update swagger

backend/src/application/services/rag_service.py

@@ -29,7 +29,7 @@ class RAGService:
         self.splitter = splitter or TextSplitter()
 
     async def index_document(self, document: Document) -> list[DocumentChunk]:
-        chunks_text = self.splitter.split(document.content)
+        chunks_text = self.splitter.split(document.content)        
         chunks: list[DocumentChunk] = []
         for idx, text in enumerate(chunks_text):
             chunks.append(
@@ -42,9 +42,18 @@ class RAGService:
                 )
             )
 
-        embeddings = self.embedding_service.embed_texts([c.content for c in chunks])
+        EMBEDDING_BATCH_SIZE = 50
+        all_embeddings: list[list[float]] = []
+        
+        for i in range(0, len(chunks), EMBEDDING_BATCH_SIZE):
+            batch_chunks = chunks[i:i + EMBEDDING_BATCH_SIZE]
+            batch_texts = [c.content for c in batch_chunks]
+            batch_embeddings = self.embedding_service.embed_texts(batch_texts)
+            all_embeddings.extend(batch_embeddings)
+        
+        print(f"Created {len(all_embeddings)} embeddings, upserting to Qdrant...")
         await self.vector_repository.upsert_chunks(
-            chunks, embeddings, model_version=self.embedding_service.model_version()
+            chunks, all_embeddings, model_version=self.embedding_service.model_version()
         )
         return chunks
 

backend/src/application/services/text_splitter.py

@@ -39,5 +39,10 @@ class TextSplitter:
 
     def _split_sentences(self, text: str) -> Iterable[str]:
         parts = re.split(r"(?<=[\.\?\!])\s+", text)
+        if len(parts) == 1 and len(text) > self.chunk_size * 2:
+            chunk_text = []
+            for i in range(0, len(text), self.chunk_size):
+                chunk_text.append(text[i:i + self.chunk_size])
+            return chunk_text
         return [p.strip() for p in parts if p.strip()]
 

backend/src/infrastructure/repositories/qdrant/vector_repository.py

@@ -36,6 +36,8 @@ class QdrantVectorRepository(IVectorRepository):
         embeddings: Sequence[list[float]],
         model_version: str,
     ) -> None:
+        BATCH_SIZE = 100
+        
         points = []
         for chunk, vector in zip(chunks, embeddings):
             points.append(
@@ -52,7 +54,13 @@ class QdrantVectorRepository(IVectorRepository):
                     },
                 )
             )
-        self.client.upsert(collection_name=self.collection_name, points=points)
+            
+            if len(points) >= BATCH_SIZE:
+                self.client.upsert(collection_name=self.collection_name, points=points)
+                points = []
+        
+        if points:
+            self.client.upsert(collection_name=self.collection_name, points=points)
 
     async def search(
         self,

tg_bot/infrastructure/telegram/handlers/collection_handler.py

@@ -3,6 +3,7 @@ from aiogram.types import Message, InlineKeyboardMarkup, InlineKeyboardButton, C
 from aiogram.filters import Command, StateFilter
 from aiogram.fsm.context import FSMContext
 import aiohttp
+from urllib.parse import unquote
 from tg_bot.config.settings import settings
 from tg_bot.infrastructure.http_client import create_http_session
 from tg_bot.infrastructure.telegram.states.collection_states import (
@@ -10,6 +11,18 @@ from tg_bot.infrastructure.telegram.states.collection_states import (
     CollectionEditStates
 )
 
+
+def decode_title(title: str) -> str:
+    if not title:
+        return "Без названия"
+    try:
+        decoded = unquote(title)
+        if decoded != title or '%' not in title:
+            return decoded
+        return title
+    except Exception:
+        return title
+
 router = Router()
 
 
@@ -243,7 +256,7 @@ async def cmd_search(message: Message):
 
     response = f"<b>Результаты поиска:</b> \"{query}\"\n\n"
     for i, doc in enumerate(results[:5], 1):
-        title = doc.get("title", "Без названия")
+        title = decode_title(doc.get("title", "Без названия"))
         content = doc.get("content", "")[:200]
         response += f"{i}. <b>{title}</b>\n"
         response += f"   <i>{content}...</i>\n\n"
@@ -378,7 +391,7 @@ async def show_collection_documents(callback: CallbackQuery):
     
     for i, doc in enumerate(documents[:10], 1):
         doc_id = doc.get("document_id")
-        title = doc.get("title", "Без названия")
+        title = decode_title(doc.get("title", "Без названия"))
         content_preview = doc.get("content", "")[:100]
         response += f"{i}. <b>{title}</b>\n"
         if content_preview:

tg_bot/infrastructure/telegram/handlers/document_handler.py

@@ -6,6 +6,7 @@ from aiogram.types import Message, InlineKeyboardMarkup, InlineKeyboardButton, C
 from aiogram.filters import StateFilter
 from aiogram.fsm.context import FSMContext
 import aiohttp
+from urllib.parse import unquote
 from tg_bot.config.settings import settings
 from tg_bot.infrastructure.http_client import create_http_session
 from tg_bot.infrastructure.telegram.states.collection_states import (
@@ -13,6 +14,18 @@ from tg_bot.infrastructure.telegram.states.collection_states import (
     DocumentUploadStates
 )
 
+
+def decode_title(title: str) -> str:
+    """Декодирует URL-encoded название документа"""
+    if not title:
+        return "Без названия"
+    try:
+        decoded = unquote(title)
+        if decoded != title or '%' not in title:
+            return decoded
+        return title
+    except Exception:
+        return title
 router = Router()
 
 
@@ -108,7 +121,7 @@ async def view_document(callback: CallbackQuery):
         )
         return
     
-    title = document.get("title", "Без названия")
+    title = decode_title(document.get("title", "Без названия"))
     content = document.get("content", "")
     collection_id = document.get("collection_id")
     
@@ -184,7 +197,7 @@ async def edit_document_prompt(callback: CallbackQuery, state: FSMContext):
     await callback.message.answer(
         "<b>Редактирование документа</b>\n\n"
         "Отправьте новое название документа или /skip чтобы оставить текущее.\n\n"
-        f"Текущее название: <b>{document.get('title', 'Без названия')}</b>",
+        f"Текущее название: <b>{decode_title(document.get('title', 'Без названия'))}</b>",
         parse_mode="HTML"
     )
     await callback.answer()
@@ -361,7 +374,7 @@ async def process_upload_document(message: Message, state: FSMContext):
         if result:
             await message.answer(
                 f"<b>✅ Документ загружен и добавлен в коллекцию</b>\n\n"
-                f"<b>Название:</b> {result.get('title', filename)}\n\n"
+                f"<b>Название:</b> {decode_title(result.get('title', filename))}\n\n"
                 f"📄 Документ сейчас индексируется. Вы получите уведомление, когда индексация завершится.\n\n",
                 parse_mode="HTML"
             )

tg_bot/infrastructure/telegram/handlers/question_handler.py

@@ -83,7 +83,9 @@ async def process_premium_question(message: Message, user: User, question_text:
                 title = source.get('title', 'Без названия')
                 try:
                     from urllib.parse import unquote
-                    title = unquote(title)
+                    decoded = unquote(title)
+                    if decoded != title or '%' in title:
+                        title = decoded
                 except:
                     pass
                 response += f"  {idx}. {title}\n"
@@ -148,7 +150,9 @@ async def process_free_question(message: Message, user: User, question_text: str
                 title = source.get('title', 'Без названия')
                 try:
                     from urllib.parse import unquote
-                    title = unquote(title)
+                    decoded = unquote(title)
+                    if decoded != title or '%' in title:
+                        title = decoded
                 except:
                     pass
                 response += f"  {idx}. {title}\n"