PromptAppGPT: A Low‑Code Framework for Rapid GPT‑Based Application Development

PromptAppGPT, developed by CCF Theory of Computer Science Committee member Zhang Changwang, is a low‑code, prompt‑driven framework that enables rapid application development using GPT for natural‑language‑based coding.

The framework provides features such as low‑code prompt application development, GPT‑3/4 text generation, DALL‑E image generation, an online code editor with compiler and runtime, automatic UI generation, bilingual (English/Chinese) interfaces, and plugin extensibility.

It supports multi‑task conditional triggering, result validation, and failure retry, turning multi‑step manual tasks into fully automated processes, and eliminates the need for users to memorize complex prompts.

Quick Start

1. Obtain an OpenAI API key. 2. Visit the PromptAppGPT website (https://promptappgpt.wangzhishi.net/) and launch the web version or download the program. 3. Configure the OpenAI key, optional proxy, and select the GPT model (gpt‑4 or gpt‑3.5‑turbo). 4. Choose and run an application, then edit and compile it using the built‑in editor.

Example Applications

1. Travel Planning Master : Generates a detailed multi‑day travel itinerary based on user‑provided destination, duration, and budget, automatically producing a bilingual plan.

2. Intelligent Image Generator : Takes a simple description, expands it with GPT‑generated prompts, and creates detailed images via DALL‑E without requiring users to write complex prompts.

3. Multilingual Translator : Translates Chinese input into English using GPT, demonstrating the framework’s language‑processing capabilities.

Application Development

PromptAppGPT uses YAML for low‑code definitions. Below is a sample application definition:

---
author: Leo
name: 旅游规划大师
description: 旅游规划大师
gptRound: single
failedRetries: 2

sysTask:

userTask:
  - executor: gpt
    prompt: |
      我想去 [$i{旅游目的地:@input}] 玩，请你以专业导游的身份，帮我做一份为期 [$i{旅游天数:@select#1/2/3/4/5/6/7}] 天的旅游攻略。
      另外，我希望整个流程不用太紧凑，我更偏向于安静的地方，可以简单的游玩逛逛。
      我的预算在 [$i{旅游预算(元):@select#1000/2000/3000/4000/5000/6000/7000}] 元左右。

extra:

The author field specifies the developer, name and description describe the app, gptRound selects single‑ or multi‑turn dialogue, and failedRetries sets retry attempts. userTask defines the user‑driven prompt and executor.

Author

Zhang Changwang is a senior researcher and CCF committee member, former SPC member of AAAI‑23, with a PhD from UCL, experience at Alibaba (LBS data mining) and Tencent (advertising recommendation), focusing on information retrieval, NLP, and big‑data mining.