test: 补充图谱抽取器、metadata 持久化与检索元数据测试；更新 roadmap

- LLM 抽取器测试：禁止自定义 Prompt、Schema 拼接、配置锁定修改
- metadata 持久化测试：验证 _save_metadata 不覆写数据库已有图谱配置
- 检索测试：验证 resource_id 写入 metadata
- roadmap 补充图谱抽取器配置优化条目

backend/test/unit/graphs/test_milvus_graph_build.py

@@ -1,11 +1,11 @@
 from __future__ import annotations
 
 from types import SimpleNamespace
-from unittest.mock import MagicMock
+from unittest.mock import AsyncMock, MagicMock
 
 import pytest
 
-from yuxi.knowledge.graphs.extractors import normalize_extraction_result
+from yuxi.knowledge.graphs.extractors import LLMGraphExtractor, normalize_extraction_result
 from yuxi.knowledge.graphs.milvus_graph_service import MilvusGraphService
 
 
@@ -63,6 +63,70 @@ def test_normalize_extraction_result_rejects_invalid_payload(payload):
         normalize_extraction_result(payload, "llm")
 
 
+def test_llm_graph_extractor_rejects_custom_prompt():
+    extractor = LLMGraphExtractor({"model_spec": "test/model", "prompt": "custom"})
+
+    with pytest.raises(ValueError, match="不支持自定义完整 Prompt"):
+        extractor.validate_options()
+
+
+def test_llm_graph_extractor_appends_schema_to_fixed_prompt():
+    extractor = LLMGraphExtractor(
+        {
+            "model_spec": "test/model",
+            "schema": "实体类型只能是 Person 或 Organization",
+            "concurrency_count": 5,
+            "model_params": {"temperature": 0.1},
+        }
+    )
+
+    prompt = extractor._build_prompt("张三任职于公司")
+
+    assert "请从下面文本中抽取实体和实体关系" in prompt
+    assert "抽取 Schema 约束" in prompt
+    assert "实体类型只能是 Person 或 Organization" in prompt
+    assert "文本：\n张三任职于公司" in prompt
+
+
+@pytest.mark.asyncio
+async def test_milvus_graph_service_configure_persists_updated_concurrency():
+    kb = SimpleNamespace(
+        kb_type="milvus",
+        additional_params={
+            "graph_build_config": {
+                "locked": True,
+                "extractor_type": "llm",
+                "extractor_options": {"model_spec": "test/model", "concurrency_count": 5},
+            }
+        },
+    )
+
+    class Repo:
+        async def get_by_id(self, db_id):
+            return kb
+
+        async def update(self, db_id, data):
+            kb.additional_params = data["additional_params"]
+            return kb
+
+    chunk_repo = SimpleNamespace(
+        count_by_db_id=AsyncMock(return_value=0),
+        count_graph_pending_by_db_id=AsyncMock(return_value=0),
+        count_graph_indexed_by_db_id=AsyncMock(return_value=0),
+    )
+    service = MilvusGraphService(kb_repo=Repo(), chunk_repo=chunk_repo)
+
+    await service.configure(
+        "kb_test",
+        extractor_type="llm",
+        extractor_options={"model_spec": "test/model", "concurrency_count": 9},
+        created_by="user_1",
+    )
+    status = await service.get_status("kb_test")
+
+    assert status["config"]["extractor_options"]["concurrency_count"] == 9
+
+
 def test_milvus_graph_service_writes_chunk_entity_and_relation():
     tx = MagicMock()
     session = MagicMock()
@@ -82,7 +146,7 @@ def test_milvus_graph_service_writes_chunk_entity_and_relation():
         end_char_pos=8,
     )
 
-    ent_ids, tags = service.write_chunk_graph(
+    entities, triples = service.write_chunk_graph(
         "kb_test",
         chunk,
         normalize_extraction_result(
@@ -104,8 +168,9 @@ def test_milvus_graph_service_writes_chunk_entity_and_relation():
         ),
     )
 
-    assert ent_ids == ["张三", "公司"]
-    assert tags == ["Organization", "Person"]
+    assert [entity["name"] for entity in entities] == ["张三", "公司"]
+    assert {entity["label"] for entity in entities} == {"Person", "Organization"}
+    assert triples[0]["relation_type"] == "WORKS_AT"
     queries = [call.args[0] for call in tx.run.call_args_list]
     assert any("MERGE (c:Chunk:MilvusKB:`kb_test`" in query for query in queries)
     assert any("MERGE (e:Entity:MilvusKB:`kb_test`" in query for query in queries)
@@ -135,4 +200,4 @@ def test_milvus_graph_service_get_stats_empty_db_id():
     import asyncio
 
     result = asyncio.get_event_loop().run_until_complete(service.get_stats(db_id=None))
-    assert result == {"total_nodes": 0, "total_edges": 0, "entity_types": []}
+    assert result == {"total_nodes": 0, "total_edges": 0, "entity_types": []}

backend/test/unit/knowledge/test_file_size_fallback.py

@@ -1,3 +1,4 @@
+from types import SimpleNamespace
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
@@ -164,3 +165,77 @@ class TestAddFileRecordSizeFallback:
             metadata = await kb.add_file_record("db1", item, params=params)
 
             assert metadata.get("size") is None
+
+
+@pytest.mark.unit
+@pytest.mark.asyncio
+async def test_save_metadata_does_not_overwrite_existing_kb_config():
+    from yuxi.knowledge.base import KnowledgeBase
+
+    class TestKB(KnowledgeBase):
+        @property
+        def kb_type(self):
+            return "test"
+
+        async def _create_kb_instance(self, db_id, config):
+            pass
+
+        async def _initialize_kb_instance(self, instance):
+            pass
+
+        async def index_file(self, db_id, file_id, operator_id=None):
+            return {}
+
+        async def update_content(self, db_id, file_ids, params=None):
+            return []
+
+        async def aquery(self, query_text, db_id, **kwargs):
+            return []
+
+        def get_query_params_config(self, db_id, **kwargs):
+            return {"type": "test", "options": []}
+
+        async def delete_file(self, db_id, file_id):
+            pass
+
+        async def get_file_basic_info(self, db_id, file_id):
+            return {}
+
+        async def get_file_content(self, db_id, file_id):
+            return {}
+
+        async def get_file_info(self, db_id, file_id):
+            return {}
+
+    class ExistingKbRepo:
+        def __init__(self):
+            self.created = []
+            self.updated = []
+
+        async def get_by_id(self, db_id):
+            return SimpleNamespace(db_id=db_id)
+
+        async def create(self, payload):
+            self.created.append(payload)
+
+        async def update(self, db_id, data):
+            self.updated.append((db_id, data))
+
+    kb_repo = ExistingKbRepo()
+    kb = TestKB(work_dir="/tmp/test_kb")
+    kb.databases_meta["db1"] = {
+        "name": "Runtime name",
+        "description": "Runtime description",
+        "kb_type": "test",
+        "metadata": {"graph_build_config": {"extractor_options": {"concurrency_count": 5}}},
+    }
+
+    with (
+        patch("yuxi.repositories.knowledge_base_repository.KnowledgeBaseRepository", return_value=kb_repo),
+        patch("yuxi.repositories.knowledge_file_repository.KnowledgeFileRepository", return_value=SimpleNamespace()),
+        patch("yuxi.repositories.evaluation_repository.EvaluationRepository", return_value=SimpleNamespace()),
+    ):
+        await kb._save_metadata()
+
+    assert kb_repo.created == []
+    assert kb_repo.updated == []

backend/test/unit/toolkits/test_kbs_tools.py

@@ -101,6 +101,7 @@ async def test_query_kb_returns_milvus_chunks_without_sandbox_paths(monkeypatch)
             "metadata": {
                 "file_id": "file-1",
                 "source": "auth-guide.pdf",
+                "resource_id": "db-1",
             },
         }
     ]
@@ -150,6 +151,7 @@ async def test_query_kb_allows_dify_knowledge_base(monkeypatch) -> None:
             "metadata": {
                 "file_id": "dify-doc-1",
                 "source": "Dify Doc",
+                "resource_id": "db-1",
             },
         }
     ]
@@ -219,6 +221,7 @@ async def test_query_kb_normalizes_file_metadata_for_open(monkeypatch) -> None:
 
     assert result[0]["metadata"] == {
         "file_id": "file-1",
+        "resource_id": "db-1",
         "chunk_id": "chunk-1",
         "chunk_index": 3,
     }