Embeddings y Búsqueda Semántica

Embeddings: vectores que capturan significado

Generar embeddings con OpenAI

import OpenAI from 'openai'

const openai = new OpenAI()

export async function generateEmbedding(text: string): Promise<number[]> {
  const response = await openai.embeddings.create({
    model: 'text-embedding-3-small',
    input: text,
  })
  return response.data[0].embedding
}

// Batch para eficiencia
export async function generateEmbeddings(texts: string[]): Promise<number[][]> {
  const response = await openai.embeddings.create({
    model: 'text-embedding-3-small',
    input: texts,
  })
  return response.data.map(d => d.embedding)
}

Búsqueda por similitud coseno

// Similitud coseno manual (para entender)
export function cosineSimilarity(a: number[], b: number[]): number {
  let dotProduct = 0
  let normA = 0
  let normB = 0

  for (let i = 0; i < a.length; i++) {
    dotProduct += a[i] * b[i]
    normA += a[i] * a[i]
    normB += b[i] * b[i]
  }

  return dotProduct / (Math.sqrt(normA) * Math.sqrt(normB))
}

// Búsqueda en array (para datasets pequeños)
export function searchSimilar(
  query: number[],
  documents: Array<{ id: string; embedding: number[]; content: string }>,
  topK: number = 5
) {
  const scored = documents.map(doc => ({
    ...doc,
    score: cosineSimilarity(query, doc.embedding),
  }))

  return scored
    .sort((a, b) => b.score - a.score)
    .slice(0, topK)
}

-- SQL para buscar similares en Supabase
select
  id,
  content,
  1 - (embedding <=> $1::vector) as similarity
from documents
order by embedding <=> $1::vector
limit 5;

Tip Pro