The astrologer (2026-11-21) ruled that ordinary readings and forecasts never
use the user's experiences or feedback; only the rectification flow may.
Her reference files still told the model to back-calculate the user's known
events first, stop predicting below 3/5 matches, and grade A by such
verification. Until she edits them upstream, a product layer cuts those
passages out of every text sent to a model: the personal reading's quoted
references, the chat's methodology excerpts, and every chat prompt before
it reaches the provider (covering skill_read). The files are not edited;
a test names any rule that stops matching once she changes them.
Grade A no longer asks for a historical back-test. The status
user_history_verification_required is no longer a gap: no verification
questions, no penalty in ranking, never shown as 「需要过往经历核对」, not
offered to the report or synastry writers. 「未见阻断项」 is caught in the
reading and in report-question answers. Contract §AU announces the layer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
Rules sharing an English name printed each other's Chinese name in the reading
facts; each row now takes its engine row's name_cn by rule id. Neecha Bhanga rows
carry the fallen planet and the cancellation witnesses as two roles in the data
chapters' words. Narayana's direction readings are read from the report packet
when BUG-2105 stores them, and skipped otherwise.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
After BUG-2100 merged the supplementary timing chapters, one chapter holds up to
~505k characters and a question read only its first 20k. A chapter over 12k
characters is now read as its lead, a list of its ### parts by id
(chapter_<n>_<k>), then parts from the top; a part can be read by its id, and a
passage selected in a long chapter is read from the part that holds it. Cut
material ends with a mark the model is told about. The TOC timing pattern gains
年运 and Timing so 「年运附表」 and "Additional timing calculations" match their
other-language twins.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
The personal report reading wrote UL, 入庙, 小运, 婚姻煞 and the like in the
sections written for ordinary readers. Owner 2026-11-20: the body uses plain
words, technique names stay in 「依据与边界」 and the data chapters.
- Rules (product contract, report-reading block, notice G-9): the eight
sections take the chat's PLAIN_TERMS and time-naming rule; dignity is judged
from dignity_zh but told by its meaning; 「现在这几年」 gives dates and the
ruling planet, no dasha names, no Narayana; names only in 「依据与边界」.
- Check: technique_name, after every fact check, through the existing one
section repair; a writer's own plain gloss next to a name is kept instead.
A section held back for a name still counts toward the draft length, so the
draft stays usable and reaches the repair.
- Computed fallback text uses plain labels and drops the Narayana paragraph.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
At 1100 px with the sidebar open and the question panel closed the left
contents kept their 240 px column and the text was 461 wide. BUG-2093
folded them only while the side panel was open, and never on reports
that take no questions. The report view now measures its own layout box
(which keeps its width whether the column shows or not) and folds the
contents into their 「目录」 button whenever 240 + 48 + 520 do not fit,
on every report; the phone pill still takes over at 860 and below. The
reader keeps its place across the change; the place helpers move to
lib/personal-report-toc-layout.ts and are shared with the panel wiring.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
The writer was told to put 「只作参考」 into each section's basis and saw
the grade cap in its material, so stored annual reports showed
「这一块等级上限为 C」 and dozens of 「只作参考」 to readers. A shared
display pass (annualBasisText) now removes those notes on the page, in
the export and in what the question feature reads, keeping every date,
house, sign, number and technique name; stored reports are untouched.
New reports are written and stored without them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
The self-hosted data client accepts only plain column names in a select; the
JSON-path alias added for the personal report's subject name threw and the
route answered 500 for every personal, annual and synastry report. The name is
now read from the report document only when a personal report's conversation
is first created. A test scans every literal select in the app.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
A personal report generated from now on keeps the calculation facts its reading was written
from (one server-owned section row, bound to the reading's facts hash) and takes questions
beside the report, as the annual and pair reports do. Older reports show no entries.
- material: each reading section with the facts it is about, written as plain lines the answer
check can read; data chapters and the eight tables on demand; 20,000-character budget
- the annual turn is shared (runReportQuestionTurn); the consult route hands a bound
conversation to the personal or the annual handler by the report's own row
- page: entries beside the nine sections and each data chapter, selectable prose, the page
entry in the title bar, quotes that turn to their chapter; the reader keeps its place when
the panel opens or closes
- layout: contents on the left; with the panel open they fold into 「目录」 at the text's top
left when the window cannot hold contents, 520 of text and the panel
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
Staging retest round twenty-one: an annual answer copied 大运 / 小运 / 细运
and the writer's 「两套计时…说法一致」 note from the saved basis lines, and
another drew 「这副担子加得是实的」 from 「八分力三、总分三十」. Both now go to
the existing reword round; the instruction line says not to draw weight
from a number; whole note sentences with no date, number or house are left
out of what the model reads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
Pair follow-ups: 他 / 她 / 乙方 for one of the two goes to the one reword round with
both names, and a stored name with a space is written back as stored. Annual
questions: a birth-time limit or a minimum-line verdict the material read does not
make goes to the reword round, and an unsupported birth-time sentence never reaches
the reader; one instruction line says so. Asking about a selected passage makes its
section the current one on both report pages.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
An annual report's questions live in a consultation conversation bound to
the report by chat_sessions.report_id (one per report, guarded by a
trigger); no settlement function is redefined. The consult route hands
every turn of such a conversation to the annual handler, which reads only
the saved report, checks dates and placements against what it read, and
settles the fixed sentence and feedback-only turns free. The page has an
entry per written section, a page-level entry in the title bar, passage
selection and the shared panel or bottom sheet.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
The round-nineteen retest found a reassurance about separation with nothing
behind it, a Saturn aspect written as Saturn sitting on Mars, one partner
addressed as 你, a lead-in phrase at the end and the fixed sentence without
its full stop. The follow-up instructions state each rule, three wording
checks feed the existing single reword round, and the analysis lines show
names and a plain aspect instead of engine keys.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
The name step saved the whole answer (「叫我老板」) as the name. It now takes
the name out of the common phrasings, asks once more when there is none,
and writes the name on its own instead of the whole profile body.
A cold Home start read the account from the layout's boot record, which
kept the account as it was when the tab opened. A new user who finished
onboarding and then saved their profile on another page came back to the
name question, and answering it cleared the stored birth data. The boot
record now follows the account.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
Selected, current and checked states across the chat profile picker,
birth-time options, preset chips, theme and ayanamsa options, gender,
the synastry segment, ephemeris dates, the people archive and the chart
page now use --color-sidebar-selected. Selected text is ink. Hover on
borderless items is a lighter neutral. Rectification-only selectors are
left for their owner.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
Round eighteen on staging: the selected passage stuck to later questions, a
question the report did not compute was charged, answers read out technique
names and hedging lines, feedback counted as a question, the chat history row
needed a reload, and the advice section had no entry.
- pending passage is cleared once answered; nothing pending = whole report
- a reply that opens with 'not computed' becomes the fixed sentence, free
- follow-up instructions take the main chat's plain-terms and no-shedding rules
- counts skip free turns; the shell reloads its session list after an answer
- advice section and a page-level entry; selection bar placement; neutral fills
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
The owner approved removing the first group of lines from the BUG-2071
sweep: the line under each annual range card, the onboarding note about
filling in one's own data later, the ayanamsa scope sentence, and the
notes above the Bhava and transit tables. Their constants and CSS go
with them; the lines the owner did not approve stay.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
A synastry follow-up used to open a new conversation per section on the chat
page with a prefilled first question. Now each section heading has an entry,
the conversation opens in a panel beside the report (a sheet on phones) with
an empty composer, and there is one conversation per report. A passage
selected in the reading text can be asked about: the server keeps it only
when it is text of that section of that report, stores it on the question,
and names it to the model as the focus of the turn. Report rows and section
entries show how many questions were asked. Charging, refunds and the free
paths are unchanged; no database change.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
A report started and then left behind finished without a word: every page
polled only its own report and only while the tab was visible.
- GET /api/reports/status: id, status and type of up to ten of the caller's
reports, read as the signed-in user; no report body.
- The app shell watches reports the user started (browser storage, tolerant
of it missing), every 10 s and every 30 s in a background tab, and stops
when they are written, failed or gone. Pages that poll a report themselves
report what they see, so it is not asked about twice.
- Done or failed is a toast with 查看, the tab title in a background tab, and
a system notification when the browser allows it. The text names the kind
of report only.
- The browser's permission dialog is asked once, inside the click that starts
a report. There is no setting.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
- 上一章 / 下一章 now scroll to the new chapter's top through
scrollToReportAnchor, the path contents picks already use.
- The closed district select shows 「市区中心」 (the list keeps the full
wording) and place selects stay on one line with an ellipsis, so the
row keeps one height.
- The person form no longer shows 「称呼和出生资料。」 under its title or the
explanatory line under 性别; description is optional.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
Aspect table in degrees with special aspect names, Ayana angle no longer called declination,
annual planets read from planet_longitudes, both neecha bhanga rows kept, PD lord counted in
house activation, sub-periods name their parent period, Saham formulas with real operators,
D30 source input, Pushkara, Graha Pada, A10, Argala undecided, Moon chart degree, Lagna nakshatra,
gana, Shadbala units, place label in the English edition.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXp2Vp62Ur3C7yRWQdSq
- A data chapter picked in the report contents now scrolls to its heading
(it only swapped the chapter before); the chapter section carries the
same top margin as the reading parts.
- A credit ledger row that opens another page closes the settings dialog.
- The contents rail keeps its last rows clear of the floating button.
- A whole redeem code typed with spaces or without hyphens gets them back
in the browser; the server is unchanged.
- The birth-time block of both start dialogs sits on a neutral surface.
- The place picker's idle hint and the two chips on the report offer are gone.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8
Below 860px the reader's two-column rule out-ranked the single-column
rule, so once the contents became a floating button the body was left
with a 70px column on a 390px phone. The narrow block now resets the
reader rule by name.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DbBXoyn7Hj3LCW3BUs5bZ8