用「輸入 → 排盤 → 解讀敘事 → 生活建議」的同一條鏈，說明各術語在盤內的職位。可與 內容指南 併讀。非專業醫療、法律或投資建議；以 服務條款 為準。
We use one pipeline: input → chart object → reading → advice. Cross-check the content guide. Not professional advice; see Terms.
免責：以下為可教學的通識層，用於幫助理解傳統命理的常見方法論；不評價任何個人或師傅。與即時產品、條款衝突時，以官網後者為準。
Disclaimer: General education on methods — not a verdict on any practitioner. The live app and Terms prevail on conflicts.
傳統命理在方法論上，多數時候可以寫成：
Methodologically, most classical frameworks follow this chain:
出生脈絡 → 排盤（建立盤面物件／符號欄位） → 在固定體系內取元件、加權、疊代 → 敘事與建議birth inputs → chart object (fields/symbols) → select & weight components → narrative & advice
用規則把時間換成可核對的結構，例如：干支、宮位、星曜、十神、四化觸發等。若此層不成立，後面長文容易變成語言推論，而非盤內有據的敘事。
What can be field-backed on a chart object (pillars, palaces, stars, class labels, event markers). If this layer is missing, prose risks becoming ungrounded invention.
把符號「翻譯」成可理解的指引：流派強調、經驗、節點先後、口氣、甚至疲勞，都屬敘事層。盤上同一套符號，可對應多種合法但不同的講法。
How symbols become guidance: school emphasis, heuristics, ordering, phrasing, attention limits. Same symbol set can support multiple coherent readings.
以下各節，依「一句定義 → 在系統裡做什麼 → 常見讀者誤會」同一格式撰寫，便於併讀 內容指南 的產品邊界。
Each term below follows: one-liner → role in the system → a common mix-up — so it pairs cleanly with the content guide.
一句話：以出生年、月、日、時各一干一支，合計八個字，是傳統命理中極常見的「盤面骨架」之一。
One line: four pillar-pairs (Year–Hour) — eight characters as a common BaZi chart skeleton.
在系統裡的角色：把出生時間代入干支曆法，年柱、月柱、日柱、時柱層層疊上；由此再談五行強弱、十神、神煞、大運等，皆須在「四柱已可呈現」的前提下進行。它可理解為「先天節律與傾向的符號地圖」，不是單一「命運判決書」。
Role: the sexagenary mapping forms pillars; all further labels hang off that object. It is a structured symbol map, not a verdict in itself.
常見誤會：以為八字「算出來一個成語就能定終身」；實務上，怎麼讀、讀多深，仍屬敘事層與學派，需要與大運、流年、甚至與他術併參（例如與紫微交叉）時的盤庫定義一致。
Common mix-up: mistaking a label for a life sentence. Reading depth is still a model choice and must cohere with the chart library in use.
一句話：以「日主」那一字代表「自己」，其餘干支與之五行生剋，歸成約十類敘事符號，用於描寫人際、事象、取向。
One line: ten-ish classifier names for the day-master’s relations to other stems/branches (resource, peer, output, wealth, authority…), not a pantheon.
在系統裡的角色：例如常見寫法以正官、正財、食神等語彙，描述你與權威、金錢、表達的敘事原型；不是宗教意義的「神」。學派不同，歸類、是否採用某些別名、喜忌權重皆可不同。
Role: a vocabulary for how chart dynamics are narrated; schools disagree on weighting and naming details.
常見誤解：把某一十神當成「好／壞標籤」；在實務上，組合、旺衰、歲運引動才決定敘事走向，單一標籤不應被口號化。
Common mix-up: treating one label as intrinsically good or bad. Combinations and time activation matter; avoid slogan readings.
一句話：以星曜＋十二宮等結構，把人生多個領域拆成可逐一檢視的盤面模型。
One line: a star–palace model that slices life topics into checkable “rooms” on a chart.
在系統裡的角色：同樣從出生時間起盤，但產出的是「宮位＋星曜＋生年四化＋大限流年疊盤」的一類高維盤庫敘事。與八字併參很常見，但兩邊是不同術語表；不應在沒有盤庫定義下硬套一句結論。學理上可粗略比擬：八字常便於看整體節律，紫微常便於看十二宮內的細讀，但流派細節仍以所用盤庫與產品版本為準。
Role: palaces + stars + Si Hua + time overlays — a different symbol sheet from BaZi. Best compared only when each chart object is actually defined for the app.
常見誤會：以為「多星＝多準」；實則是組合爆炸：元件越多、正確的讀法越要求在固定規則下權衡與疊盤，臨場取捨不同，讀者就會聽到不同重點。
Common mix-up: "more stars = more accurate." Combinatorics grow fast; the skill is weighting, not name-dropping.
一句話：紫微盤內用化祿、化權、化科、化忌等標記，描寫帶動、得勢、名聲與摩擦等「動向」的常見工具之一。
One line: four transformation tags used in many Ziwei lineages to mark momentum and tension patterns.
在系統裡的角色：簡化理解（教材式）：化祿多指向緣分與好處之處；化權多指向掌控與壓力；化科多指向能見度與名聲；化忌多指向課題、執著與需多費心力的節點。飛星派、三合派對「看誰飛、先看誰」可不同；有忌不代表「壞完」，而是敘事上要著墨的難點位。
Role: dynamic markers. Schools differ on how to fly or prioritize them. "Ji" often marks a friction/lesson locus, not a moral failure.
常見誤會：把化忌當成「全盤大凶」；實務上，與所落宮、大限、流年、星曜同看，忌也可以是「專心修一題就能進步」的描點，而非詛咒。
Common mix-up: treating a transformation as globally “bad” without palace and timing context.
一句話：主星、輔星、小星是盤內的「敘事元件」；精要在組合＋宮位＋層疊，不是單一星定調。
One line: treat stars as cast + modifiers; meaning lives in mix + palace + time overlays, not single labels.
在系統裡的角色：主星定該宮基調、輔星加減戲、小星可補上細膩鉤子；同一顆主星在命宮與在夫妻宮，語意場不同。若解盤只背「星性口訣」而忽略同宮內有誰、誰生誰、誰衝誰、哪個大限在演，兩位師傅自然講出兩套故事。
Role: same star in different palaces and cohorts = different “scene,” not a fixed keyword. Good reads are relational.
常見誤解：以為「吉星入命就一生順遂」；實盤有廟旺落陷、空劫、化忌來扣，敘事要回到欄位與盤庫，而不是一句口號。
Common mix-up: lucky-star slogans. Brightness, voids, and harms in the same palace rewrite the line.
一句話：在盤面已成立之後，把人生時間軸疊在命盤上，分段看、逐年看的兩個常見層次。
One line: time overlays on an existing chart: decade-sized bands vs year-by-year passes.
在系統裡的角色：大運常見敘法是以約十年為一節（實作與起運法依派）；流年是年的干支與盤的互動。兩者若離開了可呈現的盤庫欄位，敘事容易變成自由聯想。細節以所用曆法、盤庫與產品版本為準。
Role: you need the base chart fields first, then you stack periods. Which “luck” algorithm applies depends on the library.
常見誤解：一聽到某年沖、某年吉就慌；沖、合、吉、凶都要放回盤內敘事與整體盤續，單一流年標籤不宜口號化。
Common mix-up: one annual keyword panic. Clashes and boons are chart-relative, not standalone headlines.
因此，先建立可校對的盤庫、再在可引用欄位內導讀，是產品上與「盤續不飄、敘事有據」對齊的方向；與 內容指南 的說法一致。若無盤、僅在聊天內讓泛用模型寫劇本，盤面不成立，讀者最難對照、亦較難多輪自洽，亦見內容指南之對照。
That is why a verifiable library first, bounded reading second matches the content guide. “Chat-invented charts” are a different class of problem.
八字與紫微併參在學理上常多一維符號可讀，目的不是兩邊各說各話，而是：在盤庫可呈現、欄位可核對的前提下，看相近主題與相同時間層的敘事能否對讀、分工或需節制。以下為產學上常討論的方法論（不逐條等於產品已內建；實作以即時產品與 服務條款 為準）。
Two symbol systems in parallel are useful when a verifiable object exists: you compare themes and time-slices, then narrate with restraint. What follows is common methodology talk, not a guarantee of shipped features. The live app and Terms govern.
我們的雙系統在學理上想拉高什麼？可以粗略比做：一張是長程節律的符號圖、一張是宮內的細讀圖，兩圖在同一盤庫能成立的前提下，再讓時間疊盤、主題、語氣盡量對齊、可續讀、可自洽——但仍不宣稱在科學實驗意義上「永遠不錯」；具體產品邊界見 內容指南 與 服務條款。
Think two different maps that must both lock to the same object first: you align themes and time-slices, then speak with graded certainty. This page does not claim infallibility. The content guide and Terms are authoritative for product language.
交叉參照是在盤庫內、可引用關聯的敘事，不是兩套自由聯想各寫一齣戲。產品上，仍以老師傅與內部盤庫定錨的符號層為準，專屬解讀在可引用範圍內導讀；不在此權威頁承諾已內建真太陽時、內部「信號分數」產品化、或依用戶回饋自動重訓等——若實作階段有，當以產品介面與條款敘事為準。
“Cross-ref” is narrative that stays in-field, not two unrelated stories. The product’s scope remains the chart library and bounded reading per the guide and Terms. We do not state here that geolocation true solar time, automated scoring, or feedback-driven retraining are shipped features—follow the app if they appear later.
以下不宣稱人必定弱於系統、也不宣稱系統在科學實驗意義上「萬試萬中」或「絕對客觀」。產品敘事是：在已結構化的命盤物件上，由內部盤庫欄位支撐一致性敘事，並由專屬解讀層在可引用範圍內導讀、追問，越界可拒答（服務條款）。命盤所依之傳統體系與欄位，由具實務經驗之老師傅與盤庫邏輯定義；解讀層不取代盤面權威來源，實作以即時產品為準。
We do not claim machines beat humans, nor do we claim experimental omniscience or “perfect objectivity.” The stance is: a structured object first, a bounded reader second; refuse out-of-scope asks per the Terms. Masters anchor chart rules; the reader does not replace that authority. Live app prevails.
| 你剛學完的上文變因 | Issue from above | 產品可對照敘事（與內容指南一致） | Product stance (see guide) | 讀者較可感知的差異 | User-visible |
|---|---|---|---|---|---|
| 學派權重不同、口訣庫各異 | School weights differ | 內部盤庫依產品版本、既定規則與欄位產生一致結構，解讀在該欄位內導讀、可拒答盤外發明。非「併吞天下所有流派真理」的宣稱。 | One in-app rule set per version: structured fields, bounded narrative — not a claim to all schools at once. | 多輪追問依同一套盤庫參照，較不會前後自打臉；仍以即時產品為準。 | Same object for follow-up Q&A; app truth wins. |
| 元件組合大、人為一輪難盡掃 | Large combinatorial field space | 產品以可程式化、欄位化的盤庫為底，解讀層在盤內有據的節點上導讀。此處是降低遺漏與飄移的工程取向，不是「人機優劣」的人生判斷。 | Engineering stance: field-backed base + guided traversal — an ops lens, not a value judgement on people. | 在線、可隨時回到盤內欄位核對，而非單次口述記憶。 | Re-open the same field-backed object anytime. |
| 敘事主觀、節點取捨不同 | Narrative & emphasis vary | 專屬解讀在可引用欄位內產生自然語言層，措辭可隨版調整，但應受盤庫約束。需醫療、法律、投資承諾者不在服務內，見條款。 | NLP in scope; copy may revolve, fields should not. No medical/legal/financial promises — Terms. | 敘事力求與盤內敘事可對照、可續聊，而非盤外長篇聯想。 | Tries to keep chat aligned to the chart object, not ungrounded essays. |
產學界對「生成式模型在無約束時易產生似真但難以核對的細節」有廣泛討論。若僅從技術風險角度參考（不直接等同任何命理斷驗正確性），可閱讀 Stanford《AI Index》彙整 與 NIST AI Risk Management Framework。Ask Sifu 的設計是盤庫有欄位、解讀有邊界；外連屬產學通論參考，不構成任何命理實效承諾。詳見 服務條款；實作以即時產品為準。
Ungrounded generative text is a known risk pattern; see the Stanford AI Index and NIST AI RMF for general context — not a claim of destiny accuracy. Field-backed library + bounded reader; the live app and Terms govern.
先以傳統八字、紫微之規則產生可核對的盤庫，再由專案演練之專屬解讀層在可引用之範圍內導讀、追問。相對地，不先產生可校對盤面、只讓泛用大模型在聊天內即席寫劇本，盤面不成立，敘事易飄。完整產品邊界、免責與實作細節，以 內容指南 及 服務條款 為準。
Rule-based chart structure first, tuned reader in scope second — the opposite of ungrounded chat-only invention. Authoritative boundaries: the content guide and Terms.