Documentation
Get Started
Guide: Schema-Stitching

Guide: Schema-Stitching Project

Once you've created a Hive project of type Schema-Stitching, you can simply push your GraphQL schema to the registry. This guide will guide you through the basics of schema pushing, checking and fetching, and also the implementation of a simple schema-stitching gateway.

This guide is going to use the Stitching Directives approach (opens in a new tab) of Schema-Stitching; the advantage of this approach is that all schema and type merging configuration is represented in a single document managed by each subservice, and can be reloaded by the gateway without the need to specify any stitching configurations on the gateway.

Schemas and Services

For this guide, we are going to use the following schemas and services:

    • users.graphql
    • posts.graphql
  • Here's the GraphQL schema (SDL) for every service we are going to publish to Hive:

    The @merge directive denotes a root field used to query a merged type across services. The keyField argument specifies the name of a field to pick off origin objects as the key value.

    The @canonical directive specifies types and fields that provide a canonical definition (opens in a new tab) to be built into the gateway schema.

    You can read more about the Stitching Directives in the documentation. (opens in a new tab)

    subschemas/users.graphql
    directive @merge(
      keyField: String
      keyArg: String
      additionalArgs: String
      key: [String!]
      argsExpr: String
    ) on FIELD_DEFINITION
     
    directive @canonical on OBJECT | INTERFACE | INPUT_OBJECT | UNION | ENUM | SCALAR | FIELD_DEFINITION | INPUT_FIELD_DEFINITION
     
    type User {
      id: ID!
      username: String!
      email: String!
    }
     
    type Query {
      users(ids: [ID!]!): [User]! @merge(keyField: "id") @canonical
    }

    Hive CLI Access Token

    As you may have noticed, Hive has created three targets under your project: development, staging, and production. This guide will use the development target to explore the features Hive offers.

    To begin, select the development target under your Hive project and then choose the Settings tab. On the Settings screen, you can manage your target's settings and access tokens.

    Under the Registry Access Tokens section, click the Generate new token button. Give your access token a name and select Schema Check & Push from the list of token presets:

    Create Registry Token Form

    Click on Generate Token button and you should get your CLI token created, with permissions to publish GraphQL schemas.

    Make sure to copy your token and keep it safe. You won't be able to see it again.

    Note: You can create multiple tokens with different permissions for different use cases.

    Publish your schemas

    Now that you have your access token, you can publish your schema to the registry.

    We'll start with the Users service.

    Run the following command in your terminal, to publish your subschemas/users.graphql to the registry (replace YOUR_TOKEN_HERE with the token you have just created):

    hive schema:publish \
      --registry.accessToken YOUR_TOKEN_HERE \
      --service="users" \
      --url="http://fake.com/users/graphql" \
      --author "Me" \
      --commit "First" \
      subschemas/users.graphql

    If you are running under a NodeJS project, make sure to include the npx, yarn or pnpm prefix to the command.

    If everything goes well, you should see the following output:

     Published initial schema.

    If you'll get to your target's page on Hive dashboard, you should see that it was updated with the new schema you just published 🎉

    Stitching SDL

    You should also check out the Explorer tab: you can see the schema you just published, and explore the API types, fields, and arguments.

    Now, let's publish the Posts service schema to Hive:

    hive schema:publish \
      --registry.accessToken YOUR_TOKEN_HERE \
      --service="posts" \
      --url="http://fake.com/posts/graphql" \
      --author "Me" \
      --commit "Second" \
      subschemas/posts.graphql

    If everything goes well, you should see the following output:

     Schema published

    On your target's Explorer page now, you'll be able to see the schema of both services 🎉

    Schema Explorer

    Schema Checks

    Hive can perform several checks on your schema before publishing it to the registry. You can use Hive CLI to run these check and find potential breaking changes, and potential composition issues when a Stitching project is used.

    To see how schema checks works, let's make a small change to our schema. First, we'll start with a non-breaking change - we'll add a new field to the User type under the Users service:

    subschemas/users.graphql
    directive @merge(
      keyField: String
      keyArg: String
      additionalArgs: String
      key: [String!]
      argsExpr: String
    ) on FIELD_DEFINITION
     
    directive @canonical on OBJECT | INTERFACE | INPUT_OBJECT | UNION | ENUM | SCALAR | FIELD_DEFINITION | INPUT_FIELD_DEFINITION
     
    type User {
      id: ID!
      username: String!
      email: String!
      age: Int
    }
     
    type Query {
      users(ids: [ID!]!): [User]! @merge(keyField: "id") @canonical
    }

    Now, run the Hive CLI with the schema:check command and your modified subschemas/users.graphql file:

    hive schema:check \
      --registry.accessToken YOUR_TOKEN_HERE \
      --service="users" \
      subschemas/users.graphql

    You should see that Hive successfully detect the change you made, and exists with a 0 exit code, meaning that the schema is compatible, valid and has no breaking changes:

     Detected 1 change
     
    - Field age was added to object type User

    Now, are going to try introduce a breaking change. To do that, we'll rename an existing field in the GraphQL schema of the Users service:

    subschemas/users.graphql
    directive @merge(
      keyField: String
      keyArg: String
      additionalArgs: String
      key: [String!]
      argsExpr: String
    ) on FIELD_DEFINITION
     
    directive @canonical on OBJECT | INTERFACE | INPUT_OBJECT | UNION | ENUM | SCALAR | FIELD_DEFINITION | INPUT_FIELD_DEFINITION
     
    type User {
      id: ID!
      username: String!
      emailAddress: String! # was "email" before
    }
     
    type Query {
      users(ids: [ID!]!): [User]! @merge(keyField: "id") @canonical
    }

    Now, run the Hive CLI with the schema:check command again and the modified subschemas/users.graphql file:

    hive schema:check \
      --registry.accessToken YOUR_TOKEN_HERE \
      --service="users" \
      subschemas/users.graphql

    In that case, you'll notice that Hive CLI exists with a 1 exit code, meaning that the schema has breaking changes, and it's not compatible with the current schema in the registry:

     Detected 1 error
     
    - Breaking Change: Field email was removed from object type User
     
     Detected 2 changes
     
    - Field email was removed from object type User
    - Field emailAddress was added to object type User

    In addition to detecting simple breaking changes, Hive is capable of detecting composability issues across your services. To see how it works, let's make a small change to our schema.

    We are going to add a conflict to the Users service. We are going to add a new field (posts) to the User type, that conflicts with the User type in the Posts service.

    subschemas/users.graphql
    directive @merge(
      keyField: String
      keyArg: String
      additionalArgs: String
      key: [String!]
      argsExpr: String
    ) on FIELD_DEFINITION
     
    directive @canonical on OBJECT | INTERFACE | INPUT_OBJECT | UNION | ENUM | SCALAR | FIELD_DEFINITION | INPUT_FIELD_DEFINITION
     
    type User {
      id: ID!
      username: String!
      email: String!
      posts: Boolean # conflict with the "Posts" service
    }
     
    type Query {
      users(ids: [ID!]!): [User]! @merge(keyField: "id") @canonical
    }

    Run the Hive CLI with the schema:check command again and the modified subschemas/users.graphql file:

    hive schema:check \
      --registry.accessToken YOUR_TOKEN_HERE \
      --service="users" \
      subschemas/users.graphql

    And now you can see that the schema check process has failed, due to conflicts and inconsistencies between the schemas:

     Detected 2 errors
     
    - Failed to compare schemas: Failed to build schema: Error: Definitions of field User.posts implement inconsistent list types across subschemas and cannot be merged.
    - Definitions of field User.posts implement inconsistent list types across subschemas and cannot be merged.

    Evolve your schema

    Now that you have your schema published, you can evolve it over time. You can add new types, fields, and implement new capabilities for your consumers.

    Let's make a valid change in our schema and push it again to the registry:

    subschemas/users.graphql
    directive @merge(
      keyField: String
      keyArg: String
      additionalArgs: String
      key: [String!]
      argsExpr: String
    ) on FIELD_DEFINITION
     
    directive @canonical on OBJECT | INTERFACE | INPUT_OBJECT | UNION | ENUM | SCALAR | FIELD_DEFINITION | INPUT_FIELD_DEFINITION
     
    type User {
      id: ID!
      username: String!
      email: String!
      age: Int
    }
     
    type Query {
      users(ids: [ID!]!): [User]! @merge(keyField: "id") @canonical
    }

    And publish it to Hive:

    hive schema:publish \
      --registry.accessToken YOUR_TOKEN_HERE \
      --service="users" \
      --url="http://fake.com/users/graphql" \
      --author "Me" \
      --commit "Third" \
      subschemas/users.graphql

    You should see now that Hive accepted your published schema and updated the registry:

     Schema published

    It's a good timing to check the History tab of your Hive target. You should see that a new schema is published now, and you can see the changes you made:

    Schema History

    Fetch your schema

    Now that your GraphQL schema is stored in the Hive registry, you can access and fetch it through Hive's CDN (Content Delivery Network).

    The Hive Cloud service leverages the CloudFlare Global Network (opens in a new tab) to deliver your GraphQL schema and schema metadata. This means that your schema will be available from the nearest location to your GraphQL gateway, with 100% uptime, regardless of Hive's status. This ensures that everything required for your GraphQL API is always available, and reduces the risk of depending on Hive as a single point of failure. You can read more about Hive's CDN here.

    To get started with Hive's CDN access, you'll need to create a CDN token from your target's Settings page. You'll see a separate section for managing and creating CDN tokens, called CDN Access Token.

    Click on Create new CDN Token to create a new CDN token. Describe your token with an alias, and click Create. Please store this access token securely. You will not be able to see it again.

    Why are Registry and CDN tokens different?

    We use a separate, externally managed storage to manage CDN tokens to ensure high availability of your schemas. This approach ensures that your GraphQL schemas are fully secured and highly available. CDN tokens are read-only and can only fetch schemas from the CDN. They do not have permissions to perform any other action on your Hive organization.

    To use your access token, go to your target's page on Hive's dashboard and click on the Connect to CDN button. You will see a screen with instructions on how to obtain different types of artifacts from the CDN. For this guide, you can pick the GraphQL SDL artifact.

    CDN Token Form

    Copy the URL and let's try to fetch your schema using curl (replace YOUR_HIVE_CDN_TOKEN with your CDN token, and CDN_ENDPOINT_HERE with the endpoint you copied from Hive's dashboard):

    curl -L -H "X-Hive-CDN-Key: YOUR_HIVE_CDN_TOKEN" CDN_ENDPOINT_HERE

    You should see that Hive CDN returns your full GraphQL schema as an output for that command.

    A simple gateway

    Now that you have your schema published, and the schema is available on Hive's CDN, you can use it to build a simple GraphQL gateway.

    In addition to the SDL endpoint exposed from Hive's CDN, you can also use the /services endpoint to fetch a list of all services and their metadata (url and name).

    As part of Hive Client library, we are also exposing utilities for fetching and integrating your schema with your gateway.

    Here's a short snippet of how you can fetch the services information from Hive CDN, and start a GraphQL gateway using GraphQL-Yoga:

    import { createServer } from 'node:http'
    import { buildSchema } from 'graphql'
    import { createYoga } from 'graphql-yoga'
    import { createServicesFetcher } from '@graphql-hive/client'
    import { stitchSchemas } from '@graphql-tools/stitch'
    import { stitchingDirectives } from '@graphql-tools/stitching-directives'
     
    const { stitchingDirectivesTransformer } = stitchingDirectives()
     
    const fetchServices = createServicesFetcher({
      endpoint: process.env.HIVE_CDN_ENDPOINT,
      key: process.env.HIVE_CDN_KEY
    })
     
    async function main() {
      const services = await fetchServices()
      const subschemas = services.map(service => {
        return {
          schema: buildSchema(service.sdl, { assumeValid: true, assumeValidSDL: true }),
          executor: buildHTTPExecutor({ endpoint: service.url }),
          subschemaConfigTransforms: [stitchingDirectivesTransformer]
        }
      })
     
      const schema = stitchSchemas({
        subschemas
      })
      const yoga = createYoga({ schema })
      const server = createServer(yoga)
     
      server.listen(4000, () => {
        console.info('Gateway is running on http://localhost:4000/graphql')
      })
    }
     
    main().catch(err => {
      console.error(err)
      process.exit(1)
    })

    Make sure to add environment variables:

    HIVE_CDN_ENDPOINT is your Hive CDN endpoint (without the artficat type, just drop the /sdl or /services from the URL).

    HIVE_CDN_KEY is the Hive CDN Access Token you created earlier.

    Next Steps

    Now that you use the basic functionality of Hive as a schema registry, we recommend following other powerful features of Hive: