Product decision 2026-10-02 (TASK-upstream-sync5 R1): a time range is offered
only with at least 4 dated, primary-scoreable events covering 3 domains,
counted on all of them (training + reserved holdout). Was 3 training events /
2 domains in three TS copies and the Python acceptance gate while the policy
file already said 4/3.
- One definition: references/rectification_policy.v1.json
(minConfirmationEvents / minConfirmationDomains). TS core/types MIN_DATED_*,
rectification-decision MIN_STANDALONE_*, evidence-model MIN_ACCEPTANCE_*,
the convergence evaluator and the post-inference trainingGateOpen all read
it; Python decision_policy MIN_ACCEPTANCE_* alias MIN_CONFIRMATION_*.
- Python receipt counts all scoreable events / domains for event_quality and
domain_diversity; decision policy identity v3 -> v4 (candidate UUIDs carry
it). Candidate scores unchanged (77 v5 cases A/B identical), so the
algorithm stays rectification-v5-matrix-scoring-10.
- Memoization golden v3 written by write_golden; v2 frozen by sha256 with a
test that its scores equal v3 and only the receipt policy moved.
- Collect gap copy names the exact gap ("再来两件……其中至少一件不是……")
instead of always "再来一件"; VOICE.md updated. Legacy life-events form copy
4/3 as well.
- 30 frontend test files, 4 Python tests: fixtures extended to the same
scenario at 4/3, or assertions changed with 原值/新值/原因 notes.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
- Evidence-card and biography-backtest goldens regenerated with the capture
scripts (PYTHONHASHSEED=0, cache TTL 0; byte-identical twice); only the
functional-role fields change.
- CAREER_FIELD_ASK_RULE: A10 / AL / Sun / 10th lord in 4/6/8/12 say
resistance or cost only, never "not on stage / go deep", and houses are not
translated into fields.
- AFFLICTION_RANGE_RULE: no portrait of an afflicted person from the lord's
house or a benefic occupant ("managing / arranging / cares about your
future"), actions do not presuppose reaching them, one ask per answer.
- Examples that paired "4th lord in 10th" with "her attention is on your
future / close" now state the line is unafflicted (Monroe's chart had that
exact placement and copied it). Retrograde line also bans "drags on".
- Backtest script: several follow-up domains in one run.
Changed assertions carry 原值 / 新值 / 原因 in the test comments.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
- AFFLICTION_RANGE_RULE (one definition, consultation-thinking-plan.ts):
pressure / distance signs + evidence + a range of realities, no verdict on
facts the user knows better, no opposite reassurance, one ask per answer;
quoted by the shared reading (line 3) and NATAL_SECTION_RULE
- parents line 2, marriage line, health line 1 follow it; career gets a
'lifetime career shape first' line and neutral three layers (no 面试/职位)
- follow-up correction rule: map only lines that fit; say 盘本身分不出这一点
- voice / VOICE.md parents example rewritten to the range form
- contract tests updated with 原值/新值/原因
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
- natalAnswerShapeBody: NATAL_SECTION_RULE replaces 「这个人是什么样、你们怎么相处、
事情会怎么走」; the shared voice quotes the same constant (BUG-1164)
- product-voice / VOICE.md: no 逆行 in examples, parents example leads with the
affliction and asks to confirm, married-presupposing Good example removed (BUG-1165)
- follow-up shape: a user correction is acknowledged, never defended (BUG-1166)
- career examples carry no job shape (升职 / 专业能力 / 做深现有工作) (BUG-1168)
- biography backtest golden regenerated with the real engine so the card carries
brief 1's aspects, combustion, yoga domains and saturn_from_moon_house;
backtest script gains --correction follow-up turns
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
- T1 (BUG-1160): 「通用读法 · 先看受冲」 opens every turn's methodology once.
- T2 (BUG-1161): Chinese lists for parents, children, education, migration,
family, annual, general, health; health replaces the English route in the
web consultation only.
- T3 (BUG-1162): career / marriage / wealth each gain an affliction line.
- T4 (BUG-1163): health label 健康, prompt asks about the constitution first.
- Single-domain annual / timing turns keep the pre-brief method
(CONSULT_READING_BUDGET_BLOCKED): they would exceed the 12,000-character
assertions, which this brief may not change.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
Late night no longer remarks on being awake (产品 2026-10-01); every daypart,
weekday and returning line reworded in the same caring voice. Daily line:
each house sentence names where the Moon is in plain words, then one concrete
sentence; no stacked hedges. Sign-fallback moods rewritten the same way.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
Review-only snapshot for BUG-1115 through BUG-1117; not merge-ready. New opening and append-turn PostgreSQL permission failures remain blocked. Persisted joint replay has zero completed questions; segment ordering remains off by default. The existing offline replay JSON is retained stale and unchanged after a denied overwrite, including its CRLF line endings. Browser/provider validation and final serial gates remain pending. No deployment, role permission changes, or staging/main push.
Co-Authored-By: Claude Code <noreply@anthropic.com>
A 👎 now saves, server side, the rated turn plus the context window the
model read for it (reconstructed from the stored session with the same
consultationHistoryWindow the consult route uses), model and run facts.
👍 is only counted. Switching to 👍 or clearing deletes the snapshot.
Bodies are blanked after 90 days; the row cascades on session delete
and on account deletion.
- Optional 不满意原因 panel under the answer after a 👎 (five reasons,
200-char note, "会把这一轮对话发给我们排查").
- Admin 对话质量记录: 👍/👎 stats by day and model, list without text,
audited snapshot open, 处理状态 + note (support.quality.read/write).
- Privacy draft: what a 👎 keeps, why, 90 days, deletion.
- Migration 20260930050000 is add-only; set_reply_rating() replaced with
the same signature.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
Request signs out everywhere and freezes paid routes (423 + DB triggers);
signing in within 7 days shows the pending gate with 撤销注销. A periodic
idempotent worker purges personal content afterwards and keeps finance
rows against a tombstoned identity. Read-only admin list.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
Hide engine branding, gate supported divisional columns, align house cells, and add accessible symbol help. Reduce desktop chart width and stabilize the waiting slot.
Validated with 78 browser checks and 87 focused frontend tests. Full suite retains 29 identical baseline loader failures; no new failures.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Once the discriminator training gate is open, only choice cards are asked
and the range card goes out when they are exhausted; targeted lines, their
re-ask and guided windows no longer hold the card or invite more events.
Delivery body says how many choice questions were used instead of the event
fit percent; narration names an excluded cluster instead of "range
unchanged"; a delivered turn no longer carries a collect question.
Offline replay (v4, 3 radii x 2 directions): truth in range 20/20 in every
cell; guided-window injections give the same width in truth and opposite
directions, so red line 1 was revised by product to truth-in-range only.
Skill 10.0.31 -> 10.0.32 (10.0.31 kept as deprecated for pinned cases).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
The owner had no edit entry anywhere once /people replaced the settings
chart pane. 本人 now gets 编辑 with the same form; the save goes through
PATCH /api/account via a shared lib/self-profile-save.ts that the home
onboarding also uses. The dead saveProfile is removed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
TASK-birth-sky-polish-20260928.
- P1: chart page dialog shows only the picture on a dimmed stage, 「保存图片」 below, icon × in the corner.
- P2: BFF returns one server-built `caption` (「1994年5月18日 07:40 · 杭州」); printed under the fact line in all three places; adopted time adds 「 · 校正后」.
- P3: convergence 3 s + words 2.5 s + fade 1 s; settled frame still equals the exported cover.
- P4: redeem paywall no longer opens when the starter home appears; first send refused for credits opens it, later ones open billing; convergence dialog portalled to <body>. BUG-1080 investigating (Chrome repro rules out the overlay covering the convergence).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
TASK-birth-sky-followup-20260928 F4:
- pure timeline / seeded scatter / interpolation in lib/birth-sky/converge.ts;
drawBirthSky takes an optional frame (positions interpolated, alphas
multiplied) so the settled frame is the exported cover call for call
- lazy full-screen <dialog> overlay driven by requestAnimationFrame at
viewport x DPR; tap / click / Esc end it; reduced motion only fades
- data, serif and overlay chunk must all arrive inside the 3 s budget,
otherwise skipped silently; page.tsx wiring unchanged
- one sonner hint after it ends; dialog reveal variant (继续, is-reveal) removed
- DESIGN §17, VOICE, CHANGELOG, checklist, PROGRESS with verification
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
T1 scripts/birth_sky.py: real alt/az of the seven visible grahas (swe.calc_ut +
swe.azalt, geometric altitude, azimuth converted from swe's south origin to
north-east) and Yale BSC5 stars V<=4.0 (IAU 1976 precession); phase
day / twilight_morning / twilight_evening / night. Thin POST /api/birth_sky
registration in do_POST (no new handler method, no __new__).
T2 GET /api/birth-sky?subject=: birth truth read server side through
loadSubjectBirth + prepareChartViewProfile; zod contract; no billing, no
writes, no birth data in the response.
T3/T4 lib/birth-sky: sentence table, pure Canvas 2D drawBirthSky (night /
paper palettes from DESIGN tokens, east on the left, serif headline with
1.5 s font wait), PNG export via navigator.share or Blob + a.download.
Chart page prefetches and paints the cover once the chart is ok; the header
button appears only when the PNG is ready. Dialog is a lazy chunk.
T7 DESIGN §17, VOICE, CONTEXT glossary, CHANGELOG, real-device checklist.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0199rbQDTsUbCVw84wc8BTFe
T6: offer True Citra and Fagan-Bradley beside Raman/Lahiri/KP/True Pushya,
engine-supported values only; widen profiles_ayanamsa_check by an additive
migration; ephemeris note names; path tests so no new value falls back to
Raman.
T7: the unlayered `button { color: inherit; }` beats Tailwind's
text-primary-foreground, so solid report buttons showed page ink (light
(light 4.85:1, dark 5.77:1), grey disabled without fade, neutral selected
preset. button.tsx and theme tokens unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
One ProfileGenderField (女 / 男 / 不填, 44px, one hint line) in the
people-archive edit form (saved with the form, 不填 → null) and in
设置 → 个人资料 for the owner (saves on its own PATCH; birth-data saves never
send gender). The detail card shows 性别 (未填 when empty). Not asked in
onboarding. DESIGN/VOICE updated; 390/1280 screenshots with fictional data.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
TASK-consult-evidence-card-v2-20260927 T5. PROGRESS with the native-module
table, the final per-card table, before / after sizes (3 public AA charts x
12 questions, all single-domain turns <= 12,000), red-line tests, failure-name
comparison and three-column edits; real-device checklist; CHANGELOG; task
index row set to 待验收; VOICE / DESIGN name the new lookup labels (no UI
change); BUG-1060 investigating (Double Transit D9 house target reads the D9
lagna; not fixed here, red line 1).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
T2 of TASK-rectification-grounding-20260927 (product decision P1).
The rectification 「重试（重新生成）」 was a blind agent.generate rewrite that
persisted unchecked text (invented range / fit rate reproduced). Removed:
the regenerate route, regenerate-turn.ts, the regeneration agent and its
read-only tool set, the client regenerate action/state and the
regenerating/canRegenerate props. ChatMessageActions renders the regenerate
button only when onRegenerate is passed; ordinary consultation is unchanged.
The DB function regenerate_agentic_rectification_turn is kept (AGENTS §7.6,
retire in a later round). DESIGN.md / VOICE.md updated; source-contract
tests follow with 原值/新值/原因 notes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
read-consultation-evidence returns one closed-enum section (a formal varga,
research/extended vargas, a Western layer, yogas, Ashtakavarga, Shadbala,
transits, Chara Dasha, arudha, karakas, KP, gulika, kakshya, mahadashas,
thematic evidence) from this request's finished calculation, never
recalculates, answers unavailable on a cache miss and refuses a second call.
The receipt records the step and the write row shows 「正在多看一眼：…」.
A lookup after answer text went out keeps the released text whole: a verbatim
restart is dropped as it arrives (40-char confirmation), a continuation is
kept, and settlement still reads the step that wrote the answer; the lookup
runs on the answer clock without resetting it. A length continuation carries
the lookup result with the card. BUG-1059: the visible-text transformer's open
clause is settled at each tool call, so unpunctuated narration no longer
leaks into the answer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
- BUG-1049 (recurrence of BUG-504): the opening stem carries six examples and
an example answer again and is server-owned on the zero-evidence opening;
the body is two plain sentences (no 大运/盘面/代表分钟/精确到秒, no year, no
question). Stem de-dup compares whole sentences / near-equality instead of a
12-char prefix, which had deleted the body's examples sentence since
aa7ccb30 (BUG-604) + dd8f35f7 (BUG-648).
- BUG-1050: plain step labels; a finished step label shows once and
「已完成 N 步」counts shown rows; failed rows read 「…未完成」 from the
in-progress wording.
- Skill 10.0.30 -> 10.0.31 (OpeningPolicy); 10.0.30 kept as deprecated.
- VOICE / DESIGN / CHANGELOG / BUG_HISTORY / PROGRESS / real-device checklist
and screenshots.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
- D2: each intent-classifier attempt is capped at 10 s (same session model,
thinking untouched); a hang takes the existing retry -> classifier_unavailable
path. Success is timed too (RectificationClassifierDiagnostic).
- D3: a typed message builds the NDJSON stream first; the first line is
turn.progress "received", then the classifier and deterministic replies run
inside the stream. Preflight rejections become turn.rejected (old status,
code, message) and the client handles them like the old HTTP rejection.
Stage lines 收到，正在对照你的档案… / 正在记下这件事… / 正在重新对照盘面… /
正在准备下一个问题… are driven by existing tool events and engine calls, are
transient (live row only) and never persisted. VOICE / DESIGN updated.
- D4: RectificationRunDiagnostic records per-step start/end, provider token
usage incl. reasoning tokens, classifier timing and per-engine-call
durations (AsyncLocalStorage scope per turn); RectificationTurnDiagnostic
for deterministic turns. No user text, birth data or model text.
- D5: /v5/versions memo (30 s, complete identities only, per transport); the
exit gate skips its second persistNextInterviewIfIdle when the run's own
call found the next focus already active (provably identical).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
BUG-1045 (recurrence of BUG-585 via BUG-969): the snapshot merge now drops
the attached stem from the streamed "ack + stem" text with the same
stripQuestionSentences the GET route uses; server text unchanged.
BUG-1046: a failed choice submit (409 / network) or typed send withdraws
the local answered mark, remounts the card, re-reads the Case and shows
"这次没提交上，请再点一次。"; the persisted question hangs on the latest
settled assistant message with other copies of the same focus removed
(standalone block only when nothing can carry it); willContinue and the
send() settle merge unseen assistant turns (same gap as BUG-685).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
- D2 resumable hint becomes「上次那次校正还没完成，可以在历史对话里接着做。」
so it does not imply the homepage pill resumes (it always creates).
- The click-time non-self error in useRectificationSurface is the same
user-visible sentence as the homepage hint; add the full stop.
- Source, test, VOICE, DESIGN, CHANGELOG, BUG-1041, testing list and
PROGRESS (amendment by Claude) updated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
