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/services/document_processor.py

@@ -6,6 +6,7 @@ import traceback
 import json
 from app.db.session import SessionLocal
 from io import BytesIO
+from types import SimpleNamespace
 from typing import Optional, List, Dict, Any
 from fastapi import UploadFile
 from langchain_community.document_loaders import (
@@ -26,6 +27,7 @@ from minio.error import MinioException
 from minio.commonconfig import CopySource
 from app.services.vector_store import VectorStoreFactory
 from app.services.embedding.embedding_factory import EmbeddingsFactory
+from app.services.model_config import ModelConfigService
 
 class UploadResult(BaseModel):
     file_path: str
@@ -120,7 +122,45 @@ def _sanitize_metadata_for_vector_store(metadata: Optional[Dict[str, Any]]) -> D
 
     return sanitized
 
-async def process_document(file_path: str, file_name: str, kb_id: int, document_id: int, chunk_size: int = 1000, chunk_overlap: int = 200) -> None:
+def _resolve_model_profile(db: Session, user_id: Optional[int]) -> Any:
+    if user_id is None:
+        return None
+    return ModelConfigService.require_active_config(db, user_id)
+
+
+def _model_profile_snapshot(model_profile: Any) -> Any:
+    if model_profile is None:
+        return None
+    return SimpleNamespace(
+        provider=model_profile.provider,
+        api_key=model_profile.api_key,
+        api_base=model_profile.api_base,
+        chat_model=model_profile.chat_model,
+        embedding_model=model_profile.embedding_model,
+    )
+
+
+def _load_model_profile_for_user(user_id: Optional[int]) -> Any:
+    if user_id is None:
+        return None
+    db = SessionLocal()
+    try:
+        model_profile = ModelConfigService.require_active_config(db, user_id)
+        ModelConfigService.touch_last_used(db, model_profile)
+        return _model_profile_snapshot(model_profile)
+    finally:
+        db.close()
+
+
+async def process_document(
+    file_path: str,
+    file_name: str,
+    kb_id: int,
+    document_id: int,
+    chunk_size: int = 1000,
+    chunk_overlap: int = 200,
+    user_id: Optional[int] = None,
+) -> None:
     """Process document and store in vector database with incremental updates"""
     logger = logging.getLogger(__name__)
     
@@ -129,7 +169,8 @@ async def process_document(file_path: str, file_name: str, kb_id: int, document_
         
         # Initialize embeddings
         logger.info("Initializing OpenAI embeddings...")
-        embeddings = EmbeddingsFactory.create()
+        model_profile = _load_model_profile_for_user(user_id)
+        embeddings = EmbeddingsFactory.create(model_profile=model_profile)
         
         logger.info(f"Initializing vector store with collection: kb_{kb_id}")
         vector_store = VectorStoreFactory.create(
@@ -202,7 +243,7 @@ async def process_document(file_path: str, file_name: str, kb_id: int, document_
                 try:
                     from app.services.graph.graphrag_adapter import GraphRAGAdapter
 
-                    graph_adapter = GraphRAGAdapter()
+                    graph_adapter = GraphRAGAdapter(model_profile=model_profile)
                     source_texts = [doc.page_content for doc in documents_to_update if doc.page_content.strip()]
                     await graph_adapter.ingest_texts(kb_id, source_texts)
                     logger.info("GraphRAG ingestion completed in incremental processing")
@@ -323,7 +364,8 @@ async def process_document_background(
     task_id: int,
     db: Session = None,
     chunk_size: int = 1000,
-    chunk_overlap: int = 200
+    chunk_overlap: int = 200,
+    user_id: Optional[int] = None,
 ) -> None:
     """Process document in background"""
     logger = logging.getLogger(__name__)
@@ -348,6 +390,9 @@ async def process_document_background(
         logger.info(f"Task {task_id}: Setting status to processing")
         task.status = "processing"
         db.commit()
+        model_profile = _resolve_model_profile(db, user_id)
+        if model_profile is not None:
+            ModelConfigService.touch_last_used(db, model_profile)
         
         # 1. 从临时目录下载文件
         minio_client = get_minio_client()
@@ -416,7 +461,7 @@ async def process_document_background(
             
             # 3. 创建向量存储
             logger.info(f"Task {task_id}: Initializing vector store")
-            embeddings = EmbeddingsFactory.create()
+            embeddings = EmbeddingsFactory.create(model_profile=model_profile)
             
             vector_store = VectorStoreFactory.create(
                 store_type=settings.VECTOR_STORE_TYPE,
@@ -520,7 +565,7 @@ async def process_document_background(
                     from app.services.graph.graphrag_adapter import GraphRAGAdapter
 
                     logger.info(f"Task {task_id}: Starting GraphRAG ingestion")
-                    graph_adapter = GraphRAGAdapter()
+                    graph_adapter = GraphRAGAdapter(model_profile=model_profile)
                     source_texts = [doc.page_content for doc in documents if doc.page_content.strip()]
                     await graph_adapter.ingest_texts(kb_id, source_texts)
                     logger.info(f"Task {task_id}: GraphRAG ingestion completed")