Starter-based capabilities
Add a dependency to assemble; remove it to unload. Business modules do not rebuild infrastructure.
High-quality AI programming is not about longer prompts. It is about putting engineering constraints into code. TG-boot gives AI and human developers one verifiable boundary: modules, contracts and architecture gates.
$ git clone https://gitee.com/pub_module/tg-boot.git
$ powershell -ExecutionPolicy Bypass -File scripts\up.ps1
✓ JDK 27 · Maven · MariaDB · Redis ready
✓ runner started on :9999
✓ middle platform on :9100
[ModuleRpc] [LOCAL] ApiSysUserService#getByUserCode
[MQ] system.user.login → parentled
_
The English edition focuses on the framework’s core rules and the multi-end delivery model. Deeper implementation notes remain in the full manual.
The modular monolith, the four layers, and why one JVM can still keep real module boundaries.
The hard constraints an AI should read before generating entities, APIs, controllers or MQ consumers.
How ArchUnit and POM rules stop a module from quietly depending on another module’s implementation.
Synchronous Api services and asynchronous MQ events: where they live and how to keep them stable.
A reusable uni-app platform based on an improved unibest template, built for H5, mini programs and apps.
Actuator exposure, MQ baselines and the checks that should happen before a production release.
An improved unibest-based platform. New modules write only their business pages; the admin, mini program and app share one login, permission model, layout and component layer.
You do not need to become a framework expert first. Each step has a clear input, output and acceptance check.
Add a dependency to assemble; remove it to unload. Business modules do not rebuild infrastructure.
Cross-module calls only use *-api contracts, DTOs and enums, never another module’s implementation.
Local provider first, MQ only when needed. The business code stays the same.
Forbidden dependencies become tests, enforced by CI rather than memory.
JDK, Maven, MariaDB, Redis and Node live under .tools and are prepared by one script.
Start as one process. Split by domain later; what changes is deployment, not the business code.
AI does not lack the ability to generate code. It lacks a clear boundary of what must not be written. A good framework makes the correct path the easiest path.