Module prompt_optimizer 

Prompt Optimizer

Prompt variant testing and automatic improvement tracking using bandit algorithms: UCB1, Epsilon-Greedy, Thompson Sampling, and BestFirst.

Structs

AbExperimentResult

Result of one A/B experiment run (legacy).

AbOptimizerConfig

Configuration for the legacy prompt A/B optimizer.

ChainOfThoughtInjector

Wraps a prompt with chain-of-thought scaffolding.

FewShotExample

A single input/output example for few-shot prompting.

FewShotSelector

Selects the most relevant few-shot examples that fit within a token budget.

PromoterRegistry

Maps intent strings to the best known prompt variant (legacy).

PromptAbOptimizer

Orchestrates prompt A/B testing (legacy).

PromptCompressionOptimizer

Unified facade that combines compression, few-shot selection, and CoT injection into a single optimisation pipeline.

PromptCompressor

Applies lossless and near-lossless text compression to reduce token count.

PromptMutator

Applies simple transformations to prompt templates.

PromptOptimizer

Tracks prompt variants and selects the best using bandit algorithms.

PromptVariant

A single prompt variant being tracked by the optimizer.

ScoringEngine

Computes a weighted aggregate quality score from multiple metrics (legacy).

VariantGenerator

Generates N prompt variants from a base prompt (legacy).

VariantScore

Score for a single prompt variant (legacy).

Enums

CompressionGoal

High-level objective that guides how the compression optimizer applies its tools.

OptimizationStrategy

Strategy for selecting the next variant to evaluate.

PromptOptimizerError

Errors returned by the legacy prompt optimizer.

QualityMetric

A configurable quality metric used to score an inference response (legacy).

VariantStrategy

Strategies for generating prompt variants (legacy).