EVIDENCE / METHOD
How to read a Semantix result.
Evidence labels describe how a result was produced. They do not turn a passing test into a production guarantee.
Evidence levels
Design or editorial context
A design note, roadmap item, or explanation without a reproduced output.
Repository tests
A first-party unit or integration test with a reproducible command and visible result.
Synthetic end-to-end run
A fixed fixture or labeled query set exercised from input through output.
Independent or real-session reproduction
A reproduction by an external developer or on a real session corpus with method and raw results.
Current boundary
The current public run is E1 · repository tests. It supports claims about tested repository behavior. Real-session relevance, universal framework compatibility, and lower production cost remain unverified until a fixed corpus and independent reproduction are published.
Publishing rules
每个结果都应同时公开输入、命令、环境、原始输出、解释和限制。没有这些字段时，页面只显示 E0，而不会给出精确分数。
当前站点公开的 E2 资料是一个小型仓库 fixture：15 rows · 10 train / 5 replay · single reuse 100.0% with 20.0% error · three-region reuse 60.0% with 0.0% error · verdict PASS · grey ratio 40.0% (above the 30.0% alarm)。它用于说明评估流程，不代表真实会话相关率或生产成本下降。