All projects

Open source · Developer tools

NLUX.

An open-source JavaScript and TypeScript library for bringing conversational AI into web products. I created and maintain the library, its documentation, and its community support.

DocumentationGitHub
Period
Nov 2023 – Present
My role
Founder & maintainer
Status
Active
NLUXNLUXConversational AI interfaces
NLUX Nova theme showing conversation layouts in light and dark modes
NLUX’s Nova theme — conversation layouts in light and dark modes, from the project documentation.
Weekly npm downloads
10K+
GitHub stars
1.2K+
Supported interfaces
JS / React

The idea

Make conversational interfaces easier to build.

Adding AI to a product involves more than calling a model. Developers need to handle streaming responses, manage the conversation, and make the interface fit their application.

I created NLUX to package that work into a reusable library, with a quick setup and room to adapt the experience as a product grows.

The library

One interface. A choice of backends.

NLUX separates the interface, API, and model adapters. React and vanilla JavaScript packages let developers use the library within their existing applications. The core has no runtime dependencies or external UI libraries.

  • Streaming responses

    Render responses as they arrive, with the conversation handled in the interface.

  • Pluggable adapters

    Connect OpenAI, LangChain, Hugging Face, or a custom inference backend.

  • Themes and personas

    Customise layouts, assistant identities, and light or dark themes to fit the surrounding product.

  • Generative UI

    Use the React Server Components adapter to render interactive components as chat responses, including integrations with the Vercel AI SDK.

In a product

An assistant that understands the screen.

NLUX also supports conversations that use application context and trigger actions. The Stock Wiz demo shows an assistant alongside a portfolio table: a request can apply a stock-exchange filter, and a question about a ticker can be answered using the current view.

React hooks register relevant context and available tasks, while an adapter and backend middleware connect them to the AI system. The same approach can make the assistant part of a product’s workflow.

Stock Wiz demo: an NLUX assistant applies a Hong Kong exchange filter and explains a stock ticker beside a portfolio table
Stock Wiz — a context-aware demo from the NLUX documentation, using sample portfolio data. View full size

My role

From the first API to ongoing maintenance.

I own the architecture, implementation, documentation, and releases. Maintaining the project also means answering questions, supporting integrations, and helping developers get from an example to a working product.

The library has reached 1.2K+ GitHub stars and 10K+ weekly npm downloads. It remains an active open-source project.