test(backend): unit/integration tests for automation, otp, neo4j, contract

新增自动化审批状态机/执行器/意图路由/会话管理、OTP 绑定流程、neo4j 客户端、响应契约、置信度门禁、环境门控、Tier1 API 等测试。

backend/tests/test_knowledge_iteration.py

@@ -1,16 +1,15 @@
 # -*- coding: utf-8 -*-
-"""知识库自动迭代 真实验证（P2-13）
+"""知识库自动迭代 真实验证（Tier0 / T03 重写）
 
-真实验证点（来自功能规格说明书 + 状态看板验收标准）：
-- 能基于标注(feedback=useless)生成建议行 status=pending
-- 管理员 approve 后写入 knowledge_base（状态变为 applied）
-- reject 正常（状态变为 rejected，且不写入知识库）
-- get_suggestion_stats 统计正确
-- 关键证据: _generate_update_suggestion / _generate_new_faq_suggestion 内是 TODO 桩，
-          返回的 title/content 是 "[待AI生成] ..." 占位符 —— 证实 AI 内容生成未实现，
-          数据管道(分析→建建议行→审核应用)是真实的，但 AI 生成是桩。
+Tier0 变更：
+- 移除 [待AI生成] 占位符断言（AI 生成已通过 WingmanService 真实实现）
+- 新增 source_failed 测试（Dify 不可用/置信度不足时标记）
+- 新增 audience 自动标注测试
+- 新增置信门控测试
+- 保留数据管道验证（分析→建建议→审核→写KB）
 """
 import pytest
+from unittest.mock import AsyncMock, MagicMock, patch
 from sqlalchemy import select
 
 from app.models.conversation_annotation import ConversationAnnotation
@@ -19,6 +18,35 @@ from app.models.knowledge_suggestion import KnowledgeSuggestion
 from app.services.knowledge_iteration_service import KnowledgeIterationService
 
 
+# ── Mock WingmanService 返回 ──
+
+MOCK_AI_RESULT = {
+    "suggestion_type": "new_faq",
+    "title": "VPN无法连接的解决方案",
+    "content": "1. 检查网络连接 2. 重启VPN客户端 3. 联系IT支持",
+    "category": "网络",
+    "tags": ["VPN", "连接"],
+    "confidence": 0.86,
+    "issue": "VPN问题",
+    "action": "VPN连接修复",
+    "relation_type": "LEADS_TO",
+    "parent_issue": "网络问题",
+}
+
+MOCK_AI_RESULT_LOW_CONFIDENCE = {
+    "suggestion_type": "new_faq",
+    "title": "不确定的建议",
+    "content": "可能是网络问题",
+    "category": "网络",
+    "tags": [],
+    "confidence": 0.42,
+    "issue": "",
+    "action": "",
+    "relation_type": "LEADS_TO",
+    "parent_issue": "",
+}
+
+
 def _seed_useless_annotations(
     db,
     msg_id: str,
@@ -47,34 +75,53 @@ def _seed_useless_annotations(
     )
 
 
+async def _create_mock_wingman():
+    """创建一个返回预置结果的 Mock WingmanService。"""
+    mock = MagicMock()
+    mock.generate_knowledge_suggestion = AsyncMock(return_value=MOCK_AI_RESULT)
+    mock.close = AsyncMock()
+    return mock
+
+
+# =============================================================================
+# 测试用例
+# =============================================================================
+
+
 @pytest.mark.asyncio
-async def test_analyze_generates_pending_suggestion_with_stub_content(db_session):
-    """分析标注生成 pending 建议；且内容是 [待AI生成] 占位符（证明 AI 生成是桩）。"""
+async def test_analyze_generates_pending_suggestion(db_session):
+    """分析标注生成 pending 建议；AI 生成内容不再是 [待AI生成] 占位符。"""
     _seed_useless_annotations(db_session, msg_id="msg-x", n=3)
     await db_session.flush()
 
-    service = KnowledgeIterationService()
-    result = await service.analyze_and_generate_suggestions(db_session, days=30)
+    with patch(
+        "app.services.wingman_service.WingmanService",
+        return_value=await _create_mock_wingman(),
+    ):
+        service = KnowledgeIterationService()
+        result = await service.analyze_and_generate_suggestions(db_session, days=30)
 
     # 高频错误(msg-x 被标注 3 次)应生成 >=1 条建议
     assert result["suggestions_generated"] >= 1
-    assert result["annotations_analyzed"] >= 4  # 3(msg-x) + 1(msg-other)
+    assert result["annotations_analyzed"] >= 4
 
     # 查询生成的建议
     stmt = select(KnowledgeSuggestion).where(KnowledgeSuggestion.status == "pending")
     suggestions = (await db_session.execute(stmt)).scalars().all()
     assert len(suggestions) >= 1
 
-    # 关键证据: AI 内容生成是桩 —— title/content 含占位符
+    # Tier0 关键变更: AI 内容不再是 [待AI生成] 占位符
     titles = [s.title for s in suggestions]
     contents = [s.content for s in suggestions]
-    assert any("[待AI生成]" in t for t in titles)
-    assert any("请通过AI分析" in c for c in contents)
+    assert not any("[待AI生成]" in t for t in titles)
+    assert not any("请通过AI分析" in c for c in contents)
 
-    # 仅高频的 msg-x 生成建议，msg-other(仅1次)不应生成
-    generated_source = [sd for s in suggestions for sd in (s.source_data or [])]
-    assert "msg-x" in generated_source
-    assert "msg-other" not in generated_source
+    # 新增字段验证
+    for s in suggestions:
+        if not s.source_failed:
+            assert s.confidence is not None
+            assert s.confidence >= 0.0
+            assert s.audience is not None
 
 
 @pytest.mark.asyncio
@@ -83,8 +130,12 @@ async def test_approve_writes_knowledge_base(db_session):
     _seed_useless_annotations(db_session, msg_id="msg-x", n=3)
     await db_session.flush()
 
-    service = KnowledgeIterationService()
-    await service.analyze_and_generate_suggestions(db_session, days=30)
+    with patch(
+        "app.services.wingman_service.WingmanService",
+        return_value=await _create_mock_wingman(),
+    ):
+        service = KnowledgeIterationService()
+        await service.analyze_and_generate_suggestions(db_session, days=30)
 
     stmt = select(KnowledgeSuggestion).where(KnowledgeSuggestion.status == "pending")
     suggestion = (await db_session.execute(stmt)).scalars().first()
@@ -95,10 +146,13 @@ async def test_approve_writes_knowledge_base(db_session):
     assert approved.status == "applied"
     assert approved.reviewer_id == "reviewer-1"
 
-    # knowledge_base 应新增一行（内容仍是桩占位符）
+    # knowledge_base 应新增一行
     kb_rows = (await db_session.execute(select(KnowledgeBase))).scalars().all()
     assert len(kb_rows) == 1
-    assert "[待AI生成]" in kb_rows[0].title
+    # 内容不再是占位符
+    assert "[待AI生成]" not in kb_rows[0].title
+    # 图字段应被保留
+    assert kb_rows[0].graph_sync_status in ("pending", "synced", "failed")
 
 
 @pytest.mark.asyncio
@@ -107,8 +161,12 @@ async def test_reject_marks_rejected(db_session):
     _seed_useless_annotations(db_session, msg_id="msg-x", n=3)
     await db_session.flush()
 
-    service = KnowledgeIterationService()
-    await service.analyze_and_generate_suggestions(db_session, days=30)
+    with patch(
+        "app.services.wingman_service.WingmanService",
+        return_value=await _create_mock_wingman(),
+    ):
+        service = KnowledgeIterationService()
+        await service.analyze_and_generate_suggestions(db_session, days=30)
 
     stmt = select(KnowledgeSuggestion).where(KnowledgeSuggestion.status == "pending")
     suggestion = (await db_session.execute(stmt)).scalars().first()
@@ -127,15 +185,24 @@ async def test_reject_marks_rejected(db_session):
 
 @pytest.mark.asyncio
 async def test_stats_counts_correctly(db_session):
-    """get_suggestion_stats 统计正确。"""
+    """get_suggestion_stats 统计正确（含新增状态）。"""
     _seed_useless_annotations(db_session, msg_id="msg-x", n=3)
     await db_session.flush()
-    service = KnowledgeIterationService()
-    await service.analyze_and_generate_suggestions(db_session, days=30)
+
+    with patch(
+        "app.services.wingman_service.WingmanService",
+        return_value=await _create_mock_wingman(),
+    ):
+        service = KnowledgeIterationService()
+        await service.analyze_and_generate_suggestions(db_session, days=30)
 
     stats = await service.get_suggestion_stats(db_session)
     assert stats["total"] >= 1
     assert stats["pending"] >= 1
+    # 新增状态键应存在
+    assert "queued" in stats
+    assert "graph_synced" in stats
+    assert "expired" in stats
 
     # approve 一条后 applied +1
     stmt = select(KnowledgeSuggestion).where(KnowledgeSuggestion.status == "pending")
@@ -143,3 +210,68 @@ async def test_stats_counts_correctly(db_session):
     await service.approve_suggestion(db_session, s.id, "reviewer-1")
     stats2 = await service.get_suggestion_stats(db_session)
     assert stats2["applied"] >= 1
+
+
+@pytest.mark.asyncio
+async def test_source_failed_on_low_confidence(db_session):
+    """验证置信度 < 0.7 的提案标记 source_failed=True。"""
+    _seed_useless_annotations(db_session, msg_id="msg-low-conf", n=3)
+    await db_session.flush()
+
+    # Mock 返回低置信度结果
+    mock = MagicMock()
+    mock.generate_knowledge_suggestion = AsyncMock(
+        return_value=MOCK_AI_RESULT_LOW_CONFIDENCE
+    )
+    mock.close = AsyncMock()
+
+    with patch(
+        "app.services.wingman_service.WingmanService",
+        return_value=mock,
+    ):
+        service = KnowledgeIterationService()
+        await service.analyze_and_generate_suggestions(db_session, days=30)
+
+    stmt = select(KnowledgeSuggestion).where(
+        KnowledgeSuggestion.source_data.contains("msg-low-conf")
+    )
+    suggestions = (await db_session.execute(stmt)).scalars().all()
+    assert len(suggestions) >= 1
+    # 置信度 < 0.7 应标记 source_failed
+    for s in suggestions:
+        if s.confidence is not None and s.confidence < 0.7:
+            assert s.source_failed is True
+
+
+@pytest.mark.asyncio
+async def test_queue_suggestion_transitions(db_session):
+    """验证入队列状态转换。"""
+    # 直接创建一个 pending 的建议
+    suggestion = KnowledgeSuggestion(
+        suggestion_type="new_faq",
+        status="pending",
+        title="队列测试建议",
+        content="测试内容",
+        category="软件",
+        tags=["测试"],
+        source_type="conversation",
+        source_data=["conv-queue-test"],
+        reason="入队列测试",
+        confidence=0.8,
+        audience="employee_quick_reply",
+    )
+    db_session.add(suggestion)
+    await db_session.commit()
+    await db_session.refresh(suggestion)
+
+    service = KnowledgeIterationService()
+    result = await service.queue_suggestion(db_session, suggestion.id)
+
+    assert result is not None
+    assert result.status == "queued"
+    assert result.queued_at is not None
+
+    # 获取队列统计
+    queue_stats = await service.get_queue_stats(db_session)
+    assert queue_stats["queued_total"] >= 1
+    assert "by_audience" in queue_stats