feat: 增强知识库功能，优化索引和RAG检索

server/tests/test_knowledge_normalizer.py

@@ -44,9 +44,9 @@ def test_knowledge_normalizer_appends_structured_table(monkeypatch) -> None:
 
         enriched = service.build_enriched_text(raw_text)
 
-    assert enriched.startswith("# 结构化表格补充")
+    assert enriched.startswith(raw_text.strip())
     assert "| 餐补 | 75 | 65 | 55 | 140 |" in enriched
-    assert enriched.endswith(raw_text.strip())
+    assert enriched.endswith("| 合计 | 110 | 100 | 90 | 175 |")
 
 
 def test_knowledge_normalizer_keeps_only_markdown_table_body() -> None:
@@ -79,12 +79,12 @@ def test_knowledge_normalizer_builds_section_navigation_without_table() -> None:
         service = KnowledgeNormalizationService(db)
         enriched = service.build_enriched_text(raw_text)
 
-    assert enriched.startswith("# 章节导航")
+    assert enriched.startswith(raw_text.strip())
     assert "- 第一章 总则" in enriched
     assert "## 第二章 住宿费标准" in enriched
     assert "# 问答线索补充" in enriched
     assert "- 第二章 住宿费标准：住宿费按照出差城市档位和职级标准执行" in enriched
-    assert enriched.endswith(raw_text.strip())
+    assert "# 章节导航" in enriched
 
 
 def test_knowledge_normalizer_builds_answer_clues_from_lists_and_kv_lines() -> None:

server/tests/test_knowledge_rag_service.py

@@ -74,6 +74,28 @@ def test_build_hits_prioritizes_answer_clue_appendix_for_rule_queries() -> None:
     assert [item["candidate_id"] for item in hits] == ["clue-1", "plain-1"]
 
 
+def test_build_hits_demotes_chapter_navigation_for_specific_rule_queries() -> None:
+    hits = KnowledgeRagService._build_hits_from_query_data(
+        query="探亲差旅归哪个部门管理？",
+        chunks=[
+            {
+                "chunk_id": "toc-1",
+                "file_path": "/tmp/doc-1__费用制度.md",
+                "content": "# 章节导航\n\n- 第一章 总则\n- 第二章 职责分工\n- 第三章 支出归口",
+            },
+            {
+                "chunk_id": "body-1",
+                "file_path": "/tmp/doc-1__费用制度.md",
+                "content": "附表3：支出归口管理部门与归口业务范围\n组织人事部：探亲差旅、条件艰苦及安全风险较高区域补助等支出。",
+            },
+        ],
+        entities=[],
+        limit=2,
+    )
+
+    assert [item["candidate_id"] for item in hits] == ["body-1", "toc-1"]
+
+
 def test_resolve_default_qdrant_url_prefers_container_host(monkeypatch) -> None:
     monkeypatch.setattr(
         knowledge_rag_module.socket,
@@ -93,3 +115,29 @@ def test_resolve_default_qdrant_url_falls_back_to_loopback(monkeypatch) -> None:
     monkeypatch.setattr(knowledge_rag_module.socket, "getaddrinfo", raise_lookup_error)
 
     assert knowledge_rag_module._resolve_default_qdrant_url() == "http://127.0.0.1:6333"
+
+
+def test_is_query_ready_status_rejects_failed_status_even_with_chunks() -> None:
+    assert (
+        KnowledgeRagService.is_query_ready_status(
+            {
+                "status": "failed",
+                "chunks_count": 11,
+                "chunks_list": ["chunk-1"],
+            }
+        )
+        is False
+    )
+
+
+def test_is_query_ready_status_rejects_processing_status_even_with_chunks() -> None:
+    assert (
+        KnowledgeRagService.is_query_ready_status(
+            {
+                "status": "processing",
+                "chunks_count": 11,
+                "chunks_list": ["chunk-1"],
+            }
+        )
+        is False
+    )

server/tests/test_knowledge_service.py

@@ -0,0 +1,81 @@
+from __future__ import annotations
+
+from datetime import UTC, datetime
+
+from sqlalchemy import create_engine
+from sqlalchemy.orm import Session, sessionmaker
+from sqlalchemy.pool import StaticPool
+
+from app.api.deps import CurrentUserContext
+from app.core.agent_enums import AgentName, AgentRunSource, AgentRunStatus
+from app.db.base import Base
+from app.services.agent_runs import AgentRunService
+from app.services.knowledge import (
+    KNOWLEDGE_INGEST_STATUS_FAILED,
+    KNOWLEDGE_INGEST_STATUS_SYNCING,
+    KnowledgeService,
+)
+
+
+def build_session() -> Session:
+    engine = create_engine(
+        "sqlite+pysqlite:///:memory:",
+        connect_args={"check_same_thread": False},
+        poolclass=StaticPool,
+    )
+    Base.metadata.create_all(bind=engine)
+    session_factory = sessionmaker(bind=engine, autoflush=False, autocommit=False)
+    return session_factory()
+
+
+def test_reconcile_document_ingest_status_keeps_failed_when_linked_run_failed(
+    tmp_path,
+    monkeypatch,
+) -> None:
+    with build_session() as db:
+        service = KnowledgeService(storage_root=tmp_path, db=db)
+        uploaded = service.upload_document(
+            "报销制度",
+            "demo.txt",
+            b"hello",
+            CurrentUserContext(
+                username="admin",
+                name="管理员",
+                role_codes=["manager"],
+                is_admin=True,
+            ),
+        )
+
+        run = AgentRunService(db).create_run(
+            agent=AgentName.HERMES.value,
+            source=AgentRunSource.USER_MESSAGE.value,
+            status=AgentRunStatus.FAILED.value,
+            route_json={"job_type": "knowledge_index_sync"},
+        )
+        service.set_document_ingest_statuses(
+            [uploaded.id],
+            KNOWLEDGE_INGEST_STATUS_SYNCING,
+            agent_run_id=run.run_id,
+        )
+
+        monkeypatch.setattr(
+            "app.services.knowledge_rag.KnowledgeRagService.get_document_status_map",
+            lambda self, _document_ids: {
+                uploaded.id: {
+                    "status": "processing",
+                    "query_ready": False,
+                    "updated_at": datetime.now(UTC).isoformat(),
+                }
+            },
+        )
+
+        index = service._load_index()
+        changed = service._reconcile_document_ingest_statuses(
+            index,
+            document_ids=[uploaded.id],
+            preserve_syncing=False,
+        )
+
+        entry = next(item for item in index["documents"] if item["id"] == uploaded.id)
+        assert changed is True
+        assert entry["ingest_status"] == KNOWLEDGE_INGEST_STATUS_FAILED