Interface IPromptOptimizer<T>

Namespace

AiDotNet.Interfaces

Assembly

AiDotNet.dll

Defines the contract for optimizing prompts to improve language model performance.

public interface IPromptOptimizer<T>

Type Parameters

T

The type of numeric data used for scoring.

Remarks

A prompt optimizer automatically refines prompts to achieve better performance on a specific task. Optimization strategies include discrete search, gradient-based methods, ensemble approaches, and evolutionary algorithms.

For Beginners: A prompt optimizer automatically improves your prompts.

Think of it like automatic recipe refinement:

• You start with a basic recipe
• The optimizer tries variations (more salt, less sugar, different temperature)
• It measures which variations taste better
• It keeps refining until it finds the best version

For prompts:

• You provide a basic prompt
• Optimizer generates variations
• Tests each variation's performance
• Returns the best-performing prompt

Example: Initial prompt: "Classify this review" After optimization: "Carefully analyze the sentiment and tone of the following product review. Classify it as positive, negative, or neutral based on the overall customer satisfaction." Result: 15% accuracy improvement

Benefits:

• Better results without manual trial-and-error
• Discover optimal phrasings you wouldn't think of
• Systematic improvement process
• Measurable performance gains

Methods

GetOptimizationHistory()

Gets the optimization history showing performance over iterations.

IReadOnlyList<OptimizationHistoryEntry<T>> GetOptimizationHistory()

Returns

IReadOnlyList<OptimizationHistoryEntry<T>>

Remarks

Returns a list of prompts and their scores from the optimization process, allowing analysis of how optimization progressed.

For Beginners: Shows how the prompt improved during optimization.

Like a training log showing progress:

• Iteration 1: "Classify sentiment" → Score: 0.65
• Iteration 5: "Classify the sentiment of" → Score: 0.72
• Iteration 10: "Analyze sentiment and classify as" → Score: 0.78
• Iteration 20: "Carefully analyze the sentiment..." → Score: 0.85

Use this to:

• Visualize improvement over time
• Understand what changes helped
• Debug optimization issues
• Decide if more iterations would help

Example:

var history = optimizer.GetOptimizationHistory();

foreach (var entry in history)
{
    Console.WriteLine($"Iteration {entry.Iteration}: Score {entry.Score}");
    Console.WriteLine($"Prompt: {entry.Prompt}");
}

// Plot improvement curve
PlotScores(history.Select(h => h.Score));

Optimize(string, Func<string, T>, int)

IPromptTemplate Optimize(string initialPrompt, Func<string, T> evaluationFunction, int maxIterations = 100)

Parameters

initialPrompt string

evaluationFunction Func<string, T>

maxIterations int

Returns

IPromptTemplate

OptimizeAsync(string, Func<string, Task<T>>, int, CancellationToken)

Task<IPromptTemplate> OptimizeAsync(string initialPrompt, Func<string, Task<T>> evaluationFunction, int maxIterations = 100, CancellationToken cancellationToken = default)

Parameters

initialPrompt string

evaluationFunction Func<string, Task<T>>

maxIterations int

cancellationToken CancellationToken

Returns

Task<IPromptTemplate>