rework: switch to Qdrant (vector_store.py)

vector_store.py

@@ -1,16 +1,17 @@
-"""Векторное хранилище ChromaDB + эмбеддинги Ollama."""
+"""Векторное хранилище Qdrant + эмбеддинги Ollama."""
 from __future__ import annotations
 
 import os
-from pathlib import Path
 
-from langchain_chroma import Chroma
+from langchain_qdrant import QdrantVectorStore
 from langchain_core.documents import Document
 from langchain_ollama import OllamaEmbeddings
 from langchain_text_splitters import RecursiveCharacterTextSplitter
+from qdrant_client import QdrantClient
+from qdrant_client.http.models import Distance, VectorParams
 
 COLLECTION_NAME = "knowledge_base"
-CHROMA_PATH = os.getenv("CHROMA_PATH", "./chroma_data")
+QDRANT_PATH = os.getenv("QDRANT_PATH", "./qdrant_data")
 OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://localhost:11434")
 EMBED_MODEL = os.getenv("OLLAMA_EMBED_MODEL", "nomic-embed-text")
 
@@ -22,12 +23,18 @@ def get_embeddings() -> OllamaEmbeddings:
     )
 
 
-def get_vector_store() -> Chroma:
+def get_vector_store() -> QdrantVectorStore:
-    Path(CHROMA_PATH).mkdir(parents=True, exist_ok=True)
+    client = QdrantClient(path=QDRANT_PATH)
-    return Chroma(
+    collections = client.get_collections().collections
+    if not any(col.name == COLLECTION_NAME for col in collections):
+        client.create_collection(
             collection_name=COLLECTION_NAME,
-        embedding_function=get_embeddings(),
+            vectors_config=VectorParams(size=768, distance=Distance.COSINE),
-        persist_directory=CHROMA_PATH,
+        )
+    return QdrantVectorStore(
+        client=client,
+        collection_name=COLLECTION_NAME,
+        embedding=get_embeddings(),
     )