Serve Mesh v1 Graph with Yoga

[BabakScript]

Hi @ardatan,

Is there any way to serve the Mesh v1 unified graph via Yoga directly rather than generating the unified graph with mesh-compose cli and serving it via Hive?

I'm looking for something like this:

import { getComposedSchemaFromConfig } from "@graphql-mesh/compose-cli";
import { composeConfig } from "./mesh.config";
import { createServer } from "node:http";
import { createYoga } from "graphql-yoga";

async function startServer() {
  logger.info("Starting server...");
  const meshInstance = await getComposedSchemaFromConfig(
    composeConfig,
    {} as any,
  );

  // Create a Yoga instance with a GraphQL schema.
  const yoga = createYoga({ schema: meshInstance }); // <<<<<<<<<<<<<<<<<<<<<<<<<<<<< Send unified graph to Yoga

  // Pass it into a server to hook into request handlers.
  const server = createServer(yoga);

  // Start the server and you're done!
  server.listen(4000, () => {
    console.info("Server is running on http://localhost:4000/graphql");
  });
}

startServer();

But it seems passing the getComposedSchemaFromConfig function output like this is not working 🤔

Type 'string' is not assignable to type 'YogaSchemaDefinition<{}, {}>'.

30   const yoga = createYoga({ schema: meshInstance });

In Mesh v0, we used const { getEnveloped, schema } = await getMesh(meshConfig); to serve the generated graph on the fly without saving that in a schema registry, and I'm looking for a way to do the same in Mesh V1.

Any idea on this?

[ardatan]

No, the only way of serving them is to use Hive Gateway.
The generated SDL should be saved using Mesh Compose CLI, and Hive Gateway should be pointed to that generated SDL. Registry is not needed.

[BabakScript]

Thanks for your reply. After checking the new versions of Hive Gateway and GraphQL Mesh, and Schema Stitching now, I'm a bit confused about which one I really need to choose and upgrade to from Mesh V0.

In the past, it was clear and easy to choose Mesh for my use case because I have a couple of small GraphQL APIs, and what I need is to just unify all of these small graphs and apply some custom logic for some queries with custom resolvers without changing schema and expose the unified graph. So, I used Mesh as a proxy gateway to aggregate all small graphs and apply some custom logic on the fly to some queries.
But now, in Mesh V1, there is no way to serve the unified graph so I have to save and expose the unified graph somewhere to make it visible for Hive to serve it.

As I checked, it seems Hive supports "proxy" mode to act as a proxy BUT I'm not sure if it supports multiple sources to do the unification phase in Hive too. (so it seems Hive is not a full replacement for Mesh)

So maybe a solution is to unify all the sources in an interval, and save them with Mesh, and then serve them by Hive. (it's not ideal because, for a proxy like this, I prefer not to store the schema in both Mesh and Hive levels)

I continue searching between Hive Gateway and GraphQL Mesh, and Schema Stitching to see if I can handle the entire solution with only one of them rather than a combo solution, as I always prefer a simpler setup but I'm wondering what you think about this use case and which one is the best fit based on your knowledge?

[ardatan]

Today you generate a supergraph SDL with Mesh Compose as supergraph.graphql, then serve it with Hive Gateway with hive-gateway supergraph.
Mesh Compose acts like mesh build and Hive Gateway acts like mesh start. You can see the following working example.
The setup is pretty simple actually.
https://github.com/graphql-hive/gateway/tree/main/examples/openapi-javascript-wiki

[BabakScript]

Thanks for your reply I'll take a look at this example. 🙏