@furongjun1999/dsh-memory 0.4.11 → 0.5.0

package/md_cg/test_health_scale.py

@@ -1,173 +1,173 @@
-# -*- coding: utf-8 -*-
-"""体检读数规模纪律验证（O(1) 元数据口径，替代全量扫描）。
-
-背景（第 4 条取证）：审计日志无上界增长（实测 ~2 条/s、累计 ~21.5 M 条 / 4.0 GB），
-而健康度只需要「量级」这一 O(1) 信息。旧实现 `len(list(read_jsonl(p)))` 把只读体检
-绑到 O(n) 全量解析上（4.0 GB → RSS 5 GB+、数十分钟不返回、整条 MCP 通道被一次调用
-堵死）；改用流式数行仍要 O(n) 磁盘 IO，随运行时长线性劣化——同属「指标与代价错配」。
-
-断言面：
-  ① 缺失文件：零读数不炸（bytes=0 / events=0 / exact=True）
-  ② 阈值内：events 精确 == 日志行数（口径与物化读一致，改造不丢读数）
-  ③ 超阈值：只读元数据 + 尾部估算，exact=False，且**不触发任何全量计数**（O(n) 守卫）
-  ④ 估算忠于量级：与真实行数相对误差 < 20%（可用但不假装精确）
-  ⑤ health_os 集成：审计面走 _log_scale，**read_jsonl/count_jsonl 都不得再碰审计路径**
-     （防退回物化读/全量扫的核心回归守卫）
-  ⑥ 口径连续：小审计文件下 health_os 仍给出精确条数 + exact=True
-
-运行：python -m md_cg.test_health_scale
-"""
-from __future__ import annotations
-
-import json
-import os
-import shutil
-import tempfile
-
-from . import mdcos as mdcos_mod
-from .mdcos import MdCGOS
-
-PASS = FAIL = 0
-FAILS = []
-
-
-def ok(cond, label):
-    global PASS, FAIL
-    if cond:
-        PASS += 1
-    else:
-        FAIL += 1
-        FAILS.append(label)
-        print(f"  FAIL {label}")
-
-
-def _write_log(path, n, pad=100):
-    with open(path, "w", encoding="utf-8") as f:
-        for i in range(n):
-            f.write(json.dumps({"t": float(i), "op": "add", "i": i,
-                                "pad": "x" * pad}) + "\n")
-    return os.path.getsize(path)
-
-
-class _Spy:
-    """记录 mdcos 模块级 read_jsonl / count_jsonl 的调用路径（探针）。"""
-
-    def __enter__(self):
-        self.read, self.count = [], []
-        self._read, self._count = mdcos_mod.read_jsonl, mdcos_mod.count_jsonl
-        spy = self
-
-        def _r(p, *a, **k):
-            spy.read.append(p)
-            return spy._read(p, *a, **k)
-
-        def _c(p, *a, **k):
-            spy.count.append(p)
-            return spy._count(p, *a, **k)
-
-        mdcos_mod.read_jsonl, mdcos_mod.count_jsonl = _r, _c
-        return self
-
-    def __exit__(self, *exc):
-        mdcos_mod.read_jsonl, mdcos_mod.count_jsonl = self._read, self._count
-        return False
-
-
-def main():
-    tmp = tempfile.mkdtemp(prefix="mdcg_scale_")
-    try:
-        root = os.path.join(tmp, "root")
-        os.makedirs(root, exist_ok=True)
-        cg = MdCGOS(root)
-        audit = cg.audit_log
-
-        # ---------- ① 缺失文件 ----------
-        s = cg._log_scale(os.path.join(root, "_nope.jsonl"))
-        ok(s["bytes"] == 0 and s["events"] == 0 and s["exact"] is True,
-           "①缺失日志 → 零读数不炸（bytes=0/events=0/exact=True）")
-
-        # ---------- ② 阈值内精确 ----------
-        n = 200
-        size = _write_log(audit, n)
-        cg.AUDIT_COUNT_MAX_BYTES = size + 1
-        s = cg._log_scale(audit)
-        ok(s["exact"] is True and s["events"] == n and s["bytes"] == size,
-           "②规模在阈值内 → 精确条数（%d 行，与物化读口径一致）" % n)
-
-        # ---------- ③ 超阈值：O(1)，不触发全量计数 ----------
-        cg.AUDIT_COUNT_MAX_BYTES = 1024          # 人工压低阈值，等价于 GB 级日志
-        with _Spy() as spy:
-            s = cg._log_scale(audit)
-        ok(s["exact"] is False and s["bytes"] == size,
-           "③超阈值 → exact=False，bytes 仍为 O(1) 精确元数据")
-        ok(s["events"] is not None and s["events"] > 0,
-           "③b超阈值仍给读数（尾部采样估算），不是拒不回答")
-        ok(not spy.count and not spy.read,
-           "③c超阈值分支零全量调用（未 count_jsonl / 未 read_jsonl）——O(n) 守卫")
-
-        # ---------- ④ 估算忠于量级 ----------
-        rel = abs(s["events"] - n) / float(n)
-        ok(rel < 0.20,
-           "④估算与真实行数相对误差 %.1f%% < 20%%（量级可用，不假装精确）" % (rel * 100))
-
-        # ---------- ⑤ health_os 集成：审计路径零全量读 ----------
-        with _Spy() as spy:
-            h = cg.health_os()
-        o = h["os"]
-        ok(o["audit_events_exact"] is False and o["audit_bytes"] == size
-           and o["audit_events"] == s["events"],
-           "⑤health_os 透出审计量级（exact=False/bytes 精确/events=估算）")
-        ok(not any("_audit.jsonl" in p for p in spy.read),
-           "⑤bhealth_os 不再对审计日志做物化读（read_jsonl 未被以审计路径调用）")
-        ok(not any("_audit.jsonl" in p for p in spy.count),
-           "⑤chealth_os 不再对审计日志做全量计数（count_jsonl 未被以审计路径调用）")
-
-        # ---------- ⑥ 口径连续：小日志仍精确 ----------
-        cg.AUDIT_COUNT_MAX_BYTES = 64 << 20
-        root2 = os.path.join(tmp, "root2")
-        os.makedirs(root2, exist_ok=True)
-        cg2 = MdCGOS(root2)
-        _write_log(cg2.audit_log, 37)
-        o2 = cg2.health_os()["os"]
-        ok(o2["audit_events"] == 37 and o2["audit_events_exact"] is True,
-           "⑥小日志 → health_os 仍给精确条数（37 行，改造不丢读数）")
-
-        # ---------- ⑦ 真实写入风格：流式行数 == 物化条数（口径等价） ----------
-        import random
-
-        from .fsutil import append_jsonl, read_jsonl
-        real = os.path.join(root, "_real.jsonl")
-        for i in range(50):
-            append_jsonl(real, {"i": i, "pad": "y" * random.randint(1, 80)})
-        n_real = len(list(read_jsonl(real)))
-        s_real = cg._log_scale(real)
-        ok(s_real["events"] == n_real == 50 and s_real["exact"] is True,
-           "⑦完整写入日志：流式行数 %s == 物化条数 %d（口径逐位相等）"
-           % (s_real["events"], n_real))
-        os.remove(real)
-
-        # ---------- ⑧ 半截尾行边界：下界口径，最多差 1 行 ----------
-        bad = os.path.join(root, "_bad.jsonl")
-        with open(bad, "w", encoding="utf-8") as f:
-            f.write('{"ok": 1}\nnot-json\n{"ok": 2}')      # 末行未终止 + 非法 JSON
-        sb = cg._log_scale(bad)
-        n_material = len(list(read_jsonl(bad)))
-        ok(sb["events"] == 2 and sb["exact"] is True,
-           "⑧半截尾行不计入（换行数 2 vs 物化条数 %d）——下界口径，差 1 行在量级容差内"
-           % n_material)
-        os.remove(bad)
-
-        cg.close()
-        cg2.close()
-    finally:
-        shutil.rmtree(tmp, ignore_errors=True)
-
-    print(f"\nhealth_scale: {PASS} pass / {FAIL} fail")
-    if FAILS:
-        for f in FAILS:
-            print(f" - {f}")
-    raise SystemExit(1 if FAIL else 0)
-
-
-if __name__ == "__main__":
-    main()
+# -*- coding: utf-8 -*-
+"""体检读数规模纪律验证（O(1) 元数据口径，替代全量扫描）。
+
+背景（第 4 条取证）：审计日志无上界增长（实测 ~2 条/s、累计 ~21.5 M 条 / 4.0 GB），
+而健康度只需要「量级」这一 O(1) 信息。旧实现 `len(list(read_jsonl(p)))` 把只读体检
+绑到 O(n) 全量解析上（4.0 GB → RSS 5 GB+、数十分钟不返回、整条 MCP 通道被一次调用
+堵死）；改用流式数行仍要 O(n) 磁盘 IO，随运行时长线性劣化——同属「指标与代价错配」。
+
+断言面：
+  ① 缺失文件：零读数不炸（bytes=0 / events=0 / exact=True）
+  ② 阈值内：events 精确 == 日志行数（口径与物化读一致，改造不丢读数）
+  ③ 超阈值：只读元数据 + 尾部估算，exact=False，且**不触发任何全量计数**（O(n) 守卫）
+  ④ 估算忠于量级：与真实行数相对误差 < 20%（可用但不假装精确）
+  ⑤ health_os 集成：审计面走 _log_scale，**read_jsonl/count_jsonl 都不得再碰审计路径**
+     （防退回物化读/全量扫的核心回归守卫）
+  ⑥ 口径连续：小审计文件下 health_os 仍给出精确条数 + exact=True
+
+运行：python -m md_cg.test_health_scale
+"""
+from __future__ import annotations
+
+import json
+import os
+import shutil
+import tempfile
+
+from . import mdcos as mdcos_mod
+from .mdcos import MdCGOS
+
+PASS = FAIL = 0
+FAILS = []
+
+
+def ok(cond, label):
+    global PASS, FAIL
+    if cond:
+        PASS += 1
+    else:
+        FAIL += 1
+        FAILS.append(label)
+        print(f"  FAIL {label}")
+
+
+def _write_log(path, n, pad=100):
+    with open(path, "w", encoding="utf-8") as f:
+        for i in range(n):
+            f.write(json.dumps({"t": float(i), "op": "add", "i": i,
+                                "pad": "x" * pad}) + "\n")
+    return os.path.getsize(path)
+
+
+class _Spy:
+    """记录 mdcos 模块级 read_jsonl / count_jsonl 的调用路径（探针）。"""
+
+    def __enter__(self):
+        self.read, self.count = [], []
+        self._read, self._count = mdcos_mod.read_jsonl, mdcos_mod.count_jsonl
+        spy = self
+
+        def _r(p, *a, **k):
+            spy.read.append(p)
+            return spy._read(p, *a, **k)
+
+        def _c(p, *a, **k):
+            spy.count.append(p)
+            return spy._count(p, *a, **k)
+
+        mdcos_mod.read_jsonl, mdcos_mod.count_jsonl = _r, _c
+        return self
+
+    def __exit__(self, *exc):
+        mdcos_mod.read_jsonl, mdcos_mod.count_jsonl = self._read, self._count
+        return False
+
+
+def main():
+    tmp = tempfile.mkdtemp(prefix="mdcg_scale_")
+    try:
+        root = os.path.join(tmp, "root")
+        os.makedirs(root, exist_ok=True)
+        cg = MdCGOS(root)
+        audit = cg.audit_log
+
+        # ---------- ① 缺失文件 ----------
+        s = cg._log_scale(os.path.join(root, "_nope.jsonl"))
+        ok(s["bytes"] == 0 and s["events"] == 0 and s["exact"] is True,
+           "①缺失日志 → 零读数不炸（bytes=0/events=0/exact=True）")
+
+        # ---------- ② 阈值内精确 ----------
+        n = 200
+        size = _write_log(audit, n)
+        cg.AUDIT_COUNT_MAX_BYTES = size + 1
+        s = cg._log_scale(audit)
+        ok(s["exact"] is True and s["events"] == n and s["bytes"] == size,
+           "②规模在阈值内 → 精确条数（%d 行，与物化读口径一致）" % n)
+
+        # ---------- ③ 超阈值：O(1)，不触发全量计数 ----------
+        cg.AUDIT_COUNT_MAX_BYTES = 1024          # 人工压低阈值，等价于 GB 级日志
+        with _Spy() as spy:
+            s = cg._log_scale(audit)
+        ok(s["exact"] is False and s["bytes"] == size,
+           "③超阈值 → exact=False，bytes 仍为 O(1) 精确元数据")
+        ok(s["events"] is not None and s["events"] > 0,
+           "③b超阈值仍给读数（尾部采样估算），不是拒不回答")
+        ok(not spy.count and not spy.read,
+           "③c超阈值分支零全量调用（未 count_jsonl / 未 read_jsonl）——O(n) 守卫")
+
+        # ---------- ④ 估算忠于量级 ----------
+        rel = abs(s["events"] - n) / float(n)
+        ok(rel < 0.20,
+           "④估算与真实行数相对误差 %.1f%% < 20%%（量级可用，不假装精确）" % (rel * 100))
+
+        # ---------- ⑤ health_os 集成：审计路径零全量读 ----------
+        with _Spy() as spy:
+            h = cg.health_os()
+        o = h["os"]
+        ok(o["audit_events_exact"] is False and o["audit_bytes"] == size
+           and o["audit_events"] == s["events"],
+           "⑤health_os 透出审计量级（exact=False/bytes 精确/events=估算）")
+        ok(not any("_audit.jsonl" in p for p in spy.read),
+           "⑤bhealth_os 不再对审计日志做物化读（read_jsonl 未被以审计路径调用）")
+        ok(not any("_audit.jsonl" in p for p in spy.count),
+           "⑤chealth_os 不再对审计日志做全量计数（count_jsonl 未被以审计路径调用）")
+
+        # ---------- ⑥ 口径连续：小日志仍精确 ----------
+        cg.AUDIT_COUNT_MAX_BYTES = 64 << 20
+        root2 = os.path.join(tmp, "root2")
+        os.makedirs(root2, exist_ok=True)
+        cg2 = MdCGOS(root2)
+        _write_log(cg2.audit_log, 37)
+        o2 = cg2.health_os()["os"]
+        ok(o2["audit_events"] == 37 and o2["audit_events_exact"] is True,
+           "⑥小日志 → health_os 仍给精确条数（37 行，改造不丢读数）")
+
+        # ---------- ⑦ 真实写入风格：流式行数 == 物化条数（口径等价） ----------
+        import random
+
+        from .fsutil import append_jsonl, read_jsonl
+        real = os.path.join(root, "_real.jsonl")
+        for i in range(50):
+            append_jsonl(real, {"i": i, "pad": "y" * random.randint(1, 80)})
+        n_real = len(list(read_jsonl(real)))
+        s_real = cg._log_scale(real)
+        ok(s_real["events"] == n_real == 50 and s_real["exact"] is True,
+           "⑦完整写入日志：流式行数 %s == 物化条数 %d（口径逐位相等）"
+           % (s_real["events"], n_real))
+        os.remove(real)
+
+        # ---------- ⑧ 半截尾行边界：下界口径，最多差 1 行 ----------
+        bad = os.path.join(root, "_bad.jsonl")
+        with open(bad, "w", encoding="utf-8") as f:
+            f.write('{"ok": 1}\nnot-json\n{"ok": 2}')      # 末行未终止 + 非法 JSON
+        sb = cg._log_scale(bad)
+        n_material = len(list(read_jsonl(bad)))
+        ok(sb["events"] == 2 and sb["exact"] is True,
+           "⑧半截尾行不计入（换行数 2 vs 物化条数 %d）——下界口径，差 1 行在量级容差内"
+           % n_material)
+        os.remove(bad)
+
+        cg.close()
+        cg2.close()
+    finally:
+        shutil.rmtree(tmp, ignore_errors=True)
+
+    print(f"\nhealth_scale: {PASS} pass / {FAIL} fail")
+    if FAILS:
+        for f in FAILS:
+            print(f" - {f}")
+    raise SystemExit(1 if FAIL else 0)
+
+
+if __name__ == "__main__":
+    main()

package/md_cg/test_hive_ingest.py

@@ -0,0 +1,285 @@
+# -*- coding: utf-8 -*-
+"""M6 蜂巢 ingest 专项测试：HiveJobsSource 事件映射 / 终态补位 / watermark 幂等 /
+层归属纪律（§5.5 只落 contextual）/ fix-pair 产出（设计稿 §5.6 验收判据）。
+
+运行：python -m md_cg.test_hive_ingest   （退出码 0 = 全绿）
+"""
+from __future__ import annotations
+
+import json
+import os
+import shutil
+import sys
+import tempfile
+import time
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8", errors="replace")
+except AttributeError:
+    pass
+
+from .mdcos import MdCGOS
+from .sources import HiveJobsSource, Ingestor, run as sources_run
+
+PASS, FAIL = 0, 0
+
+
+def check(name, cond, detail=""):
+    global PASS, FAIL
+    if cond:
+        PASS += 1
+        print(f"  [ok] {name}")
+    else:
+        FAIL += 1
+        print(f"  [FAIL] {name}  {detail}")
+
+
+def make_jobs(root: str):
+    """构造设计稿 §5.6 验收场景：job1 强杀（无 final，result=error）、
+    job2 重试成功（final + result=done，content 含命令行）、job3 普通 done。"""
+    jobs = os.path.join(root, "jobs")
+    t = time.time()
+    # job1：强杀——progress 有 start/error，无 final；result=error
+    j1 = os.path.join(jobs, "h1000000000000_aaa1")
+    os.makedirs(j1)
+    json.dump({"model": "cmd", "user_prompt": "跑一次会失败的检查"},
+              open(os.path.join(j1, "spec.json"), "w", encoding="utf-8"),
+              ensure_ascii=False)
+    with open(os.path.join(j1, "progress.jsonl"), "w", encoding="utf-8") as f:
+        f.write(json.dumps({"ts": t - 60, "kind": "start",
+                            "task": "跑一次会失败的检查"}, ensure_ascii=False) + "\n")
+        f.write(json.dumps({"ts": t - 30, "kind": "error",
+                            "error": "单步超时（5s）被强杀：pytest"}, ensure_ascii=False) + "\n")
+    json.dump({"ok": False, "error": "单步超时（5s）被强杀：pytest",
+               "finished_ts": t - 29},
+              open(os.path.join(j1, "result.json"), "w", encoding="utf-8"),
+              ensure_ascii=False)
+    # job2：重试成功——final + result=done，content 是命令输出（_FIX_RE 命中）
+    j2 = os.path.join(jobs, "h1000000001000_bbb2")
+    os.makedirs(j2)
+    json.dump({"model": "cmd", "user_prompt": "跑一次会失败的检查"},
+              open(os.path.join(j2, "spec.json"), "w", encoding="utf-8"),
+              ensure_ascii=False)
+    with open(os.path.join(j2, "progress.jsonl"), "w", encoding="utf-8") as f:
+        f.write(json.dumps({"ts": t - 10, "kind": "start",
+                            "task": "跑一次会失败的检查"}, ensure_ascii=False) + "\n")
+        f.write(json.dumps({"ts": t - 5, "kind": "final",
+                            "content_head": "pytest 全部通过"}, ensure_ascii=False) + "\n")
+    json.dump({"ok": True, "content": "python hive/test_x.py\n全部通过",
+               "finished_ts": t - 4},
+              open(os.path.join(j2, "result.json"), "w", encoding="utf-8"),
+              ensure_ascii=False)
+    # job3：普通 done（带 handoff，验证续跑卡映射）
+    j3 = os.path.join(jobs, "h1000000002000_ccc3")
+    os.makedirs(j3)
+    json.dump({"model": "deepseek-flash", "user_prompt": "普通任务"},
+              open(os.path.join(j3, "spec.json"), "w", encoding="utf-8"),
+              ensure_ascii=False)
+    with open(os.path.join(j3, "progress.jsonl"), "w", encoding="utf-8") as f:
+        f.write(json.dumps({"ts": t - 3, "kind": "start", "task": "普通任务"},
+                           ensure_ascii=False) + "\n")
+        f.write(json.dumps({"ts": t - 2, "kind": "handoff",
+                            "summary": "上下文将满，交回续跑"}, ensure_ascii=False) + "\n")
+    json.dump({"ok": True, "content": "普通完成", "finished_ts": t - 1},
+              open(os.path.join(j3, "result.json"), "w", encoding="utf-8"),
+              ensure_ascii=False)
+    # job4：无 finished_ts（批次8 实测缺陷——exec_cmd 旧版形态）→ mtime 兜底
+    j4 = os.path.join(jobs, "h1000000003000_ddd4")
+    os.makedirs(j4, exist_ok=True)
+    json.dump({"model": "cmd", "user_prompt": "legacy"}, open(
+        os.path.join(j4, "spec.json"), "w", encoding="utf-8"))
+    rp4 = os.path.join(j4, "result.json")
+    json.dump({"ok": True, "content": "旧版产物", "steps": [
+        {"command": ["python", "-c", "print('legacy-ok')"]}]},
+        open(rp4, "w", encoding="utf-8"), ensure_ascii=False)
+    os.utime(rp4, (t - 0.5, t - 0.5))     # mtime 锚定到事件时间线内
+    return jobs
+
+
+def main():
+    root = tempfile.mkdtemp(prefix="mdcg_m6_")
+    try:
+        jobs = make_jobs(root)
+        cg = MdCGOS(root)
+        src = HiveJobsSource(jobs)
+        evs = list(src.events())
+
+        # ---- 1. 事件映射（§5.3） ----
+        print("[1] 事件映射")
+        check("1a start → 任务开始（role=user）",
+              any(e["role"] == "user" and "任务开始" in e["text"] for e in evs))
+        check("1b tool 条目被跳过（防流水账）",
+              all("tool_trace" not in e["text"] for e in evs))
+        check("1c 强杀 job 的 error 事件带「任务失败」前缀（fix-pair 前件）",
+              any("任务失败" in e["text"] and "aaa1" in e["text"] for e in evs))
+        check("1d handoff → 续跑卡",
+              any("续跑卡" in e["text"] for e in evs))
+        check("1e session=hive:<job_id>",
+              all(e["session"].startswith("hive:h") for e in evs), str(evs[0]["session"]))
+
+        # ---- 2. 终态补位（§5.6 判据 3） ----
+        print("[2] result 终态补位")
+        check("2a 强杀 job（progress 无 final）→ result 终态补位事件存在",
+              any("aaa1" in e["text"] and "任务失败" in e["text"] for e in evs))
+        check("2b 补位事件标注 progress 缺 final",
+              any("result 终态补位" in e["text"] for e in evs))
+
+        # ---- 3. ingest：层归属（§5.5）+ 产出 ----
+        print("[3] ingest 层归属（只落 contextual，knowledge 零污染）")
+        k_before = len([n for n, e in (cg.index.get("nodes") or {}).items()
+                        if (e or {}).get("layer") == "knowledge"])
+        ing = Ingestor(cg, layer="contextual", sensitivity="internal")
+        # §5.5 硬纪律（与 M3.2 双轨制同构）：ingest 自动 mine_fix_pairs 会把
+        # 挖掘产物直写 knowledge 层（实测 0→2 污染）——M6 路径默认关闭；
+        # fix-pair 挖掘保留为显式能力（挖掘产物入 knowledge 须走收口）
+        rep = ing.ingest(src, mine_fix_pairs=False)
+        check("3a 事件全部写入 contextual", rep["written"] == len(evs),
+              f"written={rep['written']} evs={len(evs)}")
+        check("3a2 无 finished_ts 的 result 经 mtime 兜底摄入（批次8 缺陷修复）",
+              any("legacy-ok" in (x.get("text") or "") for x in evs),
+              str([x.get("text", "")[:60] for x in evs if "legacy" in x.get("text", "")]))
+        k_after = len([n for n, e in (cg.index.get("nodes") or {}).items()
+                       if (e or {}).get("layer") == "knowledge"])
+        check("3b knowledge 层节点数不变（§5.5 反向对照）", k_before == k_after,
+              f"{k_before}→{k_after}")
+
+        # ---- 4. watermark 幂等（§5.6 判据 2） ----
+        print("[4] watermark 幂等")
+        rep2 = ing.ingest(HiveJobsSource(jobs))
+        check("4a 重跑零新增（断点续跑无重复）",
+              rep2["new_events"] == 0 and rep2["written"] == 0, str(rep2)[:150])
+
+        # ---- 5. fix-pair（§5.6 判据 1：错误→修复，显式能力验证） ----
+        print("[5] fix-pair（显式挖掘能力；M6 自动路径已按 §5.5 关闭）")
+        rep3 = ing.ingest(HiveJobsSource(jobs), mine_fix_pairs=False)
+        check("5a 幂等重跑下不再产出事件（fix-pair 无重复输入）",
+              rep3["new_events"] == 0)
+        # 事件流支持配对挖掘（错误后 4 条内出现命令行 → 配对）——能力验证：
+        # 实际启用须走收口（编排者/设计者裁决后显式调用，产物才可入 knowledge）
+        seq = [{"role": e["role"], "text": e["text"]} for e in evs]
+        fp = cg.mine_fix_pairs(seq)
+        check("5b 错误→修复对可被挖出（超时失败→重试通过）",
+              len(fp.get("pairs") or []) >= 1, str(fp)[:200])
+
+        # ---- 6. MCP/CLI 暴露口（sources.run action=hive）+ 事件级密级（双态） ----
+        print("[6] 暴露口 sources.run(action=hive) + error 事件 private（双态）")
+        root2 = tempfile.mkdtemp(prefix="mdcg_m6b_")
+        try:
+            jobs2 = make_jobs(root2)
+            # 6a-6c：高 clearance（secret）调用方 → error 事件正常落 private
+            from .mdcos import MdCGSecure
+            from .security import Principal
+            cg_sec = MdCGSecure(root2, principal=Principal(
+                actor="t_secret", clearance="secret", can_write=True,
+                role="designer", auth_mode="test"))
+            rep_h = sources_run(cg_sec, action="hive", path=jobs2)
+            check("6a run(action=hive) 摄取成功（统一暴露口）",
+                  rep_h.get("written", 0) > 0 and rep_h.get("source", "").startswith(
+                      "hive_jobs:"), str(rep_h)[:150])
+            priv = inter = 0
+            for nid, e in (cg_sec.index.get("nodes") or {}).items():
+                if not str(nid).startswith("src_"):
+                    continue
+                fm = e or {}
+                if fm.get("sensitivity") == "private":
+                    priv += 1
+                elif fm.get("layer") == "contextual":
+                    inter += 1
+            check("6b error 事件节点=private（secret clearance 落盘成功）",
+                  priv >= 2, f"private={priv}")
+            check("6c 非 error 事件保持 internal",
+                  inter >= 3, f"internal={inter}")
+            # 6d-6e：internal clearance 调用方 → private 写入 AccessDenied，
+            # denied 明细可见（不静默丢失，误差归因原料可查可重放）
+            cg_int = MdCGSecure(tempfile.mkdtemp(prefix="mdcg_m6c_"))
+            rep_i = sources_run(cg_int, action="hive", path=jobs2)
+            check("6d internal clearance 写 private 被拒且明细可见",
+                  rep_i.get("denied", 0) >= 2
+                  and len(rep_i.get("denied_events") or []) == rep_i["denied"],
+                  str(rep_i.get("denied_events"))[:160])
+            # 6f：显式降级 error_sensitivity=internal → 全部落盘（显式权衡）
+            cg_low = MdCGSecure(tempfile.mkdtemp(prefix="mdcg_m6d_"))
+            rep_l = sources_run(cg_low, action="hive", path=jobs2,
+                                error_sensitivity="internal")
+            check("6f 显式降级 error_sensitivity=internal 全部落盘",
+                  rep_l.get("denied", 0) == 0 and rep_l.get("written", 0) > 0,
+                  str({k: rep_l.get(k) for k in ("written", "denied")}))
+            # fail-closed：无 path 无 env → 人话指引
+            os.environ.pop("MDCG_HIVE_JOBS", None)
+            rep_nc = sources_run(cg_low, action="hive")
+            check("6g 无 root 时 fail-closed 人话指引（不猜布局）",
+                  rep_nc.get("ok") is False and "MDCG_HIVE_JOBS" in rep_nc["error"])
+            # dry_run：只统计不写盘
+            rep_d = sources_run(cg_low, action="hive", path=jobs2, dry_run=True)
+            check("6h dry_run 预演不写盘",
+                  rep_d.get("dry_run") is True and rep_d.get("written", 1) == 0)
+        finally:
+            shutil.rmtree(root2, ignore_errors=True)
+
+        # ---- 7. §5.5 系统纪律化（zcode 外评 break#2/P1）：默认不挖矿 + 产物入队 ----
+        print("[7] mine_fix_pairs 默认翻转 + 自动产物走 propose 队列")
+        root3 = tempfile.mkdtemp(prefix="mdcg_m7_")
+        try:
+            # 造一个含「错误→修复」对的 jsonl 会话文件
+            jl = os.path.join(root3, "sess.jsonl")
+            with open(jl, "w", encoding="utf-8") as f:
+                f.write(json.dumps({"role": "assistant",
+                                    "text": "执行失败：ModuleNotFoundError: no x"}) + "\n")
+                # 修复行必须是行首命令（_FIX_RE 行首启发式）
+                f.write(json.dumps({"role": "assistant",
+                                    "text": "执行修复：\npip install x"}) + "\n")
+            cg3 = MdCGOS(root3)
+            kn_before = sum(1 for e in (cg3.index.get("nodes") or {}).values()
+                            if (e or {}).get("layer") == "knowledge")
+            # 7a：不传 mine_fix_pairs（吃新默认 False）→ 不挖矿
+            rep_a = sources_run(cg3, action="jsonl", path=jl)
+            check("7a 默认不挖矿（file/jsonl 不再直写知识层——旧行为即红）",
+                  "fix_pairs" not in rep_a, str(rep_a.get("fix_pairs"))[:120])
+            kn_after = sum(1 for e in (cg3.index.get("nodes") or {}).values()
+                           if (e or {}).get("layer") == "knowledge")
+            check("7b 默认路径 knowledge 层零新增（污染反向对照）",
+                  kn_after == kn_before, f"{kn_before} -> {kn_after}")
+            # 7c：显式开启 → 自动产物走 propose 队列（不直写）
+            from .sources import JsonlSource
+            rep_c = Ingestor(cg3).ingest(JsonlSource(jl), mine_fix_pairs=True,
+                                         dry_run=True)
+            # dry_run 不挖矿；改真实写（换新 root 避免水位吃掉事件）
+            cg4 = MdCGOS(tempfile.mkdtemp(prefix="mdcg_m7b_"))
+            rep_c = Ingestor(cg4).ingest(JsonlSource(jl), mine_fix_pairs=True)
+            fp = rep_c.get("fix_pairs") or {}
+            check("7c 显式开启时自动产物走提案（as_proposals=True，pids 非空）",
+                  fp.get("as_proposals") is True and len(fp.get("pids") or []) >= 1
+                  and not fp.get("knowledge_ids"), str(fp)[:160])
+            kn_c = sum(1 for e in (cg4.index.get("nodes") or {}).values()
+                       if (e or {}).get("layer") == "knowledge")
+            check("7d 提案模式下 knowledge 层仍零新增（收口前不入层）",
+                  kn_c == 0, f"knowledge={kn_c}")
+            check("7e 提案入审核队列（review_list 可见）",
+                  len(cg4.review_list() or []) >= 1,
+                  str(cg4.review_list())[:120])
+            # 7f：显式工具语义直调（不带 as_proposals）→ 保持直写
+            r_tools = cg4.mine_fix_pairs([
+                {"error": "ModuleNotFoundError: no y", "fix": "pip install y"}])
+            check("7f 显式调用保持直写（knowledge_ids 非空）",
+                  len(r_tools.get("knowledge_ids") or []) == 1, str(r_tools)[:120])
+            # 7g：payload_hash 幂等——同内容再入队返回原 pid
+            r2 = cg4.mine_fix_pairs([
+                {"error": "ModuleNotFoundError: no x", "fix": "pip install x"}],
+                as_proposals=True)
+            r3 = cg4.mine_fix_pairs([
+                {"error": "ModuleNotFoundError: no x", "fix": "pip install x"}],
+                as_proposals=True)
+            check("7g 同内容重试幂等（返回原 pid 不重复入队）",
+                  r2["pids"] == r3["pids"] and len(r2["pids"]) >= 1,
+                  f"{r2['pids']} vs {r3['pids']}")
+        finally:
+            shutil.rmtree(root3, ignore_errors=True)
+    finally:
+        shutil.rmtree(root, ignore_errors=True)
+
+    print(f"\n=== hive ingest tests: {PASS} passed, {FAIL} failed ===")
+    sys.exit(1 if FAIL else 0)
+
+
+if __name__ == "__main__":
+    sys.exit(main())