WeChat Mini Programs

The starting point

WeChat is the main gateway to digital services in China: mini programs open directly inside it, with nothing to install. This know-how was put into practice on a bilingual mini program built around a conversational AI assistant.

The problem

To reach users in China, a Western web or mobile app is not enough: you need a WeChat Mini Program, with its own interface languages (WXML, WXSS), its cloud backend, its development tools and its publishing rules. Testing and evolving a product in this closed environment calls for a dedicated method.

The answer

The mini program is written in native WeChat code, bilingual in Chinese and English, and a test chain runs it automatically in the official simulator after every change. A bilingual HTML prototype serves as the visual specification before development, and AI-generated content is governed by automated checks rather than by instructions to the model alone.

• Native mini program: WeChat JavaScript, WXML and WXSS, 10 pages, a custom tab bar and custom components
• Bilingual Chinese / English interface, 208 translation keys kept in parity
• End-to-end journeys automated in the WeChat simulator (miniprogram-automator), 100 checks passing
• WeChat CloudBase cloud functions prepared on the server side
• Conversational AI assistant: content labelled “AI-generated” and verified automatically
• A bilingual 12-screen HTML prototype, generated by script, used as the visual specification
• Planned: a Chinese language model, WeChat Pay, publication on the platform

Under the hood

The mini program is written in native WeChat JavaScript (WXML, WXSS): ten pages, a custom tab bar and reusable components. Data follows an append-only model, and displayed calculations are produced by deterministic code rather than by the language model. WeChat CloudBase cloud functions are ready for the server side. Quality relies on scripts: structure checks, translation parity and end-to-end journeys driven inside the WeChat simulator.

The notable challenge

A mini program only runs inside the WeChat environment, which makes automated testing unusual. The solution was to drive the official simulator by script to replay complete journeys in both languages, complemented by structure and translation-parity checks, re-run after every change.

My role

Market research, bilingual product design, then development of the mini program and its test chain, with AI assistance (Claude Code and review agents).

Where the project stands

The mini program compiles and runs in the WeChat simulator; the latest automated run, on 15 September 2026, passes 100 checks with no failures in both languages. It still runs on demo data: the language model, the real backend and on-device tests remain to be connected.

• Connect a Chinese language model
• Test on iOS and Android phones
• Prepare publication on the WeChat platform

A few figures

The project’s tools

Mini program

WeChat Mini Program (native JavaScript, WXML, WXSS)

Backend

WeChat CloudBase (cloud functions, database)

Testing and quality

miniprogram-automator, Node.js check scripts

Prototype

Static HTML generated with Python, Chinese / English toggle

AI

Planned: a Chinese language model (Qwen-Max class)

Tooling

WeChat DevTools