Second

local_knowledge_base.py

@@ -1,6 +1,8 @@
 import os
 from qdrant_client import QdrantClient
 from qdrant_client.models import Distance, VectorParams, PointStruct
+from langchain_community.document_loaders import TextLoader, PyPDFLoader, DirectoryLoader
+from langchain.text_splitter import Language
 
 class LocalKnowledgeBase:
     def __init__(self, collection_name="documents", persist_dir="./qdrant_data"):
@@ -37,6 +39,11 @@ class LocalKnowledgeBase:
             '.pdf': PyPDFLoader,
             '.txt': lambda path: DirectoryLoader(path, glob="**/*.txt"),
             '.docx': lambda path: DirectoryLoader(path, glob="**/*.docx"),
+            '.hpp': lambda path: TextLoader(path, encoding='utf-8'),
+            '.cpp': lambda path: TextLoader(path, encoding='utf-8'),   # добавить
+            '.h': lambda path: TextLoader(path, encoding='utf-8'),     # добавить
+            '.cc': lambda path: TextLoader(path, encoding='utf-8'),    # добавить
+            
         }
         
         all_documents = []
@@ -54,7 +61,8 @@ class LocalKnowledgeBase:
                         print(f"Ошибка загрузки {file_path}: {e}")
         
         # Разбиваем на чанки
-        text_splitter = RecursiveCharacterTextSplitter(
+        text_splitter = RecursiveCharacterTextSplitter.from_language(
+            language=Language.CPP,
             chunk_size=1000,
             chunk_overlap=200,
             length_function=len
@@ -82,7 +90,8 @@ class LocalKnowledgeBase:
                 payload={
                     "text": doc.page_content,
                     "source": doc.metadata.get("source", "unknown"),
-                    "page": doc.metadata.get("page", 0)
+                    "page": doc.metadata.get("page", 0),
+                    "file_type": "cpp"
                 }
             )
             points.append(point)