Working with resolvers and Vercel Postgres

Resolvers are a powerful way to extend your Grafbase backend. We can use resolvers to compute custom business logic, make network requests, invoke dependencies, and perform various operations.

In this guide, we'll create a new GraphQL API using Grafbase with resolvers and leverage Vercel Postgres to handle OpenAI vector embeddings.

To get started, create a new directory or navigate to an existing project and run the following command:

npx grafbase init

We'll create a system to store and query vector embeddings generated by the OpenAI API.

Open the file grafbase/schema.graphql and replace the contents with the following schema:

type Embedding {
  id: ID!
  text: String!
  vector: JSON!
}

extend type Query {
  getEmbedding(id: ID!): Embedding @resolver(name: "get-embedding")
  searchEmbeddings(query: String!): [Embedding!]!
    @resolver(name: "search-embeddings")
}

extend type Mutation {
  createEmbedding(input: EmbeddingInput!): Embedding!
    @resolver(name: "create-embedding")
}

input EmbeddingInput {
  text: String!
}

Go to Vercel and create an account if you don't have one. Then, you can follow the quick start guide on Vercel.

Once your database is created, and your .env file populated by using vercel env pull .env.development.local, you are ready to continue, @vercel/postgres automatically retrieves the connection string to connect to the db from the .env file.

Before continuing, you must initialize the table that we are going to use to store our data, for that use any Postgres client and run:

CREATE EXTENSION vector;

CREATE TABLE IF NOT EXISTS embeddings (
    id SERIAL PRIMARY KEY,
    text TEXT NOT NULL,
    vector VECTOR(1536) NOT NULL
);

We'll now create the resolvers for handling the vector embeddings.

First install @vercel/postgres by typing yarn add @vercel/postgress or npm install @vercel/postgress

Create the file grafbase/resolvers/get-embedding.js and add the following code:

import { db } from '@vercel/postgres'

export default async function Resolver(_, { id }) {
  const client = await db.connect()

  return await client.sql`SELECT * FROM embeddings WHERE id = ${id}`
}

Create another file grafbase/resolvers/search-embeddings.js and type:

import { db } from '@vercel/postgres'

export default async function Resolver(_, { query }) {
  const client = await db.connect()

  return await client.sql`SELECT * FROM embeddings WHERE text ILIKE '%${query}%'`
}

Finally, create the file grafbase/resolvers/create-embedding.js:

import { db } from '@vercel/postgres'

export default async function Resolver(_, { input }) {
  const client = await db.connect()

  const { text } = input

  // Generate the vector using OpenAI embeddings API
  const response = await fetch(
    'https://api.openai.com/v1/engines/davinci-codex/completions',
    {
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
        // You must define OPENAI_API_KEY on your .env file
        Authorization: `Bearer ${process.env.OPENAI_API_KEY}`,
      },
      body: JSON.stringify({
        prompt: text,
        max_tokens: 1,
      }),
    },
  )

  const data = await response.json()
  const vector = data.choices[0].text

  return await client.sql`INSERT INTO embeddings (text, vector) VALUES ('${text}', '${vector}') RETURNING *`
}

Start the Grafbase development server by running the following command in your project directory:

npx grafbase dev

You can now interact with the GraphQL API by visiting http://localhost:4000 in your browser.

To create an embedding, run the following mutation:

mutation {
  createEmbedding(input: { text: "Embedding Example" }) {
    text
  }
}

This will create a new embedding with the provided text and vector.

To retrieve an embedding by its ID, run the following query:

query {
  getEmbedding(id: "EMBEDDING_ID") {
    id
    text
    vector
  }
}

Replace EMBEDDING_ID with the actual ID of the embedding you want to retrieve.

To search for embeddings based on a query string, run the following query:

query {
  searchEmbeddings(query: "Example") {
    id
    text
    vector
  }
}

Replace "Example" with your desired query string.

That's it! You now have a Grafbase backend with resolvers connected to Vercel Postgres, allowing you to store and query OpenAI vector embeddings. Feel free to explore and expand upon this setup to suit your needs.