@furongjun1999/dsh-memory 0.9.23 → 0.10.3

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 (118) hide show
  1. package/CHANGELOG.en.md +515 -404
  2. package/CHANGELOG.md +642 -449
  3. package/README.en.md +244 -240
  4. package/README.md +272 -262
  5. package/codebuddy/CODEBUDDY.md +1 -1
  6. package/docs/discipline/templates/full.md.tmpl +1 -1
  7. package/docs/discipline/templates/rules.mdc.tmpl +1 -1
  8. package/docs/en/mdcg/world-model-endpoint.md +1 -1
  9. package/docs/mdcg/cogmap_ops.md +19 -16
  10. package/lib/client.js +29 -3
  11. package/lib/hooks.d.ts +12 -0
  12. package/lib/hooks.js +43 -1
  13. package/lib/tools.d.ts +2 -2
  14. package/md_cg/bucketindex.py +231 -0
  15. package/md_cg/bucketscope.py +357 -0
  16. package/md_cg/cohesion.py +337 -0
  17. package/md_cg/config_registry_bulk.py +24 -10
  18. package/md_cg/config_validate.py +14 -2
  19. package/md_cg/constraint_audit.py +231 -0
  20. package/md_cg/constraint_watch.py +206 -0
  21. package/md_cg/context_audit.py +240 -0
  22. package/md_cg/context_plan.py +142 -0
  23. package/md_cg/detrerank.py +137 -0
  24. package/md_cg/distill.py +247 -0
  25. package/md_cg/dossier.py +151 -0
  26. package/md_cg/entities.py +133 -0
  27. package/md_cg/fsutil.py +135 -0
  28. package/md_cg/govern/__init__.py +23 -0
  29. package/md_cg/iface/__init__.py +46 -0
  30. package/md_cg/iface/_verify.py +609 -0
  31. package/md_cg/iface/contracts.py +337 -0
  32. package/md_cg/judgment_manifest.py +14 -0
  33. package/md_cg/kernel/__init__.py +27 -0
  34. package/md_cg/mcp_server.py +6044 -5761
  35. package/md_cg/mdcg.py +501 -16
  36. package/md_cg/mdcos.py +546 -12
  37. package/md_cg/metacognition.py +66 -13
  38. package/md_cg/mutate_extract_shipped.py +126 -0
  39. package/md_cg/mutate_ledger_e2e.py +100 -0
  40. package/md_cg/op_table.py +22 -0
  41. package/md_cg/probe_c_ledger_integrity.py +258 -0
  42. package/md_cg/probe_d_retirement_discipline.py +263 -0
  43. package/md_cg/profiles.py +298 -0
  44. package/md_cg/protocol.py +7 -1
  45. package/md_cg/reldate.py +142 -0
  46. package/md_cg/review_cli.py +6 -1
  47. package/md_cg/run_tests.py +30 -1
  48. package/md_cg/scopeselect.py +214 -0
  49. package/md_cg/scrub.py +66 -9
  50. package/md_cg/selfstmt.py +139 -0
  51. package/md_cg/state_extract.py +149 -0
  52. package/md_cg/state_slots.py +62 -10
  53. package/md_cg/state_view.py +181 -0
  54. package/md_cg/structboost.py +179 -0
  55. package/md_cg/sustain.py +28 -2
  56. package/md_cg/test_alternatives_historical.py +165 -0
  57. package/md_cg/test_b2_sensitivity_landing.py +226 -223
  58. package/md_cg/test_b3_merge_keeps_content.py +608 -583
  59. package/md_cg/test_bucket_cards.py +85 -0
  60. package/md_cg/test_bucket_day_order.py +74 -0
  61. package/md_cg/test_bucket_quota.py +79 -0
  62. package/md_cg/test_bucket_recency_guards.py +75 -0
  63. package/md_cg/test_bucket_three_causes.py +83 -0
  64. package/md_cg/test_bucket_time_rounds.py +87 -0
  65. package/md_cg/test_bucketindex.py +130 -0
  66. package/md_cg/test_bucketscope_early.py +71 -0
  67. package/md_cg/test_cohesion.py +107 -0
  68. package/md_cg/test_cond_term_fold_gate.py +1210 -0
  69. package/md_cg/test_cond_term_neg_scope.py +147 -0
  70. package/md_cg/test_constraint_audit.py +87 -0
  71. package/md_cg/test_constraint_watch.py +108 -0
  72. package/md_cg/test_context_audit.py +110 -0
  73. package/md_cg/test_context_plan.py +81 -0
  74. package/md_cg/test_detrerank.py +157 -0
  75. package/md_cg/test_gate_four_kpi.py +136 -0
  76. package/md_cg/test_generation_guard.py +14 -2
  77. package/md_cg/test_issue192_move_layer_indices_guard.py +48 -0
  78. package/md_cg/test_issue39_utf8_stdio.py +572 -572
  79. package/md_cg/test_issue413_preimage_failclosed.py +147 -9
  80. package/md_cg/test_issue768_overwrite_time_window_anchor.py +135 -0
  81. package/md_cg/test_issue789_log_plaintext_probe.py +100 -0
  82. package/md_cg/test_issue790_verb_negation.py +77 -0
  83. package/md_cg/test_issue810_root_guard.py +71 -0
  84. package/md_cg/test_issue812_info_index_stale.py +78 -0
  85. package/md_cg/test_issue827_weaken_dedup.py +361 -0
  86. package/md_cg/test_issue829_calibration_positive_gate.py +473 -0
  87. package/md_cg/test_kernel_iface_contracts.py +18 -0
  88. package/md_cg/test_md_cg_no_hive_dep.py +238 -17
  89. package/md_cg/test_probe_cd_shipped.py +385 -0
  90. package/md_cg/test_recent_tail.py +93 -0
  91. package/md_cg/test_run_tests_entry_gates.py +110 -0
  92. package/md_cg/test_scopeselect.py +70 -0
  93. package/md_cg/test_scopeselect_clock.py +70 -0
  94. package/md_cg/test_session_attribution_consistency.py +835 -813
  95. package/md_cg/test_state_extract_shipped.py +162 -0
  96. package/md_cg/test_state_ledger_chain.py +207 -0
  97. package/md_cg/test_state_ledger_e2e.py +145 -0
  98. package/md_cg/test_structboost.py +102 -0
  99. package/md_cg/test_timeboost.py +83 -0
  100. package/md_cg/test_timeboost_relativedate.py +71 -0
  101. package/md_cg/test_topicagg.py +129 -0
  102. package/md_cg/timeboost.py +223 -0
  103. package/md_cg/tokens.py +1 -1
  104. package/md_cg/topicagg.py +142 -0
  105. package/md_cg/version_anchor.py +84 -1
  106. package/md_cg/whitebox_kb/wisdom/code_compose.py +33 -5
  107. package/md_cg/whitebox_kb/wisdom/code_solidified.json +6195 -6195
  108. package/md_cg/whitebox_kb/wisdom/compose_engine.py +5 -1
  109. package/md_cg/whitebox_kb/wisdom/role_compose.py +271 -268
  110. package/package.json +1 -1
  111. package/src/hooks.ts +43 -1
  112. package/ui/minimal/payload/build-ui.mjs +41 -2
  113. package/ui/minimal/payload/lib/client.js +29 -3
  114. package/ui/minimal/payload/web/app.js +3 -1
  115. package/ui/minimal/payload/web/client.js +30 -3
  116. package/web/app.js +3 -1
  117. package/web/client.js +30 -3
  118. 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 +1,87 @@
1
+ # -*- coding: utf-8 -*-
2
+ """md_cg.constraint_audit 的守卫测试。
3
+
4
+ 固化的行为契约(2026-10-24 第四十七轮):
5
+ 1. 约束能在日志里找到对应记录 → checked
6
+ 2. 找不到 → unchecked,且**不等于没违反**(I6)
7
+ 3. 编造的约束必须落 unchecked(不能因为日志里有别的词就误判)
8
+ 4. 无清单 / 空日志 → ok=None(无法判定,不默认通过)
9
+ 5. 返回必带 note,声明「有记录/无记录 ≠ 遵守/违反」
10
+
11
+ 为什么单独立测:这是 H7 唯一的可行验证路径(第四十六轮),
12
+ 若它把「没找到」当成「没违反」,结论会反向。
13
+ """
14
+
15
+ import os
16
+ import sys
17
+
18
+ sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
19
+
20
+ from md_cg import constraint_audit as cau # noqa: E402
21
+
22
+ FAILS = []
23
+
24
+
25
+ def check(name, cond, detail=""):
26
+ if cond:
27
+ print(" ok %s" % name)
28
+ else:
29
+ print(" FAIL %s %s" % (name, detail))
30
+ FAILS.append(name)
31
+
32
+
33
+ LOG = """abc1234 feat(cohesion): 拼装接缝检测
34
+ def5678 feat(constraint-watch): 约束巡检模块
35
+ ghi9012 docs: 补适用边界
36
+ """
37
+
38
+
39
+ def main():
40
+ items = cau.parse_constraints("C1 约束巡检\nC2 拼装接缝检测\n")
41
+
42
+ # ① 能找到
43
+ rep = cau.cross_check(items, LOG)
44
+ check("全部能找到 ok=True", rep["ok"] is True, rep["counts"])
45
+ check("checked 有证据", all(c.get("evidence") for c in rep["checked"]), rep["checked"])
46
+
47
+ # ② 编造的找不到
48
+ rep = cau.cross_check("C9 不得把火星数据写入木星表\n", LOG)
49
+ check("编造约束 unchecked", rep["ok"] is False, rep["counts"])
50
+ check("编造约束 cover 低",
51
+ rep["unchecked"][0]["cover"] < cau.MATCH_THRESHOLD, rep["unchecked"])
52
+
53
+ # ③ 空输入 → None
54
+ check("无清单 ok=None", cau.cross_check("", LOG)["ok"] is None)
55
+ check("空日志 ok=None", cau.cross_check(items, "")["ok"] is None)
56
+
57
+ # ④ 必带 I6 提示
58
+ rep = cau.cross_check(items, LOG)
59
+ check("note 含 I6 提示", "不等于" in rep["note"], rep["note"][:50])
60
+
61
+ # ⑤ 部分命中
62
+ rep = cau.cross_check("C1 约束巡检\nC9 火星木星\n", LOG)
63
+ check("部分命中 checked=1", rep["counts"]["checked"] == 1, rep["counts"])
64
+ check("部分命中 unchecked=1", rep["counts"]["unchecked"] == 1, rep["counts"])
65
+
66
+ # ⑥ 留痕缺口检查(2026-10-24 第四十九轮补)
67
+ import tempfile, subprocess, os
68
+ tg = cau.trace_gap(repo=os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
69
+ check("trace_gap 返回 ok 字段", "ok" in tg, tg)
70
+ check("trace_gap 有 counts", "counts" in tg or tg.get("ok") is None, tg)
71
+ check("trace_gap note 说明不等于违反",
72
+ "不等于违反" in (tg.get("note") or ""), (tg.get("note") or "")[:40])
73
+ # 非 git 目录 → ok=None(不默认通过)
74
+ with tempfile.TemporaryDirectory() as td:
75
+ tg2 = cau.trace_gap(repo=td)
76
+ check("非 git 目录 ok=None", tg2.get("ok") is None, tg2)
77
+
78
+ print("")
79
+ if FAILS:
80
+ print("失败 %d 项:%s" % (len(FAILS), FAILS))
81
+ return 1
82
+ print("全部通过")
83
+ return 0
84
+
85
+
86
+ if __name__ == "__main__":
87
+ raise SystemExit(main())
@@ -0,0 +1,108 @@
1
+ # -*- coding: utf-8 -*-
2
+ """md_cg.constraint_watch 的守卫测试。
3
+
4
+ 固化的行为契约(2026-10-24,上下文管理研究第二十二、二十三轮):
5
+ 1. 全部在场 → ok=True
6
+ 2. 部分/全部缺失 → ok=False,且缺的逐条列出
7
+ 3. 覆盖率 0 < c < 阈值 → 建议「回取」(有痕迹但不够)
8
+ 4. 覆盖率 0 → 建议「报警」(连痕迹都没有)
9
+ 5. 空文本 / 无约束 → ok=None(无法判定,**不得**默认通过)
10
+ 6. 清单解析只认带编号短行,说明文字不算约束
11
+
12
+ 为什么这些边界要立测:漏报就是静默丢约束(第三轮、第十二轮、第二十二轮
13
+ 三次同型失败都指向「没有强制就不做」),所以「报不出」比「误报」危险得多。
14
+ """
15
+
16
+ import os
17
+ import sys
18
+
19
+ sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
20
+
21
+ from md_cg import constraint_watch as cw # noqa: E402
22
+
23
+ FAILS = []
24
+
25
+
26
+ def check(name, cond, detail=""):
27
+ if cond:
28
+ print(" ok %s" % name)
29
+ else:
30
+ print(" FAIL %s %s" % (name, detail))
31
+ FAILS.append(name)
32
+
33
+
34
+ LIST = (
35
+ "# 约束清单\n"
36
+ "C1 不得覆盖旧文件\n"
37
+ "C2 必须留下校验记录\n"
38
+ "C3 不得删除中间产物\n"
39
+ )
40
+ FULL = "导出必须留下校验记录。不得覆盖旧文件,不得删除中间产物。"
41
+ PARTIAL = "本次操作会覆盖一些临时内容。"
42
+ EMPTY = "今天天气不错。"
43
+
44
+
45
+ def main():
46
+ # ① 全部在场
47
+ rep = cw.check(LIST, FULL)
48
+ check("全部在场 ok=True", rep["ok"] is True, rep.get("counts"))
49
+ check("全部在场 missing=0", len(rep["missing"]) == 0, rep["missing"])
50
+
51
+ # ② 部分缺失:逐条列出
52
+ rep = cw.check(LIST, PARTIAL)
53
+ check("部分缺失 ok=False", rep["ok"] is False, rep.get("counts"))
54
+ check("部分缺失 有回取建议", any(m["advice"] == "回取" for m in rep["missing"]),
55
+ rep["missing"])
56
+
57
+ # ③ 全部缺失 → 报警
58
+ rep = cw.check(LIST, EMPTY)
59
+ check("全部缺失 ok=False", rep["ok"] is False, rep.get("counts"))
60
+ check("全部缺失 全为报警", all(m["advice"] == "报警" for m in rep["missing"]),
61
+ rep["missing"])
62
+
63
+ # ④ 空文本 / 无约束 → None(不默认通过)
64
+ check("空文本 ok=None", cw.check(LIST, "")["ok"] is None)
65
+ check("无约束 ok=None", cw.check("没有编号条目", FULL)["ok"] is None)
66
+ check("空清单 ok=None", cw.check("", FULL)["ok"] is None)
67
+
68
+ # ⑤ 解析只认带编号短行
69
+ items = cw.parse_constraints(LIST)
70
+ check("解析出 3 条", len(items) == 3, items)
71
+ check("编号规范化", [i["id"] for i in items] == ["C1", "C2", "C3"], items)
72
+ long_line = "C9 " + "这是一条很长的说明文字" * 8
73
+ check("超长行不算约束", len(cw.parse_constraints(long_line)) == 0)
74
+
75
+ # ⑥ 覆盖率单调性
76
+ c_hi = cw.coverage("不得覆盖旧文件", FULL)
77
+ c_lo = cw.coverage("不得覆盖旧文件", EMPTY)
78
+ check("在场覆盖高于缺失", c_hi > c_lo, (c_hi, c_lo))
79
+
80
+ # ⑦ 回取三态(2026-10-24 第三十三轮补)
81
+ r = cw.retrieve_for("不得覆盖旧文件", None)
82
+ check("回取未接入 ok=None", r["ok"] is None, r)
83
+ r = cw.retrieve_for("", None)
84
+ check("回取空文本 ok=None", r["ok"] is None, r)
85
+
86
+ class FakeCG:
87
+ def __init__(self, hits):
88
+ self._hits = hits
89
+ def search(self, q, k=3):
90
+ return (self._hits, {})
91
+ r = cw.retrieve_for("不得覆盖旧文件", FakeCG([]))
92
+ check("回取无候选 ok=False", r["ok"] is False, r)
93
+ check("无候选提示含 I6", "不等于" in r["note"], r["note"])
94
+ r = cw.retrieve_for("不得覆盖旧文件",
95
+ FakeCG([{"id": "n1", "content": "候选内容"}]))
96
+ check("回取有候选 ok=True", r["ok"] is True, r)
97
+ check("有候选仍需人判断", "需判断" in r["note"], r["note"])
98
+
99
+ print("")
100
+ if FAILS:
101
+ print("失败 %d 项:%s" % (len(FAILS), FAILS))
102
+ return 1
103
+ print("全部通过")
104
+ return 0
105
+
106
+
107
+ if __name__ == "__main__":
108
+ raise SystemExit(main())
@@ -0,0 +1,110 @@
1
+ # -*- coding: utf-8 -*-
2
+ """md_cg.context_audit 的守卫测试。
3
+
4
+ 固化的行为契约(2026-10-24,框架 v5 第⑤步「三种损失各查一遍」):
5
+ 1. 三种损失各由一个读数覆盖:dropped / diluted / broken
6
+ 2. 判定按层级裁剪——同一份上下文,L2 可 PASS 而 L3 必须 FAIL
7
+ 3. 任一 required 项 False → FAIL
8
+ 4. 无 False 但有 None → UNKNOWN(**不默认通过,也不判坏**)
9
+ 5. L1/L2 不强制 broken,但读数偏低时给出「可升层」提示
10
+ 6. 缺材料 → UNKNOWN 而非 PASS
11
+
12
+ 第 2、5 条是实测修正的结果:L2 是「半连续」(步骤要连、章节可断),
13
+ 若强制接缝,会把正常的结构化上下文判为失败。
14
+ """
15
+
16
+ import os
17
+ import sys
18
+
19
+ sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
20
+
21
+ from md_cg import context_audit as ca # noqa: E402
22
+
23
+ FAILS = []
24
+
25
+
26
+ def check(name, cond, detail=""):
27
+ if cond:
28
+ print(" ok %s" % name)
29
+ else:
30
+ print(" FAIL %s %s" % (name, detail))
31
+ FAILS.append(name)
32
+
33
+
34
+ # 两块内容不同、但都含目标词的段落(避免依赖外部文件)
35
+ A = ("第一段讨论摘要压缩的机制。摘要保留率等于 1 除以平均段长。"
36
+ "这一结论由位置效应推出,段首存活 93.2%。" * 3)
37
+ B = ("第二段讨论上下文稀释。上下文稀释达 36.7 倍。"
38
+ "稀释不单调,因为同主题重复会抬高词频。" * 3)
39
+ LIST = "C1 摘要保留率等于 1 除以平均段长\nC2 上下文稀释达 36.7 倍\n"
40
+ TGT = "摘要保留率等于 1 除以平均段长,上下文稀释达 36.7 倍。"
41
+
42
+
43
+ def main():
44
+ # ① 三种损失都有读数
45
+ r = ca.audit("分析影响,必须给出可核对读数", blocks=[A, B], checklist=LIST, target_text=TGT)
46
+ check("三种损失都有读数",
47
+ set(r["losses"].keys()) == {"dropped", "diluted", "broken"}, r["losses"].keys())
48
+
49
+ # ② 层级裁剪:同一上下文 L2 与 L3 判定可不同
50
+ l2 = ca.audit("分析影响,必须给出可核对读数", blocks=[A, B], checklist=LIST, target_text=TGT)
51
+ l3 = ca.audit("导出数据,但不能覆盖旧文件", blocks=[A, B], checklist=LIST, target_text=TGT)
52
+ check("L2 判为 L2", l2["level"] == "L2", l2["level"])
53
+ check("L3 判为 L3", l3["level"] == "L3", l3["level"])
54
+ check("L2 不强制 broken", l2["required"]["broken"] is False, l2["required"])
55
+ check("L3 强制 broken", l3["required"]["broken"] is True, l3["required"])
56
+
57
+ # ③ 缺材料 → UNKNOWN(不默认通过)
58
+ r = ca.audit("导出数据,但不能覆盖旧文件")
59
+ check("缺材料 UNKNOWN", r["verdict"] == "UNKNOWN", r["verdict"])
60
+ check("缺材料 无 PASS", r["verdict"] != "PASS", r)
61
+
62
+ # ④ 约束缺失 → FAIL
63
+ bad = LIST + "C9 不得把火星数据写入木星表\n"
64
+ r = ca.audit("导出数据,但不能覆盖旧文件", blocks=[A, B], checklist=bad, target_text=TGT)
65
+ check("约束缺失 FAIL", r["verdict"] == "FAIL", r["verdict"])
66
+ check("约束缺失 列出 dropped", "dropped" in r["failed"], r["failed"])
67
+
68
+ # ⑤ L1 连 diluted 之外的 broken 也不要求
69
+ r = ca.audit("帮我查一下是哪个版本", blocks=[A, B], checklist=LIST, target_text=TGT)
70
+ check("L1 不强制 broken", r["required"]["broken"] is False, r["required"])
71
+
72
+ # ⑥ 只报不做:必带 note
73
+ r = ca.audit("分析影响", blocks=[A, B], checklist=LIST, target_text=TGT)
74
+ check("必带 note", bool(r["note"]) and "只报不做" in r["note"], r["note"][:40])
75
+
76
+ # ⑦ 可选回取(2026-10-24 第四十一轮补)
77
+ class FakeCG:
78
+ def search(self, q, k=2):
79
+ return ([{"id": "cand_1", "content": "候选出处内容"}], {})
80
+ bad = LIST + "C9 不得把火星数据写入木星表\n"
81
+ r_no = ca.audit("导出数据,但不能覆盖旧文件", blocks=[A, B], checklist=bad, target_text=TGT)
82
+ check("不传 cg 时无 sources",
83
+ all("sources" not in m for m in r_no["losses"]["dropped"].get("missing", [])),
84
+ r_no["losses"]["dropped"].get("missing"))
85
+ r_yes = ca.audit("导出数据,但不能覆盖旧文件", blocks=[A, B], checklist=bad,
86
+ target_text=TGT, cg=FakeCG())
87
+ missing = r_yes["losses"]["dropped"].get("missing", [])
88
+ check("传 cg 时补 sources",
89
+ any(m.get("sources") for m in missing), missing)
90
+ check("actions 含候选出处",
91
+ any("候选出处" in a for a in r_yes["actions"]), r_yes["actions"][:2])
92
+ check("note 声明只查不改", "只查不改" in r_yes["note"], r_yes["note"][:60])
93
+ # 限量:max_lookup 生效
94
+ many = LIST + "".join("C%d 不得写入木星表%d\n" % (i, i) for i in range(100, 110))
95
+ r_lim = ca.audit("导出数据,但不能覆盖旧文件", blocks=[A, B], checklist=many,
96
+ target_text=TGT, cg=FakeCG(), max_lookup=2)
97
+ withsrc = sum(1 for m in r_lim["losses"]["dropped"].get("missing", [])
98
+ if m.get("sources"))
99
+ check("max_lookup 限量生效", withsrc <= 2, withsrc)
100
+
101
+ print("")
102
+ if FAILS:
103
+ print("失败 %d 项:%s" % (len(FAILS), FAILS))
104
+ return 1
105
+ print("全部通过")
106
+ return 0
107
+
108
+
109
+ if __name__ == "__main__":
110
+ raise SystemExit(main())
@@ -0,0 +1,81 @@
1
+ # -*- coding: utf-8 -*-
2
+ """md_cg.context_plan 的守卫测试。
3
+
4
+ 固化的行为契约(2026-10-24,上下文管理研究框架 v4 五步判据链):
5
+ 1. 命中不可逆词 → L3(后果不可撤销,哪怕同时含强制词)
6
+ 2. 只命中强制词 → L2
7
+ 3. 查事实型且无强制/不可逆 → L1
8
+ 4. 都没有 → 灰区(**不按 L1 兜底**)
9
+ 5. 灰区的声明按 L3、策略按实际(解耦)
10
+ 6. 每层输出都带 caveat(这是建议不是裁决)
11
+
12
+ 第 4 条为什么重要:误判为高损失可逆有界,误判为低损失不可逆无界(研究第六轮)。
13
+ 所以「判不出」必须落到灰区/L2,绝不能落到 L1。
14
+ """
15
+
16
+ import os
17
+ import sys
18
+
19
+ sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
20
+
21
+ from md_cg import context_plan as cp # noqa: E402
22
+
23
+ FAILS = []
24
+
25
+
26
+ def check(name, cond, detail=""):
27
+ if cond:
28
+ print(" ok %s" % name)
29
+ else:
30
+ print(" FAIL %s %s" % (name, detail))
31
+ FAILS.append(name)
32
+
33
+
34
+ def main():
35
+ # ① 不可逆 → L3(即使同时有强制词)
36
+ r = cp.plan("导出数据,但不能覆盖旧文件,必须留下校验记录")
37
+ check("不可逆优先于强制 → L3", r["level"] == "L3", r["level_reason"])
38
+ check("L3 声明含外置清单", "外置" in r["declaration"], r["declaration"])
39
+ check("L3 核对不得静默", "不得静默" in r["check"], r["check"])
40
+
41
+ # ② 只强制 → L2
42
+ r = cp.plan("提交前必须跑一遍测试")
43
+ check("只强制 → L2", r["level"] == "L2", r["level_reason"])
44
+
45
+ # ③ 查事实 → L1
46
+ r = cp.plan("帮我查一下智能论是哪个版本")
47
+ check("查事实 → L1", r["level"] == "L1", r["level_reason"])
48
+
49
+ # ④ 判不出 → 灰区,不落 L1
50
+ r = cp.plan("把这份报告整理成三页")
51
+ check("判不出 → 灰区", r["level"] == "灰区", r["level_reason"])
52
+ check("灰区声明按 L3", "按 L3 声明" in r["declaration"], r["declaration"])
53
+ check("灰区策略按实际", "策略按实际" in r["strategy"], r["strategy"])
54
+
55
+ # ⑤ 各层形态互不相同(判据链第②步真的分流了)
56
+ forms = {lv: cp.plan(t)["form"] for lv, t in [
57
+ ("L1", "查一下版本号是多少"),
58
+ ("L2", "必须检查一遍"),
59
+ ("L3", "删除旧文件"),
60
+ ]}
61
+ check("三层形态不同", len(set(forms.values())) == 3, forms)
62
+
63
+ # ⑥ 每层都带 caveat
64
+ for t in ("查一下版本", "必须检查", "删除文件", "整理成三页"):
65
+ r = cp.plan(t)
66
+ check("caveat 存在(%s)" % r["level"], bool(r["caveat"]))
67
+
68
+ # ⑦ 空输入不炸,落灰区
69
+ r = cp.plan("")
70
+ check("空输入 → 灰区", r["level"] == "灰区", r["level"])
71
+
72
+ print("")
73
+ if FAILS:
74
+ print("失败 %d 项:%s" % (len(FAILS), FAILS))
75
+ return 1
76
+ print("全部通过")
77
+ return 0
78
+
79
+
80
+ if __name__ == "__main__":
81
+ raise SystemExit(main())