AI Search Core Library 🧠

A Context-Aware, Intent-Driven Search Engine designed for any domain (E-commerce, Travel, SaaS, etc.).

This library provides a flexible "brain" for your search bar. Unlike standard search engines (Algolia, Fuse.js) that only match text, AI Search Core understands:

1. Context: Remembers what the user just saw or asked.
2. Intent: Detects if user wants to "buy", "compare", or "find info".
3. Preferences: Boosts results based on user stats (e.g., "Apple fan", "Budget shopper").

Installation

npm install ai-search-core fuse.js

Quick Start

1. Define your Data

Your data can be anything. We just need a standard structure to index.

const products = [
  { id: 1, title: 'iPhone 15', category: 'Smartphone', price: 999 },
  { id: 2, title: 'MacBook Air', category: 'Laptop', price: 1200 },
  { id: 3, title: 'Samsung S24', category: 'Smartphone', price: 899 },
];

2. Configure the Brain

Define how the AI should "think".

import { CoreEngine, AIConfig } from 'ai-search-core';
import Fuse from 'fuse.js';

const config: AIConfig = {
    // 1. Synonym Mapping
    synonyms: {
        'hp': ['smartphone', 'phone', 'mobile'],
        'mac': ['macbook', 'laptop']
    },

    // 2. Define Intents (What is the user trying to do?)
    intents: [
        { name: 'check_price', patterns: ['how much', 'price', 'cost', 'biaya'] },
        { name: 'compare', patterns: ['vs', 'difference', 'compare', 'beda'] }
    ],

    // 3. Boosting Rules (Business Logic)
    boostingRules: [
        // Rule: If asking for price, boost items with prices displayed
        { 
            condition: (item, ctx, intent) => intent === 'check_price' && item.price > 0,
            score: 20 
        },
        // Rule: If user is an "Apple Fan" (tracked in context), boost Apple products
        {
            condition: (item, ctx) => ctx.userPreferences.brand === 'Apple' && item.title.includes('iPhone'),
            score: 50 
        }
    ]
};

3. Initialize & Search

const engine = new CoreEngine(products, Fuse, config);

// Simulation
const result = engine.search("berapa harga hp?");

console.log(result.intent); // 'check_price'
console.log(result.results); // [iPhone 15, Samsung S24, ...] (Boosted by intent)

Advanced Features

Context Memory

The engine remembers the conversation turn-by-turn.

// Turn 1
engine.search("Lihat iPhone 15"); 
// Engine remembers: lastTopic = 'Smartphone', lastItem = 'iPhone 15'

// Turn 2 (Follow-up)
engine.search("Warnanya apa aja?");
// Engine knows "Warnanya" refers to "iPhone 15" from previous turn.

Setting User Preferences Dynamically

You can feed external signals into the engine (e.g., from buttons or user profile).

// User clicked "Apple" filter button in UI
engine.setPreference('brand', 'Apple');

// Next search will heavily boost Apple products due to the BoostingRule we defined above.
engine.search("laptop terbaik"); // -> MacBook will be top result

API Reference

CoreEngine

• constructor(data, FuseClass, config)
• search(query: string): SearchResult
• setPreference(key, value): void

AIConfig

• synonyms: Record<string, string[]>
• stopWords: string[]
• intents: IntentDefinition[]
• boostingRules: BoostingRule[]

License

MIT