DEV Community
Follow
从 v1.0 到 v2.0：一个 Vibe Coding 管线的架构级反思
The author released a patch for the vibe-coding-universal project by correcting a version number typo in a comparison table. This oversight highlighted a significant problem: even the developer was unclear about the project's current version, making it difficult for users to understand its value. The original v1.0 aimed to help AIs generate code by first clarifying requirements through structured questions, outputting a single guide file. However, it lacked design specifications, had a single file for architecture and tasks, and relied solely on the AI's internal knowledge for output quality.
This limited v1.0's utility, leading to the development of v2.0. The upgrade wasn't just an addition of features but a fundamental shift in the process. V2.0 incorporates an additional seven rounds of clarifying design preferences alongside the initial seven rounds for requirements. It now produces a comprehensive DESIGN_SPEC.md document covering ten aspects, including colors, fonts, components, and layout.
Furthermore, v2.0 consolidates information into a multi-document Build Spec package, separating the Product Requirements Document, Design Specifications, Architecture, and Task planning. A key innovation is the integration of 71 brand design systems, offering a robust reference point. The structured Q&A in v2.0 is a significant improvement, utilizing single-choice questions to guide design choices.
This step-by-step approach, presenting users with limited options per round, allows the AI to precisely match design tokens from established brand systems. The output is a usable CSS variable table, directly consumable by the vibe coding tools. The author emphasizes that clear version labeling and a user-friendly comparison table are essential, especially when new versions offer drastically different capabilities from older ones. New users need to understand the project's value proposition within seconds, necessitating immediate clarity on which version to use.
