SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI codingenglish-compiler

english-compiler

English Compiler is an AI-based proof-of-concept compiler that can directly convert English Markdown specifications into runnable code. Through carefully designed prompt chain engineering, it breaks through the input and output length limits of existing OpenAI models and supports generating complete applications spanning multiple files or ultra-long Java classes. The project is positioned as a feasibility validation of AI as a programming abstraction layer in the future, rather than a production-ready tool. It requires configuring an OpenAI API key, and compilation is currently relatively slow, making it suitable for developers interested in AI code generation to explore.

Open website

About this tool

Purpose

English Compiler aims to directly compile Markdown functional specifications written in English (or other natural languages) into functional code. It attempts to validate an idea: AI can become a new abstraction compilation layer on top of existing modern programming languages, just as modern programming languages abstract operating system APIs.

Main capabilities

  • Convert Markdown-format .spec.md specification files into code files in the corresponding language.
  • Support generating complete cross-file applications, such as a Twitter clone without login authentication (including UI and API endpoints).
  • Support generating large Java classes exceeding 4000 tokens in length, which is difficult to do by directly calling the OpenAI API.
  • Bypass OpenAI model input and output length limits through complex prompt chain engineering to achieve large file generation.
  • Optionally generate JavaScript test scripts (avoiding contamination of source code through a separate process).
  • Provide a command-line tool that can build for a single specification file or an entire project.

Usage

  • Install via NPM: specific commands are not yet provided, but the official statement says the simplest installation method is to install from NPM.
  • Create a specs folder and a src source code folder in the project, and set up a english-compiler.jsonc configuration file, which needs to include settings such as the OpenAI API key. The official reminder is not to commit the API key to a public repository.
  • Specification files are placed in the specification folder, named $NAME.spec.md, corresponding one-to-one with individual code files.
  • Specification files contain at least Markdown frontmatter, in which the type parameter specifies the target language; the optional test parameter (only JavaScript is supported) will attempt to generate related test scripts.
  • Command-line examples:
    • Ask the AI for suggestions on a specification file:EnglishCompiler spec suggest spec/ui/index.spec.md
    • Build a single specification file:EnglishCompiler build file spec/core/db.spec.md
    • Build all specification files:EnglishCompiler build all
  • Provide demo projects:demo/twitter-clone(Twitter clone) and demo/java-class(Large Java class), the demo has pre-built caches and specification files for the AI compilation process, as well as pre-built code output.

Applicable scenarios

  • Experimentation and proof-of-concept by developers interested in AI-driven code generation paradigms.
  • Scenarios that require automatically generating code prototypes from detailed functional specifications.
  • Scenarios where you want to generate large code files that exceed the usual API limits (for example, large Java classes); this project provides an example of bypassing the limits.

Limitations

  • Explicitly a "proof of concept" (POC), far from production-ready.
  • The output code may contain a small number of errors; developers trying to fix these errors take a long time, and the write-compile-test loop is slow.
  • Low performance: generating a large code file requires multiple calls to the AI process (possibly more than 10 calls per file), and the overall process is very slow.
  • When ChatGPT is overloaded, compilation may fail.
  • Depends on the OpenAI API, requires providing an API key, and the key must not be made public.
  • The test generation feature only supports JavaScript.
  • It is officially described as inefficient, and the current iteration may have no practical commercial value.
  • The project already has a successor, smol-dev-js (https://github.com/PicoCreator/smol-dev-js），但本仓库仍作为独立 POC exists.

Related tools

Browse category