s own LLM).
• evaluator - The evaluator used to score the agent against the eval set.
• config - Optimization configuration (max iterations, target threshold, output path).

pub async fn optimize( &self, agent: Arc<dyn Agent>, eval_set: &EvalSet, ) -> Result<OptimizationResult, EvalError>

Run the optimization loop.

Evaluates the agent, proposes improvements via the optimizer LLM, applies the best improvement, and repeats until the target threshold is met or max iterations are exhausted.

On completion, writes the best-performing instructions to the configured output file.