Core Philosophy
CosmicTao separates calculation from interpretation. The calculation layer handles things a computer can check: calendar boundaries, solar terms, true solar time, hexagram line states, zodiac pairings, point balances, and report access. The interpretation layer explains traditional symbols in modern language.
That split matters because each layer has a different job. A chart should calculate consistently first; the reading then explains the terms, relationships, and source logic that sit on top of it.
1. Calculation Layer
Where a rule can be computed, we keep it explicit. The main examples are:
Bazi calendar boundaries
The Four Pillars engine uses Jieqi solar-term boundaries and birth-time inputs instead of only Western calendar dates. When location is available, true solar time can adjust the hour branch near edge cases.
I Ching coin casting
The coin oracle uses the Web Crypto API for digital tosses, then maps each toss to old yin, young yang, young yin, or old yang. Changing lines are carried into the relating hexagram.
Compatibility scoring
Relationship tools combine zodiac pair rules, Bazi element balance, and report-specific scoring constants. The score is a comparison model with visible dimensions.
Points and paid access
Point balances, Plus benefits, paid report access, and refunds are checked server side. The frontend only displays the current result of those checks.
2. Interpretation Layer
Interpretations are generated from structured inputs, source excerpts, and fixed calculation results. AI helps restate and connect those materials, but the page should still show which tradition or rule it is using. Core references include:
Foundational hexagram judgments, line texts, and later commentarial language used in oracle readings.
Wan Min Ying's Ming dynasty encyclopedia, used as a reference for Ten Gods, clashes, combinations, and penalties.
A major Bazi text consulted for Day Master strength, elemental flow, and Yong Shen discussion.
When a result mixes sources, the page should avoid pretending that every school agrees. If a rule is a modern comparison layer, we label it as such.
3. Editorial & AI Standards
We use Large Language Models as translators, summarizers, and style helpers. The model should explain the provided chart, hexagram, or source excerpt rather than invent new facts about the user.
- Input scope: the model receives specific hexagram or pillar data, plus relevant source excerpts.
- No cold reading: reports should avoid personal flattery, fear pressure, and claims about private details the user did not provide.
- Plain-language editing: a useful report should explain terms such as Day Master, changing line, Useful God, clash, harmony, and void without flattening them into generic advice.
4. Quality Checks
We test the parts that can be tested. Calendar and chart logic is checked against saved fixtures, solar-term transitions, and edge cases around midnight or time zones. Report prompts are checked for structure, terminology consistency, and refund behavior when generation fails.
The goal is simple: the software should process inputs consistently, keep terminology stable, and make failure states recoverable.
5. Product Responsibility
CosmicTao keeps account logic, point spending, paid access, and refund handling on the server. User-facing pages should also state which calculation rule or source family they are using, especially when a result combines Bazi, I Ching, Liuyao, compatibility scoring, or wish features.