增加代码知识库；修复文档处理内容；增加API设置

rag-web-ui/backend/app/api/api_v1/testing.py

@@ -12,6 +12,8 @@ from app.models.knowledge import Document, KnowledgeBase
 from app.models.user import User
 from app.schemas.testing import TestingPipelineRequest, TestingPipelineResponse
 from app.services.embedding.embedding_factory import EmbeddingsFactory
+from app.services.llm.llm_factory import LLMFactory
+from app.services.model_config import ModelConfigService
 from app.services.retrieval.multi_kb_retriever import MultiKBRetriever, format_retrieval_context
 from app.services.testing_pipeline import run_testing_pipeline
 from app.services.vector_store import VectorStoreFactory
@@ -21,8 +23,12 @@ logger = logging.getLogger(__name__)
 MODEL_PIPELINE_TIMEOUT_SECONDS = 300
 
 
-async def _build_kb_vector_stores(db: Session, knowledge_bases: List[KnowledgeBase]) -> List[Dict[str, Any]]:
-    embeddings = EmbeddingsFactory.create()
+async def _build_kb_vector_stores(
+    db: Session,
+    knowledge_bases: List[KnowledgeBase],
+    model_profile: Any,
+) -> List[Dict[str, Any]]:
+    embeddings = EmbeddingsFactory.create(model_profile=model_profile)
     kb_vector_stores: List[Dict[str, Any]] = []
 
     for kb in knowledge_bases:
@@ -47,10 +53,9 @@ async def generate_testing_content(
     current_user: User = Depends(get_current_user),
     db: Session = Depends(get_db),
 ) -> Any:
-    _ = current_user
-
+    model_profile = ModelConfigService.get_active_config(db, current_user.id)
     knowledge_context = (payload.knowledge_context or "").strip()
-    if payload.knowledge_base_ids:
+    if payload.knowledge_base_ids and model_profile is not None:
         try:
             knowledge_bases = (
                 db.query(KnowledgeBase)
@@ -61,7 +66,7 @@ async def generate_testing_content(
                 .all()
             )
 
-            kb_vector_stores = await _build_kb_vector_stores(db, knowledge_bases)
+            kb_vector_stores = await _build_kb_vector_stores(db, knowledge_bases, model_profile)
             if kb_vector_stores:
                 retriever = MultiKBRetriever(
                     reranker_weight=settings.RERANKER_WEIGHT,
@@ -81,13 +86,39 @@ async def generate_testing_content(
                 payload.knowledge_base_ids,
                 exc,
             )
+    elif payload.knowledge_base_ids:
+        logger.warning(
+            "Testing generation skipped knowledge retrieval because no active model config exists: user=%s knowledge_base_ids=%s",
+            current_user.id,
+            payload.knowledge_base_ids,
+        )
+
+    use_model_generation = bool(payload.use_model_generation and model_profile is not None)
+    llm_model = None
+    if use_model_generation:
+        try:
+            llm_model = LLMFactory.create(streaming=False, model_profile=model_profile)
+            ModelConfigService.touch_last_used(db, model_profile)
+        except Exception as exc:
+            logger.exception(
+                "Testing generation LLM initialization failed for user=%s, falling back to rule-based output: %s",
+                current_user.id,
+                exc,
+            )
+            use_model_generation = False
+    elif payload.use_model_generation:
+        logger.warning(
+            "Testing generation falling back to rule-based output because no active model config exists: user=%s",
+            current_user.id,
+        )
 
     pipeline_kwargs = {
         "user_requirement_text": payload.requirement_text,
         "requirement_type_input": payload.requirement_type,
         "debug": payload.debug,
         "knowledge_context": knowledge_context,
-        "use_model_generation": payload.use_model_generation,
+        "use_model_generation": use_model_generation,
+        "llm_model": llm_model,
         "max_items_per_group": payload.max_items_per_group,
         "cases_per_item": payload.cases_per_item,
         "max_focus_points": payload.max_focus_points,