GQLPT

Leverage AI to generate GraphQL queries from plain text.

Read more about GQLPT on our blog rconnect.tech/blog/gqlpt

Image showing the online playground for gqlpt.dev. GQLPT is a npm package that allows you to generate GraphQL queries from plain text using AI.

Quick View

• EcoSystem
• Installation
• Usage
• From Introspection
• Generate and Send
• Type Generation
• FAQs
• License

EcoSystem

gqlpt

• Core library for generating GraphQL queries from plain text. GitHub

@gqlpt/cli

• TypeScript type generation for GQLPT. GitHub

@gqlpt/adapter-anthropic

• Adapter for Anthropic AI. GitHub

@gqlpt/adapter-openai

• Adapter for OpenAI GPT-3. GitHub

Installation

https://www.npmjs.com/package/gqlpt

npm install gqlpt @gqlpt/adapter-openai

Usage

import { AdapterOpenAI } from "@gqlpt/adapter-openai";

import { GQLPTClient } from "gqlpt";

const typeDefs = /* GraphQL */ `
  type User {
    id: ID!
    name: String!
  }

  type Query {
    user(id: ID!): User
  }
`;

const client = new GQLPTClient({
  typeDefs,
  adapter: new AdapterOpenAI({
    apiKey: process.env.OPENAI_API_KEY,
  }),
});

async function main() {
  await client.connect();

  const query = "Find users by id 1";

  const response = await client.generateQueryAndVariables(query);

  console.log(response);
  /*
    {
        query: 'query ($id: ID!) {\n  user(id: $id) {\n    id\n    name\n  }\n}',
        variables: { id: '1' }
    }
  */
}

main();

From Introspection

You can specify a URL and headers in the options to perform introspection on the .connect command. This allows you to fetch the GraphQL schema directly from your endpoint.

Note: When specifying a URL for initial introspection, you must call the connect method.

const client = new GQLPTClient({
  url: "http://localhost:4000/graphql", // Your GraphQL endpoint
  headers: {
    Authorization: `Bearer ${process.env.ACCESS_TOKEN}`,
  },
  adapter: new AdapterOpenAI({
    apiKey: process.env.OPENAI_API_KEY,
  }),
});

await client.connect(); // Performs introspection using the URL

Generate and Send

The GQLPTClient class includes a generateAndSend method that simplifies generating a GraphQL query from plain text and sending it directly to a specified endpoint. This method leverages the generateQueryAndVariables function and then posts the generated query to the endpoint.

Example

import { AdapterOpenAI } from "@gqlpt/adapter-openai";

import { GQLPTClient } from "gqlpt";

const client = new GQLPTClient({
  url: "http://localhost:4000/graphql",
  adapter: new AdapterOpenAI({
    apiKey: process.env.OPENAI_API_KEY,
  }),
});

async function main() {
  await client.connect();

  const response = await client.generateAndSend("Find users by id 1");

  console.log(response); // Logs the server's response to the query

  /*
    Example response structure:
    {
      "data": {
        "user": {
          "id": "1",
          "name": "John Doe"
        }
      },
      "errors": []
    }
  */
}

main();

Advanced Usage

You can override the URL or headers when calling generateAndSend:

const response = await client.generateAndSend("Find users by id 1", {
  urlOverride: "http://another-server.com/graphql",
  headersOverride: {
    Authorization: "Bearer custom-token",
  },
});

Type Generation

GQLPT offers seamless type generation without requiring changes to your existing code. The CLI tool automatically analyzes your codebase, generates appropriate types based on your usage patterns, and updates the type definitions in your node_modules.

Installation

npm install -g @gqlpt/cli

Usage

Run the following command in your project root:

npx @gqlpt/cli generate ./src

This command will:

1. Scan your ./src directory for GQLPT usage
2. Generate TypeScript types based on your plain text queries
3. Update the types in node_modules/gqlpt/build/types.d.ts

You don't need to manually import or reference these types in your code. GQLPT will automatically use them to provide type safety.

Example: Querying GitHub API with Automatic Type Safety

Here's an example of how to use GQLPT with the GitHub GraphQL API, leveraging automatically generated types:

import { AdapterOpenAI } from "@gqlpt/adapter-openai";

import { GQLPTClient } from "gqlpt";

const client = new GQLPTClient({
  url: "https://api.github.com/graphql",
  headers: {
    Authorization: `Bearer ${process.env.GITHUB_TOKEN}`,
  },
  adapter: new AdapterOpenAI({
    apiKey: process.env.OPENAI_API_KEY,
  }),
});

async function searchGitHubRepos() {
  await client.connect();

  const query = "Find repositories with the name gqlpt";

  const response = await client.generateAndSend(query);
  // response will be typed based on the generated types
}

In this example:

1. We create a GQLPTClient instance, specifying the GitHub GraphQL API endpoint and including an authorization header with a GitHub token.

2. We don't need to explicitly import or specify any type information. The types are automatically applied based on the generated definitions.

3. After connecting to the client (which performs introspection on the GitHub API), we use a natural language query to search for repositories with "gqlpt" in the name.

4. The generateAndSend method generates the appropriate GraphQL query, sends it to the GitHub API, and returns the result with full type information.

5. TypeScript provides full type safety and autocompletion when working with the response, based on the automatically generated types.

This seamless integration of type generation allows you to leverage the power of TypeScript's type system without any additional overhead in your development process. Just run the CLI tool whenever you update your GraphQL schema or change your GQLPT usage patterns, and enjoy automatic type safety in your code.

FAQs

1. I'm seeing the error "429 - Rate limit reached for requests" on making requests.

This error is related to sending too many requests to OpenAI, and them trying to rate-limit you. Try limiting the requests to gqlpt, or upgrading to a paid OpenAI API plan. In case of persistent errors, create an issue with us, or reach out to OpenAI support for OpenAI related errors.

License

MIT - Rocket Connect - https://github.com/rocket-connect