Решение готово к публикации: add vectorstore.py

vectorstore.py

@@ -0,0 +1,46 @@
+"""
+RAG vector store using ChromaDB and Ollama embeddings.
+"""
+
+from pathlib import Path
+from typing import List, Iterable
+
+import chromadb
+from langchain.embeddings.ollama import OllamaEmbeddings
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.schema.document import Document
+
+CHROMA_DIR = "./chroma_db"
+EMBED_MODEL = "nomic-embed-text"
+
+
+def create_vectorstore(persist_directory: str = CHROMA_DIR):
+    """Create or load a Chroma vector store.
+
+    Parameters
+    ----------
+    persist_directory:
+        Directory where the Chroma database is stored. If it does not exist, it will be created.
+    """
+    embeddings = OllamaEmbeddings(model=EMBED_MODEL)
+    client = chromadb.PersistentClient(path=persist_directory)
+    # Use a single collection named "documents"
+    return client.get_or_create_collection(name="documents", embedding_function=embeddings)
+
+
+def load_documents(directory: str, vectorstore) -> None:
+    """Load all .txt and .md files from *directory*, chunk them and add to the vector store.
+
+    The function does not return anything; it mutates the provided collection.
+    """
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
+    docs: List[Document] = []
+    for path in Path(directory).rglob("*.txt"):
+        content = path.read_text(encoding="utf-8")
+        docs.extend(text_splitter.create_documents([content], metadata={"source": str(path)}))
+    for path in Path(directory).rglob("*.md"):
+        content = path.read_text(encoding="utf-8")
+        docs.extend(text_splitter.create_documents([content], metadata={"source": str(path)}))
+
+    if docs:
+        vectorstore.add(documents=docs)