Embeddings

import { platform } from '@/lib/platform'

// Single text
const result = await platform.ai.embed({
  model: 'openai/text-embedding-3-small',
  input: 'The quick brown fox jumps over the lazy dog.',
})

console.log(result.embeddings[0]) // Float array [0.023, -0.012, ...]
console.log(result.dimensions)     // 1536

// Multiple texts (more efficient)
const results = await platform.ai.embed({
  model: 'openai/text-embedding-3-small',
  input: [
    'First document text...',
    'Second document text...',
    'Third document text...',
  ],
})

// results.embeddings[0], results.embeddings[1], results.embeddings[2]

// 1. Index your documents (do this once)
const documents = [
  { id: '1', text: 'How to reset your password' },
  { id: '2', text: 'Billing and payment methods' },
  { id: '3', text: 'Getting started with the API' },
]

const docEmbeddings = await platform.ai.embed({
  model: 'openai/text-embedding-3-small',
  input: documents.map(d => d.text),
})

// Store embeddings in your database (e.g., with pgvector)

// 2. Search at query time
const query = 'How do I change my password?'
const queryEmbedding = await platform.ai.embed({
  model: 'openai/text-embedding-3-small',
  input: query,
})

// Find similar documents using cosine similarity
// With pgvector: SELECT * FROM documents ORDER BY embedding <=> $1 LIMIT 5

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))
}

// Use it
const similarities = docEmbeddings.embeddings.map((docEmb, i) => ({
  id: documents[i].id,
  score: cosineSimilarity(queryEmbedding.embeddings[0], docEmb),
}))

const topResults = similarities.sort((a, b) => b.score - a.score).slice(0, 5)

// 1. User asks a question
const question = 'What is your refund policy?'

// 2. Find relevant documents
const queryEmbed = await platform.ai.embed({
  model: 'openai/text-embedding-3-small',
  input: question,
})

const relevantDocs = await db.query(`
  SELECT content FROM documents
  ORDER BY embedding <=> $1
  LIMIT 3
`, [queryEmbed.embeddings[0]])

// 3. Generate answer with context
const context = relevantDocs.map(d => d.content).join('\n\n')

const answer = await platform.ai.chat({
  model: 'anthropic/claude-3.5-sonnet',
  messages: [
    {
      role: 'system',
      content: `Answer the user's question based on the following context:

${context}

If the answer is not in the context, say you don't know.`,
    },
    { role: 'user', content: question },
  ],
})

// text-embedding-3-* supports dimension reduction
const result = await platform.ai.embed({
  model: 'openai/text-embedding-3-small',
  input: 'Your text here',
  dimensions: 512, // Reduce from 1536 to 512
})

// Smaller vectors = less storage, faster search
// Trade-off: slightly lower accuracy