How to Build a Qwen3 Chat UI with Chainlit: Hooks, Auth, and Persistence

Introduction

This guide shows how to build a simple web UI for the Qwen3 large language model using the Chainlit framework, a Python library designed specifically for AI chat applications. The goal is to avoid writing custom front‑end code while still keeping the inference logic in Python.

Project source: https://github.com/jxd134/qwen3-local-chat

What is Chainlit?

Chainlit provides a ready‑made, responsive chat UI with Markdown rendering, code highlighting, and image support. It abstracts away front‑end development and offers built‑in features such as streaming token output, data persistence, user authentication, custom step visualization, and lifecycle hooks.

Core Features

Built‑in chat UI : beautiful, responsive interface.

Streaming output : native support for LLM token‑by‑token streaming.

Data persistence layer : pluggable storage (PostgreSQL, SQLite, etc.) for conversation history.

User authentication : password, OAuth, JWT, multi‑user management.

Custom step display : break complex processing into visible steps.

Hook mechanism : lifecycle callbacks for extending behavior.

Hook Mechanism Overview

Chainlit defines hooks as decorator‑based callbacks that run at specific events (application start, user login, chat start, message receipt, chat resume). The following sections implement each hook needed for a functional Qwen3 chat service.

1. Data Layer Hook (@cl.data_layer)

This hook registers a SQLite‑backed data layer using SQLAlchemyDataLayer. The implementation uses aiosqlite for async access and stores the database at .qwen3_chainlit.db in the user's home directory.

@cl.data_layer