Serving & Client

Melony is designed to be flexible. It provides a core runtime that can be served in various environments and consumed by different clients.

Serving with `.streamResponse()`

The simplest way to serve a Melony agent is using the .streamResponse() method on the builder. It automatically executes the runtime and returns a streaming Response.

Next.js Example

// app/api/chat/route.ts
import { agent } from "./agent";

export async function POST(req: Request) {
  const { event } = await req.json();
  return agent.streamResponse(event);
}

Manual Streaming with `createStreamResponse`

If you need more control (e.g., using a custom runtime instance), you can use the built-in createStreamResponse utility from melony. It converts an async generator into a streaming Response.

import { createStreamResponse } from "melony";
import { agent } from "./agent";

export async function POST(req: Request) {
  const { event } = await req.json();
  const runtime = agent.build();
  return createStreamResponse(runtime.run(event));
}

Melony Client

The MelonyClient is a lightweight, framework-agnostic library for communicating with your Melony server. It is available via the melony/client export.

import { MelonyClient } from "melony/client";

const client = new MelonyClient({
  url: "/api/chat",
  // Optional: custom headers for auth
  headers: () => ({
    Authorization: `Bearer ${getToken()}`,
  }),
});

`send(event: Event)`

Sends an event to the server and returns an AsyncGenerator of events.

const stream = client.send({
  type: "user:text",
  data: { content: "Hello!" },
});

for await (const event of stream) {
  // Handle events as they arrive
  console.log(event);
}

Configuration Options

url: The endpoint of your Melony server.
headers: A static object or a function that returns headers (useful for dynamic tokens).

Serving with .streamResponse()