@anvia/pgvector

Store and search Anvia documents in Postgres with pgvector. Pair the store with your choice of embedding model to add semantic retrieval to agents and applications.

Install

pnpm add @anvia/pgvector @anvia/core @anvia/openai

@anvia/openai supplies the embedding model in this example; other embedding adapters work too.

Quickstart

Set DATABASE_URL and OPENAI_API_KEY. Your Postgres server must provide the pgvector extension. The database role needs permission to create the extension and table when using ensure().

import { embedDocuments } from "@anvia/core/embeddings";
import { retrieveDocuments } from "@anvia/core/vector-store";
import { OpenAIClient } from "@anvia/openai";
import { PgVectorClient } from "@anvia/pgvector";

const openai = new OpenAIClient({ apiKey: process.env.OPENAI_API_KEY! });
const model = openai.embeddingModel({ modelId: "text-embedding-3-small" });
const client = new PgVectorClient({ connectionString: process.env.DATABASE_URL! });
const store = client.vectorStore<{ id: string; text: string }>({
  tableName: "support_docs",
  dimensions: 1536,
  metric: "cosine",
});

try {
  await store.ensure();
  const { documents } = await embedDocuments({
    model,
    documents: [{ id: "password-reset", text: "Reset links expire after 30 minutes." }],
    id: (document) => document.id,
    content: (document) => document.text,
  });
  await store.upsert({ documents });

  const results = await retrieveDocuments({
    store,
    model,
    query: "How long does a reset link last?",
    topK: 3,
  });
  console.log(results);
} finally {
  await client.close();
}

Store capabilities

• Explicit provisioning with ensure() and readiness checks with validate().
• Document replacement, vector search, and metadata filtering.
• Bring your own embedding model; store.search() accepts vectors directly.

Keep semantic retrieval alongside application data in Postgres.

Pass client to use a preconfigured native SDK client. Injected clients remain caller-owned.

Learn more

• Anvia overview
• Contributing