merge: integrate ZSY customer service and profile capabilities

tests/unit/tools/test_foundation_migration_preflight.py

@@ -0,0 +1,95 @@
+"""验证迁移前状态证据采集器只读取 Git 元数据。"""
+
+# 导入测试所需的标准库 JSON、路径和子进程结果类型。
+import json
+from pathlib import Path
+from subprocess import CompletedProcess
+
+# 导入 pytest 补丁类型以及待验证的预检工具接口。
+from pytest import MonkeyPatch
+
+from tools.foundation_migration_preflight import (
+    collect_workspace_state,
+    run_git,
+    write_preflight_report,
+)
+
+
+# 验证未跟踪客服文件会被记录，且采集结果不含环境变量名称或值。
+def test_collect_workspace_state_records_untracked_paths_without_environment_values(
+    tmp_path: Path,
+) -> None:
+    # 构造确定性的 Git 命令替身，不调用真实 Git 或环境变量。
+    def fake_git_runner(_worktree: Path, *args: str) -> str:
+        # 为分支查询返回客服功能分支名称。
+        if args == ("branch", "--show-current"):
+            return "feature/customer-service-rag\n"
+        # 为提交查询返回固定的非敏感提交标识。
+        if args == ("rev-parse", "HEAD"):
+            return "abc123\n"
+        # 为状态查询返回一个未跟踪客服文件。
+        if args == ("status", "--short"):
+            return "?? app/service/agent/customer_service_agent.py\n"
+        # 防止测试静默接受未定义的 Git 查询。
+        raise AssertionError(f"unexpected git arguments: {args}")
+
+    # 使用替身采集临时工作区的只读状态。
+    state = collect_workspace_state(tmp_path, runner=fake_git_runner)
+
+    # 断言采集到预期的分支名称。
+    assert state["branch"] == "feature/customer-service-rag"
+    # 断言未跟踪客服文件完整保留在状态清单中。
+    assert state["status"] == ["?? app/service/agent/customer_service_agent.py"]
+    # 断言序列化结果不包含任何环境变量敏感字段。
+    assert "MYSQL_PASSWORD" not in json.dumps(state)
+
+
+# 验证报告写入器只输出传入的非敏感 Git 状态结构。
+def test_write_preflight_report_persists_utf8_json(tmp_path: Path) -> None:
+    # 指定临时报告文件，避免写入任何真实工作目录。
+    report_path = tmp_path / "preflight.json"
+    # 构造只含安全 Git 元数据的状态条目。
+    states = [{"path": "D:/workspace", "branch": "develop", "head": "abc123", "status": []}]
+
+    # 写入迁移前报告。
+    write_preflight_report(report_path, states)
+
+    # 以 UTF-8 读取并解析报告正文。
+    report = json.loads(report_path.read_text(encoding="utf-8"))
+    # 断言报告保留原始状态条目。
+    assert report == states
+
+
+# 验证每次 Git 查询只信任当前显式工作区，而不写入全局 Git 配置。
+def test_run_git_scopes_safe_directory_to_the_requested_worktree(
+    tmp_path: Path,
+    monkeypatch: MonkeyPatch,
+) -> None:
+    # 保存被测函数交给子进程层的命令参数。
+    captured_commands: list[list[str]] = []
+
+    # 构造返回固定分支名的无副作用子进程替身。
+    def fake_run(command: list[str], **_kwargs: object) -> CompletedProcess[str]:
+        # 记录命令以便后续断言安全目录范围。
+        captured_commands.append(command)
+        # 返回模拟的 Git 成功结果。
+        return CompletedProcess(command, 0, "develop\n", "")
+
+    # 让测试不执行真实 Git。
+    monkeypatch.setattr("tools.foundation_migration_preflight.subprocess.run", fake_run)
+
+    # 执行一个受限 Git 分支查询。
+    output = run_git(tmp_path, "branch", "--show-current")
+
+    # 断言调用仍返回 Git 输出。
+    assert output == "develop\n"
+    # 断言命令仅为该临时工作区附加安全目录。
+    assert captured_commands == [[
+        "git",
+        "-c",
+        f"safe.directory={tmp_path.resolve().as_posix()}",
+        "-C",
+        str(tmp_path.resolve()),
+        "branch",
+        "--show-current",
+    ]]

tests/unit/tools/test_knowledge_import_preflight.py

@@ -0,0 +1,75 @@
+import json
+
+import pytest
+
+from tools.knowledge_import_preflight import build_import_manifest
+
+
+def vector_candidate() -> dict[str, object]:
+    return {
+        "qa_id": "RAG-PUB-001",
+        "title": "基金交易确认时间说明",
+        "question": "基金什么时候确认？",
+        "paraphrases": ["申购何时确认", "赎回多久确认"],
+        "answer": "交易确认时间以产品规则和实际交易日为准。",
+        "scope": "public",
+        "intent": "policy_explain",
+        "collection": "fin_policy_collection",
+        "execution_mode": "vector_search",
+        "retrieval_status": "approved_candidate",
+        "audience": ["visitor", "authenticated_user"],
+        "agent_data_access": "none",
+        "tags": ["交易规则", "确认"],
+        "phase": "phase_1",
+        "source_type": "qa_pair",
+        "source_file": "qa-v5.8.txt",
+        "source_url": None,
+        "source_version": "v5.8",
+        "review_status": "approved_candidate",
+        "status": "active",
+        "effective_date": None,
+        "expire_date": None,
+    }
+
+
+def rule_only_record() -> dict[str, object]:
+    return {
+        "qa_id": "RAG-SEC-001",
+        "scope": "security_notice",
+        "collection": None,
+        "execution_mode": "fixed_route",
+        "retrieval_status": "rule_only",
+    }
+
+
+def test_manifest_contains_only_eligible_public_records_pending_admin_review() -> None:
+    manifest = build_import_manifest(
+        [vector_candidate(), rule_only_record()], source_name="qa-v5.8.jsonl"
+    )
+
+    assert manifest["summary"] == {
+        "total_records": 2,
+        "eligible_records": 1,
+        "excluded_rule_records": 1,
+        "publication_state": "pending_review",
+    }
+    entry = manifest["records"][0]
+    assert entry["qa_id"] == "RAG-PUB-001"
+    assert entry["knowledge_type"] == "policy_explain"
+    assert entry["milvus_collection"] == "fin_policy_collection"
+    assert entry["review_status"] == "pending_review"
+    assert entry["status"] == "active"
+    assert entry["retrieval_text"] == (
+        "标准问题：基金什么时候确认？\n"
+        "相似问法：申购何时确认；赎回多久确认\n"
+        "标签：交易规则、确认"
+    )
+    assert json.loads(entry["content_text"])["answer"] == "交易确认时间以产品规则和实际交易日为准。"
+
+
+def test_invalid_public_record_is_rejected_instead_of_silently_entering_manifest() -> None:
+    invalid = vector_candidate()
+    invalid["agent_data_access"] = "account"
+
+    with pytest.raises(ValueError, match="RAG-PUB-001"):
+        build_import_manifest([invalid], source_name="qa-v5.8.jsonl")

tests/unit/tools/test_publish_customer_service_knowledge.py

@@ -0,0 +1,40 @@
+import pytest
+
+from tools.publish_customer_service_knowledge import load_pending_manifest
+
+
+def manifest() -> dict[str, object]:
+    return {
+        "summary": {"eligible_records": 1, "publication_state": "pending_review"},
+        "records": [{
+            "qa_id": "FAQ-001",
+            "milvus_collection": "fin_faq_collection",
+            "retrieval_text": "标准问题：基金开户",
+            "title": "基金开户",
+            "snippet": "基金开户",
+            "tags": ["开户"],
+            "version": "v5.8",
+            "content_text": "{\"answer\":\"请在官方页面开户。\"}",
+            "source_file": "qa-v5.8.txt",
+            "effective_date": None,
+            "expire_date": None,
+            "review_status": "pending_review",
+            "status": "active",
+        }],
+    }
+
+
+def test_pending_manifest_is_converted_to_a_publishable_administrator_payload() -> None:
+    records = load_pending_manifest(manifest())
+
+    assert len(records) == 1
+    assert records[0].qa_id == "FAQ-001"
+    assert records[0].metadata["content_text"] == "{\"answer\":\"请在官方页面开户。\"}"
+
+
+def test_manifest_that_claims_to_be_published_is_rejected() -> None:
+    invalid = manifest()
+    invalid["summary"] = {"eligible_records": 1, "publication_state": "published"}
+
+    with pytest.raises(ValueError, match="pending_review"):
+        load_pending_manifest(invalid)