Why this package exists

As prompts grow in complexity, they often accumulate filler words, redundant context, and weak constraints. This inflates token costs and can confuse LLMs, degrading output quality.

@promptforgee/optimizer takes a @promptforgee/core Prompt instance and returns a highly optimized equivalent. It trims the fat, deduplicates information, and transforms passive constraints into strict directives, ensuring your prompts are as lean and effective as possible.

Features

• Token Reduction: Strips out filler words ("please", "can you", etc.) without losing intent.
• Deduplication: Merges identical or highly similar context blocks and constraints.
• Constraint Strengthening: Converts weak language ("try to avoid") into strict imperatives ("DO NOT").
• Seamless Integration: Operates directly on Prompt instances from @promptforgee/core.

Installation

# npm
npm install @promptforgee/optimizer

# pnpm
pnpm add @promptforgee/optimizer

# yarn
yarn add @promptforgee/optimizer

# bun
bun add @promptforgee/optimizer

Quick Start

import { Prompt, MarkdownFormatter } from '@promptforgee/core';
import { optimizePrompt } from '@promptforgee/optimizer';

async function main() {
  // A poorly constructed, verbose prompt
  const messyPrompt = Prompt.create()
    .task('Can you please write a function to sort an array for me?')
    .constraint('Try to avoid using O(n^2) time complexity if you can.')
    .constraint("Please don't use O(n^2) time.") // Redundant
    .context('The array contains numbers.')
    .context('The array contains numbers.'); // Redundant

  console.log('--- Before Optimization ---');
  console.log(messyPrompt.build(new MarkdownFormatter()));

  // Optimize the prompt
  const optimizedPrompt = await optimizePrompt(messyPrompt);

  console.log('\n--- After Optimization ---');
  console.log(optimizedPrompt.build(new MarkdownFormatter()));
}

main();

API Overview

optimizePrompt(prompt: Prompt): Promise<Prompt>

The primary exported function. Accepts a Prompt instance and returns a new, optimized Prompt instance.

HeuristicOptimizer

The underlying optimizer class. Currently utilizes heuristic rules (Regex, deduplication logic) to clean the prompt state.

| Method | Description | | --------------------------- | --------------------------------------- | | .optimize(prompt: Prompt) | Returns an optimized Prompt instance. |

Real-world Example

Building a self-healing prompt pipeline where generated prompts are automatically cleaned before execution:

import { Prompt } from '@promptforgee/core';
import { optimizePrompt } from '@promptforgee/optimizer';

// User-provided inputs are often messy
function createQueryPrompt(userTask: string, extraContext: string[]) {
  const basePrompt = Prompt.create().task(userTask);

  extraContext.forEach((ctx) => basePrompt.context(ctx));

  // Clean up user messiness automatically
  return optimizePrompt(basePrompt);
}

Ecosystem

@promptforgee/optimizer is the refinement step in the PromptForge lifecycle.

@promptforgee/core (Builds the prompt) ↓ @promptforgee/analyzer (Analyzes the prompt) ↓ @promptforgee/optimizer (You are here) ↓ @promptforgee/registry (Stores the optimized prompt)

Documentation

For full documentation and advanced usage, visit promptforge.dev/docs/optimizer.

Examples

Check out our Examples directory for more real-world use cases.

Roadmap

• 🚧 LLM-backed optimization engine for deep semantic deduplication.
• 🚧 Configuration options to tune optimization aggressiveness.

Contributing

We welcome contributions! Please read our Contributing Guide to get started.

License

MIT © Omnikon-Org