fix(consult): put locally computed techniques in the model packet
Independent Staging Quality Gate / validate (push) Successful in 11m10s
Independent Staging Quality Gate / publish (push) Successful in 12m30s

Web consult already computed Yoga, Arudha, KP, and related layers, then
stripped them before the spoken model. Keep those executed structures in
toModelOutput so answers cannot invent degrees from parametric knowledge.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
Jesse_Chen
2026-08-19 16:53:13 +08:00
co-authored by Cursor
parent 9efd847d2f
commit 0eedf92992
10 changed files with 418 additions and 9 deletions
+16
View File
@@ -4490,3 +4490,19 @@
- 相关记录:BUG-265、BUG-269、BUG-283
- 复发自:BUG-283(不再等模型,但仍把完整排盘当首页可用性)
- 修复版本:待提交
## BUG-300 | 网页咨询把已算技法从模型包裁掉,模型用参数知识填洞
- 状态:resolved
- 首次发现:2026-08-19
- 最近更新:2026-08-19
- 影响面:普通咨询 `toModelOutput``_attach_local_consultation_layers`、口头回答密度
- 用户现象:线上咨询缺 Yoga、Arudha/UL/A10、Chara、行运等本地 skill 会用的结构,模型用常识补度数和宫位。
- 触发条件:有出生分钟的个人盘咨询。计算工作流已挂上分盘谱、Narayana、功能性吉凶,但发给模型的证据包没有这些层。
- 根因:`toModelOutput` 的本命投影只保留升一/行星/宫位/Shadbala/AV/功能性吉凶/分盘谱。Yoga、Arudha、Gulika、Kakshya、KP、Chara、行运留在 `toAgentConsultationContext` 里却不进模型包。咨询路径也未把 Yoga/Chara/Sade Sati 写入 modules。
- 修复:咨询 attach 补 Yoga、路线相关 Chara、Sade Sati 行运。模型包投影这些层,并给出 `must_use_layers` 已执行短名单。未交付层禁止用常识补算。MEVG / VedAstro 仍诚实 `blocked`。不绑 AGENTS.md 或整份 SKILL。
- 验证:`frontend/tests/consultation-context.test.ts``consultation-spectrum-parity.test.ts``consultation-voice-contract.test.ts``tests/test_consultation_consumer_context.py`
- 防复发:咨询 `toModelOutput` 必须包含 skill 共同基础里本地已执行层;`local_layers` 不得再读引擎未写入的 module 键。
- 相关记录:BUG-279、BUG-287、BUG-298
- 复发自:BUG-279(算了却不交给模型)
- 修复版本:待提交