VAULT // NEXT: Developer Docs Starter

Overview

Embed documents

Convert your document corpus into vector embeddings using the Vault embedding endpoint.

Store in a vector DB

Upsert embeddings into a vector database (Pinecone, pgvector, Qdrant).

Retrieve at query time

Embed the user query and retrieve the top-k most similar document chunks.

Generate with context

Pass the retrieved chunks as context to vault.infer() to produce a grounded response.

Embeddings

lib/embed.ts

import { vault } from '@/lib/vault';

export async function embedDocuments(chunks: string[]): Promise<number[][]> {
  const results = await Promise.all(
    chunks.map((chunk) =>
      vault.embed({
        model: 'vault-embed-v1',
        input: chunk,
      })
    )
  );
  return results.map((r) => r.embedding);
}

export async function embedQuery(query: string): Promise<number[]> {
  const result = await vault.embed({
    model: 'vault-embed-v1',
    input: query,
  });
  return result.embedding;
}

Retrieval

lib/retrieve.ts

import { embedQuery } from './embed';
import { vectorDb } from './vector-db'; // your DB client

export async function retrieve(query: string, topK = 5): Promise<string[]> {
  const queryEmbedding = await embedQuery(query);

  const results = await vectorDb.query({
    vector: queryEmbedding,
    topK,
    includeMetadata: true,
  });

  return results.matches.map((m) => m.metadata.text as string);
}

Generation

lib/generate.ts

import { vault } from '@/lib/vault';

export async function generate(query: string, context: string[]): Promise<string> {
  const contextBlock = context
    .map((c, i) => `[${i + 1}] ${c}`)
    .join('\n\n');

  const result = await vault.infer({
    model: 'vault-3-pro',
    prompt: `Answer the question using only the provided context. \
If the context does not contain the answer, say so.

Context:
${contextBlock}

Question: ${query}`,
    maxTokens: 500,
  });

  return result.text;
}

Full Pipeline

app/api/rag/route.ts

import { retrieve } from '@/lib/retrieve';
import { generate } from '@/lib/generate';

export async function POST(req: Request) {
  const { query } = await req.json();

  const context  = await retrieve(query);
  const response = await generate(query, context);

  return Response.json({ response, sources: context.length });
}

For production, add source citations by storing document IDs alongside embeddings and returning them with the retrieved chunks.