Deterministic upstream
Candidate generation, event backtesting, and rectification stay in auditable code. AI cannot alter the selected chart.
Evidence-first BaZi research prototype
Test uncertainty before interpretation.
Recorded birth time is evidence, not ground truth. This prototype keeps chart derivation and rectification deterministic, gives people control over context, and permits AI only after the upstream evidence is fixed.
先处理出生时间的不确定性，再进入解释：校盘由确定性代码完成，上下文由用户控制，AI 只在下游工作。
- 01Derivestructured chart data
- 02Rectifydated-event evidence
- 03Contextuser-controlled facts
- 04Forecastbounded downstream use
No login
No API key
No network request
Fixture-based demo
Interactive public walkthrough
Inspect the separation of evidence and context.
This page demonstrates the information contract and export flow with an illustrative fixture. It does not calculate a real chart or make a personal prediction.
02 · Output contract
Illustrative report
Ready to inspect or export.
Why the boundary matters
One pipeline, three explicit responsibilities.
User-controlled context
Context can be hidden from forecasts, excluded from exports, or deleted. It never leaks back into chart ranking.
Measured downstream
Offline A/B/C/D evaluation separates derivative-only, context-only, default-chart, and selected-chart modes.