Lightweight JSON-RPC solution for TypeScript projects
- 👩🔧 Service definition through Zod-based contracts
- 📜 JSON-RPC 2.0 protocol
- 🕵️ Full IDE autocompletion
- 🪶 Tiny footprint (< 1kB)
- 🌎 Support for Deno and edge runtimes
- 🚫 No code generation step
Define a shared service contract and export the type of service:
import { z } from "zod";
import type { ZServiceType } from "@lorefnon/ts-json-rpc/lib/zod";
import { ZService } from "@lorefnon/ts-json-rpc/lib/zod";
export const MyServiceDef = ZService.define({
hello: z.function()
.args(z.string())
.returns(z.string().promise()),
// More methods here ...
});
export type MyService = ZServiceType<typeof MyServiceDef>
// { hello: (arg: string) => Promise<string>, ... }
Define a server-side implementation of this service:
export const MyServiceImpl = MyServiceDef.implement(() => ({
async hello(name) { // name is inferred as string
return `Hello ${name}!`;
},
// Implement other methods...
}));
Create a server with a route to handle the API requests:
import express from "express";
import { rpcHandler } from "@lorefnon/ts-json-rpc/lib/express";
const app = express();
app.use(express.json());
app.post("/api", rpcHandler(MyServiceImpl));
app.listen(3000);
Note You can also use @lorefnon/ts-json-rpc in servers other than Express. Check out to docs below for examples.
On the client-side, import the shared type and create a typed rpcClient
with it:
import { rpcClient, HttpPostTransport } from "@lorefnon/ts-json-rpc/lib/client";
// Import the type (not the implementation!)
import type { MyService } from "../shared/MyService";
// Create a typed client:
const client = rpcClient<MyService>({
transport: new HttpPostTransport({
url: "http://localhost:3000/api"
})
});
// Call a remote method:
console.log(await client.hello("world"));
That's all it takes to create a type-safe JSON-RPC API. 🎉
You can play with a live example over at StackBlitz:
It is common for services to be created per request. DefaultServiceImpl
in above example
is a service factory ie. a function that creates and returns an implementation of the service contract.
rpcHandler will invoke this function for every request to create a service object that handles that particular request.
This is convenient if you need to access the request (more on this below) but if you don't, instead of a function you could also create an instance and pass that to rpcHandler:
app.post("/api", rpcHandler(DefaultServiceImpl({})));
Now, we have a singleton service that handles all requests.
The function passed to ServiceDef.implement
can accept a context argument which is available to all the methods.
interface ServiceContext {
currentUser?: { name: string }
}
export const MyServiceImpl = ServiceDef.implement((context: ServiceContext) => ({
async hello() {
return `Hello ${context.currentUser?.name ?? "Stranger"}!`;
},
// Implement other methods...
}));
You are responsible for passing this context
to DefaultServiceImpl
.
Most common use case for context is to get access to the request object.
So, by default rpcHandler will simply pass the request object to service factory as context.
However, if you want to ensure that your service implementation is not tied to a specific server implementation (eg. express) you can also extract what you need from the request and pass it to the service factory.
app.post(
"/api",
rpcHandler((req) => MyServiceImpl(req.headers))
);
This is also useful if you need to inject any additional objects (eg. database pool instance) into the service.
The generic @lorefnon/ts-json-rpc/server
package can be used with any server framework or (edge-) runtime.
With Fastify, you would use @lorefnon/ts-json-rpc
like this:
import { handleRpc, isJsonRpcRequest } from "@lorefnon/ts-json-rpc/lib/server";
fastify.post("/api", async (req, reply) => {
if (isJsonRpcRequest(req.body)) {
const res = await handleRpc(req.body, Service(req));
reply.send(res);
}
});
A client can send custom request headers by providing a getHeaders
function:
const client = rpcClient<MyService>(apiUrl, {
getHeaders() {
return {
Authorization: auth,
};
},
});
Note The
getHeaders
function can also beasync
.
To include credentials in cross-origin requests, pass credentials: 'include'
as option.
While @lorefnon/ts-json-rpc
itself does not provide any built-in UI framework integrations,
you can pair it with react-api-query,
a thin wrapper around TanStack Query. A type-safe match made in heaven. 💕
MIT
This implementation is based on past work by Felix Gnass in typed-rpc.
The typed-rpc repo is more minimal in its focus (eg. runtime type checking is explicitly not a goal) and does not appear to be accepting pull requests.