https://github.com/ddaras/melony
Fast, unopinionated, minimalist event-driven framework
https://github.com/ddaras/melony
ai event-driven event-streaming generative-ui genui react typescript vercel-ai-sdk
Last synced: 2 months ago
JSON representation
Fast, unopinionated, minimalist event-driven framework
- Host: GitHub
- URL: https://github.com/ddaras/melony
- Owner: ddaras
- Created: 2025-09-09T22:50:53.000Z (9 months ago)
- Default Branch: main
- Last Pushed: 2026-03-15T22:39:02.000Z (3 months ago)
- Last Synced: 2026-03-16T09:56:07.644Z (3 months ago)
- Topics: ai, event-driven, event-streaming, generative-ui, genui, react, typescript, vercel-ai-sdk
- Language: TypeScript
- Homepage: https://melony.dev
- Size: 4.92 MB
- Stars: 28
- Watchers: 0
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Melony
Fast, unopinionated, minimalist event-based framework for building AI agents.
[](https://discord.gg/j2uF5n8vJK)
Melony is to AI agents what Express is to web servers — a tiny, flexible orchestration loop: `Event → Handler → Events`.
## What you get
- **Event-first runtime**: a tiny orchestration loop: `Event → Handler → Events`.
- **Fluent Builder API**: build agents with a simple, type-safe API.
- **Plugin system**: easily modularize and reuse handlers across agents.
- **HITL-friendly**: approvals and guardrails belong in **event handlers**.
- **Frontend-ready**: `@melony/react` provides the glue (providers/hooks) to connect your React app to the Melony stream.
## Getting Started
The recommended way to start building a Melony app is by cloning our [Next.js Starter](https://github.com/ddaras/melony-nextjs-starter). It comes pre-configured with the core framework, React hooks, and an example agent.
```bash
git clone https://github.com/ddaras/melony-nextjs-starter
cd melony-nextjs-starter
pnpm install
pnpm dev
```
## Quick start (full stack)
### Backend (Next.js)
```ts
// app/api/chat/route.ts
import { melony } from "melony";
const agent = melony()
.on("user:text", async function* (event, { runtime }) {
// Simple router: respond with a greeting
yield { type: "assistant:text", data: { content: "Hey there!" } };
});
export async function POST(req: Request) {
const { event } = await req.json();
return agent.streamResponse(event);
}
```
### Frontend (React)
```tsx
import { MelonyClient } from "melony/client";
import { MelonyProvider, useMelony } from "@melony/react";
const client = new MelonyClient({ url: "/api/chat" });
function Chat() {
const { messages, send } = useMelony();
return (
{messages.map(m => (
{m.role}: {m.content}
))}
send({ type: "user:text", data: { content: "Hello!" } })}>
Send
);
}
export default function App() {
return (
);
}
```
## Packages
- **`melony`**: The core framework: builder, runtime, and streaming utilities.
- **`melony-react`**: React hooks and providers to connect your app to a Melony stream.
## Apps in this repo
- **`apps/landing`**: Minimal marketing site for Melony (Vite + React).
- **`apps/studio`**: Interactive UI for working with Melony flows.
- **`apps/express`**: Backend sandbox for runtime and workflow experiments.
## Why Melony
Most "agent frameworks" are heavy and opinionated about how you should build your agent's brain. Melony is different:
**It focuses on the communication protocol.** By treating everything as an event stream, Melony makes it easy to bridge the gap between LLM thoughts and the actual product UX your users need—whether that's text, structured data, or complex interactive flows.
## Community
Join our Discord to connect with the team and other developers: [https://discord.gg/j2uF5n8vJK](https://discord.gg/j2uF5n8vJK)