@furongjun1999/dsh-memory 0.9.22 → 0.10.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (179) hide show
  1. package/CHANGELOG.en.md +450 -0
  2. package/CHANGELOG.md +112 -0
  3. package/README.en.md +244 -0
  4. package/README.md +272 -260
  5. package/codebuddy/CODEBUDDY.md +1 -1
  6. package/docs/README.md +2 -0
  7. package/docs/discipline/templates/full.md.tmpl +1 -1
  8. package/docs/discipline/templates/rules.mdc.tmpl +1 -1
  9. package/docs/en/README.md +190 -0
  10. package/docs/en/TRANSLATION.md +78 -0
  11. package/docs/en/TRANSLATION_MANIFEST.json +170 -0
  12. package/docs/en/TRANSLATION_REPORT.md +11 -0
  13. package/docs/en/dev/README.md +10 -0
  14. package/docs/en/dev/coding-testing.md +72 -0
  15. package/docs/en/dev/switch-matrix.md +10 -0
  16. package/docs/en/glossary.md +97 -0
  17. package/docs/en/harnesses/claude.md +100 -0
  18. package/docs/en/harnesses/codebuddy.md +93 -0
  19. package/docs/en/harnesses/codex.md +88 -0
  20. package/docs/en/harnesses/dsh.md +84 -0
  21. package/docs/en/harnesses/zcode.md +42 -0
  22. package/docs/en/mdcg/ccg.md +32 -0
  23. package/docs/en/mdcg/generated-docs.md +20 -0
  24. package/docs/en/mdcg/guardrail-charter.md +170 -0
  25. package/docs/en/mdcg/mcp-tool-names.md +43 -0
  26. package/docs/en/mdcg/multi-harness.md +43 -0
  27. package/docs/en/mdcg/recall-metadata.md +30 -0
  28. package/docs/en/mdcg/world-model-endpoint.md +379 -0
  29. package/docs/en/mdcg/write-aliases.md +130 -0
  30. package/docs/en/ops/README.md +11 -0
  31. package/docs/en/ops/environment-known-issues.md +244 -0
  32. package/docs/en/ops/install.md +72 -0
  33. package/docs/en/ops/operations-troubleshooting.md +86 -0
  34. package/docs/en/ops/permissions-review.md +79 -0
  35. package/docs/en/ops/release-governance.md +49 -0
  36. package/docs/en/ops/tool-name-migration.md +30 -0
  37. package/docs/en/ops/ui-install-troubleshooting.md +53 -0
  38. package/docs/en/standards/ai-coding-standard-checklist.md +71 -0
  39. package/docs/en/standards/ai-coding-standard-lingshu.md +47 -0
  40. package/docs/en/standards/ai-coding-standard.md +185 -0
  41. package/docs/en/ui/minimal.md +62 -0
  42. package/docs/mdcg/cogmap_ops.md +16 -15
  43. package/lib/client.js +29 -3
  44. package/lib/hooks.d.ts +34 -0
  45. package/lib/hooks.js +167 -33
  46. package/lib/index.js +19 -3
  47. package/lib/lib/console_web.mjs +30 -3
  48. package/lib/lib/hook_audit.d.ts +3 -2
  49. package/lib/lib/hook_audit.js +14 -4
  50. package/lib/lib/mdcg_client.d.ts +5 -2
  51. package/lib/lib/mdcg_client.js +8 -4
  52. package/lib/lib/prompt_safety.d.ts +24 -0
  53. package/lib/lib/prompt_safety.js +77 -1
  54. package/lib/lib/token_store.js +5 -2
  55. package/lib/tools.js +5 -3
  56. package/md_cg/bucketindex.py +231 -0
  57. package/md_cg/bucketscope.py +339 -0
  58. package/md_cg/config_registry_bulk.py +19 -10
  59. package/md_cg/config_validate.py +14 -2
  60. package/md_cg/detrerank.py +137 -0
  61. package/md_cg/fsutil.py +215 -25
  62. package/md_cg/govern/__init__.py +23 -0
  63. package/md_cg/iface/__init__.py +46 -0
  64. package/md_cg/iface/_verify.py +609 -0
  65. package/md_cg/iface/contracts.py +337 -0
  66. package/md_cg/judgment_manifest.py +14 -0
  67. package/md_cg/kernel/__init__.py +27 -0
  68. package/md_cg/mcp_server.py +457 -23
  69. package/md_cg/mdcg.py +1034 -79
  70. package/md_cg/mdcos.py +263 -21
  71. package/md_cg/metacognition.py +66 -13
  72. package/md_cg/mutate_extract_shipped.py +126 -0
  73. package/md_cg/mutate_ledger_e2e.py +100 -0
  74. package/md_cg/nodefile.py +147 -23
  75. package/md_cg/op_table.py +22 -4
  76. package/md_cg/probe_c_ledger_integrity.py +258 -0
  77. package/md_cg/probe_d_retirement_discipline.py +263 -0
  78. package/md_cg/reconcile.py +9 -11
  79. package/md_cg/review_cli.py +6 -1
  80. package/md_cg/rollback.py +16 -1
  81. package/md_cg/run_tests.py +30 -1
  82. package/md_cg/scopeselect.py +214 -0
  83. package/md_cg/scrub.py +160 -23
  84. package/md_cg/security.py +4 -2
  85. package/md_cg/state_extract.py +149 -0
  86. package/md_cg/state_slots.py +58 -10
  87. package/md_cg/structboost.py +179 -0
  88. package/md_cg/test_alternatives_historical.py +162 -0
  89. package/md_cg/test_b2_sensitivity_landing.py +226 -223
  90. package/md_cg/test_b3_merge_keeps_content.py +608 -583
  91. package/md_cg/test_bucket_cards.py +85 -0
  92. package/md_cg/test_bucket_day_order.py +74 -0
  93. package/md_cg/test_bucket_quota.py +79 -0
  94. package/md_cg/test_bucket_recency_guards.py +75 -0
  95. package/md_cg/test_bucket_three_causes.py +83 -0
  96. package/md_cg/test_bucketindex.py +130 -0
  97. package/md_cg/test_bucketscope_early.py +71 -0
  98. package/md_cg/test_cond_term_fold.py +183 -0
  99. package/md_cg/test_cond_term_fold_gate.py +1210 -0
  100. package/md_cg/test_cond_term_neg_scope.py +147 -0
  101. package/md_cg/test_cond_terms_colon.py +348 -0
  102. package/md_cg/test_detrerank.py +157 -0
  103. package/md_cg/test_gate_four_kpi.py +136 -0
  104. package/md_cg/test_generation_guard.py +76 -43
  105. package/md_cg/test_issue192_move_layer_indices_guard.py +48 -0
  106. package/md_cg/test_issue362_neg_condition.py +44 -7
  107. package/md_cg/test_issue39_utf8_stdio.py +572 -572
  108. package/md_cg/test_issue413_preimage_failclosed.py +147 -9
  109. package/md_cg/test_issue444_scrub_evidence_plaintext.py +14 -8
  110. package/md_cg/test_issue677_flush_batch_io.py +139 -0
  111. package/md_cg/test_issue677_info_fast.py +82 -0
  112. package/md_cg/test_issue677_remember_dedupe.py +106 -0
  113. package/md_cg/test_issue712_tokens_sessions_read.py +199 -0
  114. package/md_cg/test_issue722_strict_conditions.py +96 -0
  115. package/md_cg/test_issue730_narrowing_audit.py +121 -0
  116. package/md_cg/test_issue733_verify_override.py +193 -0
  117. package/md_cg/test_issue733_verify_override_and_overwrite_inherit.py +315 -0
  118. package/md_cg/test_issue740_filelock_fd_leak.py +218 -0
  119. package/md_cg/test_issue741_content_hash_same_source.py +158 -0
  120. package/md_cg/test_issue747_slot_contradiction.py +125 -0
  121. package/md_cg/test_issue754_nested_root_guard.py +197 -0
  122. package/md_cg/test_issue757_index_hash_compare.py +85 -0
  123. package/md_cg/test_issue758_disk_hash_single.py +67 -0
  124. package/md_cg/test_issue759_twophase_hash.py +67 -0
  125. package/md_cg/test_issue768_overwrite_time_window_anchor.py +135 -0
  126. package/md_cg/test_issue789_log_plaintext_probe.py +100 -0
  127. package/md_cg/test_issue790_verb_negation.py +77 -0
  128. package/md_cg/test_issue810_root_guard.py +71 -0
  129. package/md_cg/test_issue812_info_index_stale.py +78 -0
  130. package/md_cg/test_issue827_weaken_dedup.py +361 -0
  131. package/md_cg/test_issue829_calibration_positive_gate.py +473 -0
  132. package/md_cg/test_kernel_iface_contracts.py +18 -0
  133. package/md_cg/test_md_cg_no_hive_dep.py +238 -17
  134. package/md_cg/test_n204_n205_n226_n227_n228_n229_exit_gates.py +545 -542
  135. package/md_cg/test_n209_verify_write_face_gates.py +5 -2
  136. package/md_cg/test_neg_condition_hits.py +336 -333
  137. package/md_cg/test_probe_cd_shipped.py +385 -0
  138. package/md_cg/test_recent_tail.py +93 -0
  139. package/md_cg/test_run_tests_entry_gates.py +110 -0
  140. package/md_cg/test_scopeselect.py +70 -0
  141. package/md_cg/test_scopeselect_clock.py +70 -0
  142. package/md_cg/test_scrub_source_line_false_positive.py +64 -0
  143. package/md_cg/test_session_attribution_consistency.py +835 -813
  144. package/md_cg/test_state_extract_shipped.py +162 -0
  145. package/md_cg/test_state_ledger_chain.py +207 -0
  146. package/md_cg/test_state_ledger_e2e.py +145 -0
  147. package/md_cg/test_structboost.py +102 -0
  148. package/md_cg/test_timeboost.py +83 -0
  149. package/md_cg/test_timeboost_relativedate.py +71 -0
  150. package/md_cg/test_topicagg.py +129 -0
  151. package/md_cg/test_twophase.py +286 -286
  152. package/md_cg/timeboost.py +223 -0
  153. package/md_cg/tokens.py +55 -7
  154. package/md_cg/topicagg.py +142 -0
  155. package/md_cg/twophase.py +15 -6
  156. package/md_cg/whitebox_kb/wisdom/code_compose.py +33 -5
  157. package/md_cg/whitebox_kb/wisdom/compose_engine.py +5 -1
  158. package/md_cg/whitebox_kb/wisdom/role_compose.py +271 -268
  159. package/md_cg/writepipe.py +35 -3
  160. package/package.json +4 -1
  161. package/scripts/workspace_index.py +2 -0
  162. package/src/hooks.ts +182 -34
  163. package/src/index.ts +19 -3
  164. package/src/lib/console_web.mjs +30 -3
  165. package/src/lib/hook_audit.ts +15 -5
  166. package/src/lib/mdcg_client.ts +8 -4
  167. package/src/lib/prompt_safety.ts +73 -1
  168. package/src/lib/token_store.ts +5 -2
  169. package/src/tools.ts +5 -3
  170. package/ui/minimal/integrity.json +2 -2
  171. package/ui/minimal/payload/build-ui.mjs +41 -2
  172. package/ui/minimal/payload/lib/client.js +29 -3
  173. package/ui/minimal/payload/lib/lib/console_web.mjs +30 -3
  174. package/ui/minimal/payload/src/lib/console_web.mjs +30 -3
  175. package/ui/minimal/payload/web/app.js +3 -1
  176. package/ui/minimal/payload/web/client.js +30 -3
  177. package/web/app.js +3 -1
  178. package/web/client.js +30 -3
  179. package/zcode/AGENTS.md +1 -1
@@ -0,0 +1,147 @@
1
+ # -*- coding: utf-8 -*-
2
+ """#753 续:否定词 + **非连接字**(使用/采用/用)作用域的守卫。
3
+
4
+ 背景
5
+ ====
6
+ `_negated_at` 原先把「否定词与被否定成分之间」的允许字符限定为连接字
7
+ (`NEG_LINK_CHARS` = 的/这/种/是/…)。于是:
8
+
9
+ 条件写 `PE`,问句 `这里未使用 pe 估值`
10
+ #761 归一前:`pe` 不命中 `PE` → DEFER ✓
11
+ #761 归一后:命中 → ACCEPT ✗(否定句被当成"条件已确认")
12
+
13
+ 缺口在**语义**:`未`否定的是动词`使用`,`pe估值`是`使用`的宾语。
14
+ 宾语被否定动词管辖,故仍是"条件未确认"。只认连接字会把这类全漏掉。
15
+
16
+ 本文件钉三件事
17
+ ==============
18
+ 1. **动词字集生效**:`未使用X` / `不是用X` / `没有使用X` / `未采用X` 等必须判为否定。
19
+ 2. **不误伤肯定句**:`使用X` / `用X`(无否定词)必须判为命中。
20
+ 3. **不碰 #362 既有语义**:`不适合梅雨季节的施工` **照旧命中**(`适合`不是动词字,
21
+ 否定词作用的是`适合`而非`梅雨季节`);A-not-A(`是不是X`)照旧命中。
22
+
23
+ 判别力:删掉 `NEG_VERB_CHARS` 的任一字符 → 第 1 组红。
24
+ 把 `NEG_VERB_CHARS` 塞进 `NEG_LINK_CHARS`(合并两表)→ 结构断言红。
25
+ """
26
+ import sys
27
+
28
+ import md_cg.mdcg as M
29
+ from md_cg.mdcg import MdCG, neg_condition_hits
30
+
31
+ PASS = FAIL = 0
32
+
33
+
34
+ def ok(cond, label):
35
+ global PASS, FAIL
36
+ if cond:
37
+ PASS += 1
38
+ else:
39
+ FAIL += 1
40
+ print(" FAIL %s" % label)
41
+
42
+
43
+ print("=" * 68)
44
+ print("R14. 否定词 + 非连接字(虚化动词)作用域")
45
+ print("=" * 68)
46
+
47
+ # ---------------------------------------------------------------- R14a 结构
48
+ # 两表**分开**存放:连接字(纯语法)与虚化动词(承载语义)词性不同,
49
+ # 合并会掩盖「否定词作用于动作、动作的宾语才是词项」这层结构。
50
+ ok(hasattr(M, "NEG_VERB_CHARS"), "R14a NEG_VERB_CHARS 应存在(独立于 NEG_LINK_CHARS)")
51
+ ok(hasattr(M, "NEG_SCOPE_CHARS"), "R14a NEG_SCOPE_CHARS 应存在(判据准入集的单点)")
52
+ ok(
53
+ set(M.NEG_SCOPE_CHARS) == set(M.NEG_LINK_CHARS) | set(M.NEG_VERB_CHARS),
54
+ "R14a NEG_SCOPE_CHARS 必须恰为 连接字 ∪ 虚化动词字(不多不少)",
55
+ )
56
+ ok(
57
+ not (set(M.NEG_VERB_CHARS) <= set("是这那该此种个些的了在为之其")),
58
+ "R14a 虚化动词字不得全部落在连接字表内(否则分表无意义)",
59
+ )
60
+ # 逐字符断言,防止有人写成多字串(frozenset 里多字串永不匹配单个字符)
61
+ ok(
62
+ all(len(c) == 1 for c in M.NEG_VERB_CHARS),
63
+ "R14a NEG_VERB_CHARS 必须逐**单字**存放(多字串在逐字符判据里永不匹配)",
64
+ )
65
+ for ch in "使用采":
66
+ ok(ch in M.NEG_VERB_CHARS, "R14a NEG_VERB_CHARS 应含 %r" % ch)
67
+
68
+ # ------------------------------------------------------- R14b 否定句须判否定
69
+ NEG_SENTS = [
70
+ ("未使用pe估值", "pe"),
71
+ ("不是用pe估值", "pe"),
72
+ ("没有使用pe估值", "pe"),
73
+ ("未采用pe估值", "pe"),
74
+ ("不是采用pe估值", "pe"),
75
+ ("未使用这种pe估值", "pe"), # 动词 + 连接字混合
76
+ ("没有用这个pe估值", "pe"), # 动词 + 连接字混合
77
+ ]
78
+ for sent, needle in NEG_SENTS:
79
+ i = sent.find(needle)
80
+ ok(i >= 0, "R14b 用例自检:%r 应含 %r" % (sent, needle))
81
+ if i < 0:
82
+ continue
83
+ ok(M._negated_at(sent, i), "R14b %r 中 %r 应判为被否定" % (sent, needle))
84
+
85
+ # ---------------------------------------------- R14c 肯定句不得被误判否定
86
+ POS_SENTS = [
87
+ ("使用pe估值", "pe"),
88
+ ("用pe估值", "pe"),
89
+ ("采用pe估值", "pe"),
90
+ ("pe估值", "pe"),
91
+ ]
92
+ for sent, needle in POS_SENTS:
93
+ i = sent.find(needle)
94
+ ok(i >= 0, "R14c 用例自检:%r 应含 %r" % (sent, needle))
95
+ if i < 0:
96
+ continue
97
+ ok(not M._negated_at(sent, i), "R14c %r 中 %r 不应被判否定" % (sent, needle))
98
+
99
+ # ------------------------------------------------- R14d #362 既有语义不变
100
+ # 这是**回归防线**:本补丁只扩字符集,不改判定结构。
101
+ # `不适合梅雨季节的施工`:否定词是`不`,其后`适合`不在准入集 ⇒ 仍判「未被否定」。
102
+ ok(
103
+ bool(neg_condition_hits(["梅雨季节"], "不适合梅雨季节的施工", "")),
104
+ "R14d `不适合梅雨季节的施工` 必须**照旧命中**(#362 语义,不得因本补丁收紧)",
105
+ )
106
+ ok(
107
+ not bool(neg_condition_hits(["梅雨季节"], "不是梅雨季节", "")),
108
+ "R14d `不是梅雨季节` 仍须判为否定(取消命中)",
109
+ )
110
+ ok(
111
+ not bool(neg_condition_hits(["梅雨季节"], "不是这种梅雨季节", "")),
112
+ "R14d `不是这种梅雨季节` 仍须判为否定",
113
+ )
114
+ ok(
115
+ bool(neg_condition_hits(["梅雨季节"], "是不是梅雨季节", "")),
116
+ "R14d A-not-A(`是不是X`)仍须命中(不得被本补丁误判为否定)",
117
+ )
118
+
119
+ # ------------------------------------- R14e 端到端:条件面在否定句上须 DEFER
120
+ CONTENT = """# 功能名:T 节点
121
+ # 生效条件:PE
122
+ # 子功能:x
123
+ # 执行:y
124
+ # 验证方式:z
125
+ # 不适用条件:无
126
+ """
127
+ NODE = {
128
+ # verification_basis 取闭集内合法值 `test`(`nodefile.VERIFICATION_BASIS`);
129
+ # 不用自由文本:字面量须在 issue #466 登记表覆盖域内,否则 B01 判红。
130
+ "frontmatter": {"non_applicable_conditions": [], "verification_basis": "test"},
131
+ "content": CONTENT,
132
+ }
133
+ for q in ["未使用 pe 估值", "不是用 pe 估值", "没有使用 pe 估值",
134
+ "未采用 pe 估值", "不是采用 pe 估值"]:
135
+ st = MdCG.judge_qualification(NODE, q, None)["state"]
136
+ ok(st == "DEFER", "R14e 否定句 %r 应 DEFER(实得 %s)" % (q, st))
137
+
138
+ # 肯定句必须保持 ACCEPT(否则是矫枉过正)
139
+ for q in ["pe 估值", "使用 pe 估值", "用 pe 估值"]:
140
+ st = MdCG.judge_qualification(NODE, q, None)["state"]
141
+ ok(st == "ACCEPT", "R14e 肯定句 %r 应 ACCEPT(实得 %s)" % (q, st))
142
+
143
+ print()
144
+ print("=" * 68)
145
+ print("cond_term_neg_scope: %d passed, %d failed" % (PASS, FAIL))
146
+ print("=" * 68)
147
+ sys.exit(1 if FAIL else 0)
@@ -0,0 +1,348 @@
1
+ # -*- coding: utf-8 -*-
2
+ """条件短语切分:冒号只对已知槽标签分隔 + 空 terms 不 fail-open。
3
+
4
+ 对应缺陷(实测于 0.9.16 与 0.9.21,两版行为一致):
5
+
6
+ ① 冒号误当槽分隔
7
+ `condition_space_text()` 产出的槽标签是固定四项
8
+ (CONDITION_SLOTS = 载体/位置、时间、方法、约束),
9
+ 而旧 `_cond_terms` 把「第一个冒号之前的一切」都当标签剥掉:
10
+ 「…必须按此处置:」 尾冒号 ⇒ 整条条件剥成空串
11
+ 「…判 ≥3:1」 正文冒号 ⇒ 前半句丢失
12
+ 「到了上午 9:00 开始开会」 正文冒号 ⇒ 只剩 "00 开始开会",前缀被吞
13
+ ② 空 terms 静默 fail-open
14
+ `judge_qualification` 的旧守卫 `if cond_hit is None and terms_:`
15
+ 在 terms_ 为空时跳过整段,直落第 5 步 ACCEPT。
16
+ ⇒ 「条件写坏了」被当成「没有条件」,对任何查询都授资格。
17
+
18
+ 本文件锁的是修好之后的行为,同时保证既有正确行为不被改坏。
19
+
20
+ 运行:python -m md_cg.test_cond_terms_colon
21
+ """
22
+ from __future__ import annotations
23
+
24
+ from .mdcg import STATE_ACCEPT, STATE_DEFER, MdCG
25
+
26
+ PASS = FAIL = 0
27
+ FAILS = []
28
+
29
+
30
+ def ok(cond, label):
31
+ global PASS, FAIL
32
+ if cond:
33
+ PASS += 1
34
+ else:
35
+ FAIL += 1
36
+ FAILS.append(label)
37
+ print(f" FAIL {label}")
38
+
39
+
40
+ def _doc(cond_line="# 生效条件:问浮力", neg_line="# 不适用条件:问完全无关主题"):
41
+ return ("# 功能名:浮力计算\n" + cond_line + "\n# 子功能:按阿基米德原理算\n"
42
+ "# 执行:代入公式\n# 验证方式:编译器/静态检查通过\n" + neg_line + "\n")
43
+
44
+
45
+ def _node(cond, neg=None):
46
+ fm = {"verification_basis": "compiler"}
47
+ if neg:
48
+ fm["non_applicable_conditions"] = [neg]
49
+ return {"frontmatter": fm,
50
+ "content": _doc(cond_line="# 生效条件:" + cond)}
51
+
52
+
53
+ def _state(cond, query="今天天气怎么样", context=None):
54
+ r = MdCG.judge_qualification(_node(cond), query, context)
55
+ return r["state"], r["reason"]
56
+
57
+
58
+ def main():
59
+ global PASS, FAIL
60
+
61
+ # ---------- A 冒号:整条条件必须保住 ----------
62
+ ok(MdCG._cond_terms("三个条件同时成立时必须按此处置:")
63
+ == ["三个条件同时成立时必须按此处置"],
64
+ "A1 全角冒号结尾:整条保住,首尾标点剥离")
65
+ ok(MdCG._cond_terms("三个条件同时成立时必须按此处置:")
66
+ == ["三个条件同时成立时必须按此处置"],
67
+ "A2 半角冒号结尾:同上")
68
+ ok(MdCG._cond_terms("到了上午 9:00 开始开会时")
69
+ == ["到了上午 9:00 开始开会时"],
70
+ "A3 正文含时间 9:00:前缀不被吞")
71
+ ok(any("3:1" in t for t in
72
+ MdCG._cond_terms("页面含承载信息的线条,需按 WCAG 判 ≥3:1")),
73
+ "A4 正文含比值 3:1:内容保留")
74
+ # ⚠ URL 里的 `//` 仍会被切(`/` 是槽内分隔符,上游原行为,本补丁**不涉及**)。
75
+ # 这里如实断言现状,不假装已修。若要修,需另议「/ 何时算分隔符」。
76
+ ok(MdCG._cond_terms("参考 http://example.com 时")
77
+ == ["参考 http", "example.com 时"],
78
+ "A5 URL 的 // 仍按 / 切开(已知未修,如实锁定现状)")
79
+
80
+ # ---------- B 四槽标签仍被剥离(不能被 A 改坏) ----------
81
+ cs = MdCG._cond_terms(
82
+ "载体/位置:本机进程;时间:全时窗(任意时刻成立);"
83
+ "方法:代码走查;约束:仅测试库")
84
+ ok("本机进程" in cs and "代码走查" in cs and "仅测试库" in cs,
85
+ "B1 四槽实词保留")
86
+ ok("载体/位置" not in cs and "时间" not in cs and "方法" not in cs,
87
+ "B2 槽标签不进 terms")
88
+ ok(all("全时窗" not in t and "任意时刻" not in t for t in cs),
89
+ "B3 全时窗哨兵被过滤")
90
+
91
+ # ---------- C 空 terms 不再 fail-open ----------
92
+ for bad in ("k", "1", "{", "a", ":", "。"):
93
+ st, rs = _state(bad)
94
+ ok(st != STATE_ACCEPT,
95
+ f"C1 不可用条件 {bad!r} 不再恒 ACCEPT(得 {st})")
96
+ ok("切不出可用短语" in rs,
97
+ f"C2 不可用条件 {bad!r} 的 reason 说明原因")
98
+
99
+ # ---------- D 既有正确行为不变 ----------
100
+ st, rs = _state("换会话")
101
+ ok(st == STATE_DEFER, "D1 条件无关 → DEFER")
102
+ st, rs = _state("换会话", context={"topic": "换会话"})
103
+ ok(st == STATE_ACCEPT and "生效条件已确认" in rs, "D2 context 命中 → ACCEPT")
104
+ st, rs = _state("三个条件同时成立时必须按此处置:",
105
+ "现在三个条件同时成立时必须按此处置,怎么办")
106
+ ok(st == STATE_ACCEPT and "生效条件已确认" in rs,
107
+ "D3 冒号条件命中时仍 ACCEPT(首尾标点不挡匹配)")
108
+
109
+ # ---------- E 文档例外仍成立 ----------
110
+ st, rs = _state("无条件")
111
+ ok(st == STATE_ACCEPT, "E1 显式『无条件』→ ACCEPT(设计豁免)")
112
+ st, rs = _state("全时窗")
113
+ ok(st == STATE_ACCEPT, "E2 只含全时窗 → 不降级(文档口径)")
114
+ st, rs = _state("", "")
115
+ ok(st == STATE_ACCEPT and "未做正条件确认" in rs,
116
+ "E3 空条件 + 无情境 → ACCEPT 且如实标注未确认")
117
+ st, rs = _state("")
118
+ ok(st == STATE_ACCEPT and "切不出可用短语" not in rs,
119
+ "E4 空条件不算『切不出』(不误报)")
120
+
121
+ # ---------- F 负条件仍先于正条件 REJECT ----------
122
+ r = MdCG.judge_qualification(_node("问浮力", "问ZXQ7"), "问ZXQ7",
123
+ {"query": "问ZXQ7"})
124
+ ok(r["state"] == "REJECT", "F1 负条件命中仍 REJECT")
125
+
126
+ # ---------- G 引号 / 感叹号是代码字符,不是首尾标点 ----------
127
+ # 旧口径把 ' " ! 放进首尾标点表:`''` 剥成空串被丢、`!v` 剥成 `v`
128
+ # 再被 len<2 丢掉 —— 两者都是有效 token(空字符串字面量 / 逻辑非)。
129
+ ok(MdCG._cond_terms("''") == ["''"],
130
+ "G1 空字符串字面量 '' 保留(不是首尾标点)")
131
+ ok(MdCG._cond_terms('""') == ['""'], 'G2 空字符串字面量 "" 保留')
132
+ ok(MdCG._cond_terms("'!v'") == ["!v"],
133
+ "G3 成对引号包裹时剥外层,内层 !v 保留")
134
+ ok(MdCG._cond_terms("!v 是逻辑非") == ["!v 是逻辑非"],
135
+ "G4 句中 ! 不被当标点剥")
136
+ ok(MdCG._cond_terms("''与'!v'") == ["''与'!v'"],
137
+ "G5 相邻两个字面量不被外层引号粘连(内层仍带引号 ⇒ 不剥)")
138
+
139
+ # ---------- H 空值语义哨兵是合法声明,不是「切不出」 ----------
140
+ # `约束:无` 里的「无」只有 1 字:旧实现靠 len<2 静默滤掉、恰好 ACCEPT。
141
+ # 若补丁只用长度判定,会把它误判成 DEFER —— 必须在长度过滤前认出哨兵。
142
+ ok(MdCG._cond_terms("约束:无") == [],
143
+ "H1 「约束:无」terms 为空(1 字哨兵被过滤)")
144
+ st, rs = _state("约束:无")
145
+ ok(st == STATE_ACCEPT,
146
+ "H2 「约束:无」仍 ACCEPT(合法空值声明,不降级)")
147
+ ok("切不出可用短语" not in rs, "H3 「约束:无」不报『切不出』")
148
+ ok(MdCG._cond_terms("约束:不适用") == ["不适用"],
149
+ "H4 多字哨兵仍作为可匹配短语(口径未放宽)")
150
+ ok(MdCG._cond_terms("约束:无依赖") == ["无依赖"],
151
+ "H5 同上:无依赖 未被哨兵规则吞掉")
152
+
153
+ # ---------- I 硬不可用桶仍不豁免 ----------
154
+ for bad, label in (("载体/位置:。;方法:()", "纯标点"),
155
+ ("1", "纯数字"),
156
+ ("a", "单字碎片")):
157
+ st, rs = _state(bad)
158
+ ok(st == STATE_DEFER and "切不出可用短语" in rs,
159
+ f"I 硬不可用({label})仍 DEFER 且说明原因")
160
+
161
+ # ---------- J 计数可归因(不是合成一个 short) ----------
162
+ d = MdCG._cond_terms_diag("约束:无")
163
+ ok(d["n_decl_sentinel"] == 1 and d["n_short"] == 0,
164
+ "J1 合法的 1 字哨兵记进 n_decl_sentinel,不记 n_short")
165
+ d = MdCG._cond_terms_diag("a")
166
+ ok(d["n_short"] == 1 and d["n_decl_sentinel"] == 0,
167
+ "J2 真正的 1 字碎片记进 n_short")
168
+ d = MdCG._cond_terms_diag("时间:全时窗(任意时刻成立)")
169
+ ok(d["n_time_sentinel"] >= 1, "J3 时间维哨兵单独计数")
170
+ ok(MdCG._cond_terms_diag("")["n_empty"] == 0,
171
+ "J4 空条件不记 n_empty(『没写条件』≠『写了纯标点』)")
172
+
173
+ # ---------- K 切分单点:三处入口必须同一口径 ----------
174
+ import md_cg.nodefile as _nf
175
+ probe = ["约束:无", "''", "'!v'", "三个条件同时成立时必须按此处置:",
176
+ "载体/位置:。;方法:()", ""]
177
+ same = all(
178
+ MdCG._cond_terms(p) == _nf.cond_terms(p) == _nf.condition_terms_diag(p)["terms"]
179
+ for p in probe)
180
+ ok(same, "K1 MdCG._cond_terms / nodefile.cond_terms / condition_terms_diag 同口径")
181
+
182
+ # ---------- L 标签真源守卫 ----------
183
+ # L1 原本断言 `CONDITION_SLOTS 的标签列 == COND_SLOT_LABELS`。两者在
184
+ # nodefile 里都由 CONDITION_SLOTS 派生(`COND_SLOT_LABELS = tuple(label for
185
+ # _key, label in CONDITION_SLOTS)`),所以该断言**恒真**,且 `python -O`
186
+ # 会把类体里那条同名 assert 整体剥离 —— 防护力为 0(2026-11-02 review)。
187
+ # 换成真正会红的两条:
188
+ # L1 标签列必须是「每项非空、互不相同」的四元组 —— 手滑抄重复/抄空会红;
189
+ # L2 切片真源只能是 nodefile 一处 —— MdCG 不再挂同名别名(别名已删),
190
+ # 一旦有人把标签又抄回 mdcg.py,这条会红。
191
+ _labels = _nf.COND_SLOT_LABELS
192
+ ok(len(_labels) == len(_nf.CONDITION_SLOTS),
193
+ "L1a 标签列与槽数同长")
194
+ ok(all(isinstance(x, str) and x.strip() for x in _labels),
195
+ "L1b 标签列每项都是非空字符串")
196
+ ok(len(set(_labels)) == len(_labels),
197
+ "L1c 标签列无重复(抄重会红)")
198
+ ok(not hasattr(MdCG, "_SLOT_LABELS") and not hasattr(MdCG, "_EDGE_PUNCT"),
199
+ "L2 标签/标点真源只在 nodefile,MdCG 不挂可漂移的别名")
200
+
201
+ # ---------- M legacy「观测位置:」单槽冒充要剥掉 ----------
202
+ # 旧口径残留行 `观测位置:X` 是单槽冒充;`观测位置` 不在四槽标签里,
203
+ # 若不显式剥,它会作为一个短语进入匹配面(维护者例4 指的就是这条)。
204
+ legacy = MdCG._cond_terms("观测位置:PmWiki;时间:9:00;方法:浏览器;约束:只读")
205
+ ok(legacy == ["PmWiki", "9:00", "浏览器", "只读"],
206
+ "M5-legacy legacy 前缀被剥,不残留在 terms 里")
207
+ ok(not any(t.startswith("观测位置") for t in legacy),
208
+ "M6-legacy 无 terms 以「观测位置」开头")
209
+
210
+ # ---------- N 括号产生的结构性空段不得计入 n_empty ----------
211
+ # 2026-11-02 review 后补:维护者指出「5b 可观测性缺口」时,我按同一方向
212
+ # 又查出一条**本 PR 自己的回归**:`re.split(r"[,,、/()()]", s)` 在
213
+ # `(…)` 处会产生**结构性空段**(括号本身被当分隔符切掉后留下的空串)。
214
+ # 而原实现是「先 strip 再 if not seg: n_empty += 1」,把
215
+ # 「split 出来的空串」与「真被剥成空的纯标点段」混为一谈,
216
+ # 于是 `时间:全时窗(任意时刻成立)` 多出 n_empty=1,
217
+ # unusable = n_empty+n_digit+n_short > 0 成立,
218
+ # 把**本应豁免**的时间维哨兵误判成 DEFER —— 而源码注释明写
219
+ # 「时间维哨兵……不因其未命中而降级」。
220
+ # 修法:在**剥边之前**判结构性空段(`if not seg: continue`),
221
+ # 剥边之后变空的才计入 n_empty。
222
+ _paren = MdCG._cond_terms_diag("时间:全时窗(任意时刻成立)")
223
+ ok(_paren["n_empty"] == 0,
224
+ "N1 括号产生的结构性空段不计入 n_empty")
225
+ ok(_paren["n_time_sentinel"] >= 1,
226
+ "N2 时间维哨兵被识别(不计入不可用桶)")
227
+ ok(_paren["terms"] == [],
228
+ "N3 哨兵不进 terms(本就不该参与匹配)")
229
+ _unusable = _paren["n_empty"] + _paren["n_digit"] + _paren["n_short"]
230
+ ok(_unusable == 0,
231
+ "N4 哨兵条件走豁免路径(unusable==0 才不会被判 DEFER)")
232
+
233
+ _seg = MdCG._cond_terms_diag("载体/位置:页面(含线条)")
234
+ ok(_seg["terms"] == ["页面", "含线条"],
235
+ "N5 括号内内容仍被正常切出(修法不吞内容)")
236
+ ok(_seg["n_empty"] == 0,
237
+ "N6 有括号但无纯标点残段时 n_empty 为 0")
238
+
239
+ # 反向:真纯标点段**必须**仍计入 n_empty,否则修法把「不可用」漏掉
240
+ _punct = MdCG._cond_terms_diag(":。;、")
241
+ ok(_punct["n_empty"] >= 1,
242
+ "N7 真纯标点残段仍计入 n_empty(修法不放宽不可用判据)")
243
+
244
+ # ---------- O 三种「跳过正条件比对」的原因必须可区分 ----------
245
+ # review 5b:`约束:无`(空值声明)/ 全时窗哨兵 / 根本没写条件
246
+ # 三者语义不同,但 ACCEPT 的 reason 曾完全相同 —— 日志上无法归因。
247
+ _TPL = (
248
+ "# 功能名:探测\n\n# 生效条件:{c}\n\n# 子功能:s\n\n"
249
+ "# 执行:e\n\n# 验证方式:v\n\n# 不适用条件:情境不匹配时\n"
250
+ )
251
+
252
+ def _nd(c, cs=None):
253
+ fm = {"non_applicable_conditions": ["情境不匹配时"],
254
+ "verification_basis": "test"}
255
+ if cs is not None:
256
+ fm["condition_space"] = cs
257
+ return {"frontmatter": fm, "content": _TPL.format(c=c)}
258
+
259
+ _cs = {"observation_position": "本机进程",
260
+ "time_window": [0, 9999999999],
261
+ "observation_tool": "代码走查",
262
+ "existence_constraint": "仅测试库"}
263
+ _r_none = MdCG.judge_qualification(_nd(""), "本机进程", None)
264
+ _r_decl = MdCG.judge_qualification(_nd("约束:无"), "随便", None)
265
+ _r_time = MdCG.judge_qualification(
266
+ _nd("时间:全时窗(任意时刻成立)"), "随便", None)
267
+ ok(_r_none["state"] == "ACCEPT" and _r_decl["state"] == "ACCEPT"
268
+ and _r_time["state"] == "ACCEPT",
269
+ "O1 三种豁免都仍 ACCEPT(修的是可观测性,不改判定)")
270
+ _reasons = {_r_none["reason"], _r_decl["reason"], _r_time["reason"]}
271
+ ok(len(_reasons) == 3,
272
+ "O2 三种豁免的 reason 互不相同(5b 可观测性缺口已堵)")
273
+ ok("未声明生效条件" in _r_none["reason"],
274
+ "O3 「没写条件」有显式标记")
275
+ ok("空值声明" in _r_decl["reason"],
276
+ "O4 「约束:无」有显式标记")
277
+ ok("全时窗" in _r_time["reason"],
278
+ "O5 时间维哨兵有显式标记")
279
+
280
+ # ---------- P ASCII 词边界:路径碎片不得咬住无关英文词 ----------
281
+ # PR #735 把切分修好之后,`backend/app/core/...`、`/api/uzi/...` 这类路径
282
+ # 被切成 `app` / `core` / `api` 三个碎片进入匹配面。旧判据是裸子串
283
+ # `t in scene_str`,于是:
284
+ # "capital".find("api") == 1
285
+ # "mapping".find("app") == 1
286
+ # "score".find("core") == 4
287
+ # 三条都会在**条件完全没被确认**的情况下误判成「已确认」→ 假 ACCEPT。
288
+ # 本段锁的是「纯 ASCII 字母词按词边界,其余保持子串」这一口径。
289
+ ok(MdCG._term_in_scene("api", "capital 结构是否合理") is False,
290
+ "P1 api 不得命中 capital")
291
+ ok(MdCG._term_in_scene("app", "这个 mapping 对不对") is False,
292
+ "P2 app 不得命中 mapping")
293
+ ok(MdCG._term_in_scene("core", "计算 score 的方式") is False,
294
+ "P3 core 不得命中 score")
295
+ ok(MdCG._term_in_scene("api", "后端 /api/uzi/peers 有没有问题") is True,
296
+ "P4 api 在真路径里仍命中(不能为了收紧而漏掉)")
297
+ ok(MdCG._term_in_scene("core", "backend/app/core/uxi") is True,
298
+ "P5 core 在真路径里仍命中")
299
+ ok(MdCG._term_in_scene("UTC", "时区是 UTC 还是本地") is True,
300
+ "P6a UTC 独立成词命中")
301
+ ok(MdCG._term_in_scene("UTC", "时区是 UTCX 还是本地") is False,
302
+ "P6b UTC 右侧紧邻字母则不命中")
303
+ ok(MdCG._term_in_scene("本机进程", "载体/位置:本机进程;时间:全时窗") is True,
304
+ "P7 中文仍走子串(不收紧)")
305
+ ok(MdCG._term_in_scene("9:00", "到了上午 9:00 开始开会") is True,
306
+ "P8 非字母 ASCII(时间)仍走子串")
307
+ ok(MdCG._term_in_scene("3:1", "需按 WCAG 判 ≥3:1") is True,
308
+ "P9 非字母 ASCII(比值)仍走子串")
309
+ ok(MdCG._term_in_scene(None, "任意") is False
310
+ and MdCG._term_in_scene("", "任意") is False,
311
+ "P10 None / 空串按不命中")
312
+
313
+ # P14~P16 锁的是「收紧范围不能外溢」。三条都是把词边界写宽/写窄的
314
+ # 典型错法:写宽(对所有词都用 \b)会把中文整条判死;写窄(只认空白/标点)
315
+ # 会让中文紧邻英文时漏命中;把非字母 ASCII 也纳入收紧会连 `9:00` 一起废掉。
316
+ ok(MdCG._term_in_scene("本机进程", "cpu本机进程是否可用") is True,
317
+ "P14 中文短语紧邻拉丁字母时仍命中(收紧不得外溢到中文)")
318
+ ok(MdCG._term_in_scene("本机进程", "本机进程x 是否可用") is True,
319
+ "P15 中文短语右侧紧邻拉丁字母仍命中")
320
+ ok(MdCG._term_in_scene("api", "调用api接口时") is True,
321
+ "P16a ASCII 词紧邻中文仍命中(中文不是 ASCII 字母,不算越界)")
322
+ ok(MdCG._term_in_scene("api", "/api1/x") is True,
323
+ "P16b ASCII 词紧邻数字仍命中(只挡字母,不挡数字)")
324
+ ok(MdCG._term_in_scene("9:00", "x9:00开始") is True,
325
+ "P17 非字母 ASCII 不纳入收紧(9:00 左侧拉丁字母也照旧命中)")
326
+
327
+ # 端到端:同一条条件,三条问句必须分开判
328
+ _api_node = _node("端点 /api/uzi/peers 与 backend/app/core 两个入口")
329
+ _r_bad = MdCG.judge_qualification(_api_node, "请帮我看看 capital 结构是否合理")
330
+ ok(_r_bad["state"] == STATE_DEFER,
331
+ "P11 条件讲 /api/…,问句讲 capital ⇒ DEFER(不再假 ACCEPT)")
332
+ _r_good = MdCG.judge_qualification(_api_node, "后端 api 有没有问题")
333
+ ok(_r_good["state"] == STATE_ACCEPT
334
+ and "生效条件已确认" in _r_good["reason"],
335
+ "P12 同一条条件,问句真讲 api ⇒ 仍 ACCEPT(未过度收紧)")
336
+ _r_map = MdCG.judge_qualification(_api_node, "这个 mapping 对不对")
337
+ ok(_r_map["state"] == STATE_DEFER,
338
+ "P13 问句讲 mapping ⇒ DEFER(app 不得咬住 mapping)")
339
+
340
+ print(f"\ncond_terms_colon: {PASS} passed, {FAIL} failed")
341
+ if FAILS:
342
+ for f in FAILS:
343
+ print(f" - {f}")
344
+ raise SystemExit(1 if FAIL else 0)
345
+
346
+
347
+ if __name__ == "__main__":
348
+ main()
@@ -0,0 +1,157 @@
1
+ # -*- coding: utf-8 -*-
2
+ """确定性重排(detrerank)守卫。
3
+
4
+ 钉三件事:
5
+ 1. 四信号可数且确定(同输入同输出);
6
+ 2. **无模型分**——模块源码不得出现 embedding/similarity/模型调用;
7
+ 3. **时钟显式**——rerank 必须由调用方传 now,函数内不得取 time.time()。
8
+ """
9
+ from __future__ import annotations
10
+
11
+ import inspect
12
+ import io
13
+ import os
14
+ import sys
15
+
16
+ sys.stdout.reconfigure(encoding="utf-8", errors="replace")
17
+ _REPO = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
18
+ sys.path.insert(0, _REPO)
19
+
20
+ from md_cg import detrerank as D # noqa: E402
21
+
22
+ FAILS = []
23
+ NOW = 1_800_000_000.0
24
+
25
+
26
+ def ok(cond, name, detail=""):
27
+ if cond:
28
+ print(" PASS " + name)
29
+ else:
30
+ print(" FAIL " + name + " " + str(detail)[:200])
31
+ FAILS.append(name)
32
+
33
+
34
+ def _code_only(path):
35
+ """把源码解析成 AST 后**去掉 docstring**再 unparse —— 只留代码。
36
+
37
+ 为什么需要:本模块的说明文字会提到「不得出现 embedding / time.time」这类词,
38
+ 直接 grep 源码会把**守卫自己的说明**判成违规(假红)。守卫要盯实现,不是措辞。
39
+ """
40
+ import ast
41
+ tree = ast.parse(io.open(path, encoding="utf-8").read())
42
+ for nd in ast.walk(tree):
43
+ body = getattr(nd, "body", None)
44
+ if (isinstance(body, list) and body
45
+ and isinstance(body[0], ast.Expr)
46
+ and isinstance(body[0].value, ast.Constant)
47
+ and isinstance(body[0].value.value, str)):
48
+ body[0] = ast.Pass()
49
+ return ast.unparse(tree)
50
+
51
+
52
+ def node(nid, content, tags=None, created_at=None, edges=None, layer="knowledge", session=None):
53
+ fm = {"layer": layer, "tags": tags or []}
54
+ if created_at is not None:
55
+ fm["created_at"] = created_at
56
+ if edges is not None:
57
+ fm["edges"] = edges
58
+ if session:
59
+ fm["session"] = session
60
+ return {"id": nid, "content": content, "frontmatter": fm}
61
+
62
+
63
+ def main():
64
+ print("=" * 66)
65
+ print("确定性重排(detrerank)")
66
+ print("=" * 66)
67
+ q = "灯箱 报价 制作"
68
+ terms = D._terms(q)
69
+ ok(terms == ["灯箱", "报价", "制作"], "D1 分词确定", terms)
70
+
71
+ a = node("a", "灯箱已有报价,尚未安排制作", ["灯箱"], created_at=NOW - 3600)
72
+ b = node("b", "无关内容", [], created_at=NOW - 86400 * 365)
73
+ sa = D.signals(a, terms, NOW)
74
+ sb = D.signals(b, terms, NOW)
75
+ ok(sa["lexical"] == 3 and sb["lexical"] == 0,
76
+ "D2 词法信号 = query 词命中数", (sa["lexical"], sb["lexical"]))
77
+ ok(0.0 < sa["temporal"] <= 1.0, "D3 时间信号落在 (0,1]", sa["temporal"])
78
+ ok(sb["temporal"] < sa["temporal"], "D4 旧条目时间分更低", (sa["temporal"], sb["temporal"]))
79
+
80
+ n_edge = node("e", "灯箱 报价", ["灯箱"], created_at=NOW - 10,
81
+ edges=[{"target": "x", "relation_type": "causal"}])
82
+ se = D.signals(n_edge, terms, NOW)
83
+ ok(se["interaction"] == 1, "D5 交互信号 = 出边数(真的发生过作用)", se["interaction"])
84
+
85
+ n_same = node("s", "灯箱", ["灯箱"], created_at=NOW, session="sess-1")
86
+ ss = D.signals(n_same, terms, NOW, session="sess-1")
87
+ sn = D.signals(n_same, terms, NOW, session="sess-2")
88
+ ok(ss["spatial"] > sn["spatial"], "D6 空间信号含同会话加成", (ss["spatial"], sn["spatial"]))
89
+
90
+ # D7 全序 + 幂等:同输入两次结果逐位相同
91
+ items = [(a, 0.9, {}, []), (b, 0.9, {}, []), (n_edge, 0.9, {}, [])]
92
+ r1 = D.rerank(items, q, NOW)
93
+ r2 = D.rerank(items, q, NOW)
94
+ ok([x[0][0]["id"] for x in r1] == [x[0][0]["id"] for x in r2], "D7 同输入同输出(可回放)")
95
+ ok([x[0][0]["id"] for x in r1][0] != "b", "D8 词法强的排在前面", [x[0][0]["id"] for x in r1])
96
+
97
+ # D9 分数相同按 id 字典序(全序兜底)
98
+ twins = [(node("z", "同样内容", []), 0.5, {}, []),
99
+ (node("a", "同样内容", []), 0.5, {}, [])]
100
+ rt = D.rerank(twins, q, NOW)
101
+ ok([x[0][0]["id"] for x in rt] == ["a", "z"], "D9 同分按 id 字典序",
102
+ [x[0][0]["id"] for x in rt])
103
+
104
+ # D10/D11b:只扫**代码**(去 docstring/注释)——说明文字里出现这些词是允许的,
105
+ # 守卫要盯的是「实现里有没有真的引入模型分/隐式取时钟」。
106
+ code = _code_only(os.path.join(_REPO, "md_cg", "detrerank.py")).lower()
107
+ banned = [w for w in ("embedding", "cosine", "similarity", "openai", "model.")
108
+ if w in code]
109
+ ok(not banned, "D10 代码内无模型分/相似度项", banned)
110
+
111
+ sig = inspect.signature(D.rerank)
112
+ ok("now" in sig.parameters, "D11a rerank 形参含 now(时钟显式)", list(sig.parameters))
113
+ ok("time.time" not in code and "datetime.now" not in code,
114
+ "D11b 代码内不取系统时钟")
115
+
116
+ # D12 权重是固定常量(改值必须走评审)
117
+ ok(isinstance(D.WEIGHTS, dict) and set(D.WEIGHTS) == {"lexical", "interaction", "spatial", "temporal"},
118
+ "D12 四信号权重常量齐全", D.WEIGHTS)
119
+
120
+ # ---------- 端到端:接线层(recall 的 MDCG_DET_RERANK 开关) ----------
121
+ import tempfile
122
+ from md_cg.mdcos import MdCGOS
123
+ root = tempfile.mkdtemp(prefix="det_e2e_")
124
+ g2 = MdCGOS(root, actor="test")
125
+ ccg = ("# 功能名:{n}\n# 生效条件:{c}\n# 子功能:s\n# 执行:e\n"
126
+ "# 验证方式:test\n# 不适用条件:无\n\n{b}\n")
127
+ g2.add("a", ccg.format(n="灯箱报价记录", c="灯箱 报价",
128
+ b="灯箱 已有报价 制作 待安排"), tags=["灯箱"])
129
+ g2.add("b", ccg.format(n="灯箱制作安排", c="灯箱 制作",
130
+ b="制作 安排 灯箱"), tags=["灯箱"])
131
+ g2.add("c", ccg.format(n="其它项目", c="别的",
132
+ b="别的内容 灯箱 提了一句"), tags=["其它"])
133
+ os.environ["MDCG_DET_RERANK"] = "1"
134
+ try:
135
+ p1 = g2.recall("灯箱 报价 制作", budget_tokens=3000, k=10)
136
+ ids1 = [e["id"] for e in p1["pack"]]
137
+ ok(p1["meta"].get("det_reranked") is True, "E1 开启后 meta.det_reranked=True")
138
+ ok(all("det" in e for e in p1["pack"]), "E2 每条带四信号读数(可审计)")
139
+ ok(ids1[0] == "a", "E3 词法最强的排第一", ids1)
140
+ sig0 = p1["pack"][0]["det"]["signals"]
141
+ ok(set(sig0) == {"lexical", "interaction", "spatial", "temporal"},
142
+ "E4 信号四键齐全", sig0)
143
+ p2 = g2.recall("灯箱 报价 制作", budget_tokens=3000, k=10)
144
+ ok([e["id"] for e in p2["pack"]] == ids1, "E5 同输入同输出(可回放)")
145
+ finally:
146
+ os.environ.pop("MDCG_DET_RERANK", None)
147
+ p3 = g2.recall("灯箱 报价 制作", budget_tokens=3000, k=10)
148
+ ok(p3["meta"].get("det_reranked") in (False, None), "E6 缺省关闭")
149
+ ok(not any("det" in e for e in p3["pack"]), "E7 关闭时不注入 det 读数")
150
+
151
+ print()
152
+ print("detrerank:%d 项失败" % len(FAILS) if FAILS else "detrerank:全部通过")
153
+ return 1 if FAILS else 0
154
+
155
+
156
+ if __name__ == "__main__":
157
+ sys.exit(main())