Icon

03 Update a vector store

<p>This workflow demonstrates how you can <strong>update an existing vector store</strong> <strong>without recreating embeddings</strong> for the entire knowledge base.</p><p>The workflow reads an existing vector store and imports new data, converting only the new content into embeddings. It then combines the original and new embeddings and writes them back to the vector store.</p>

Connect to an LLM provider and select the embedding model

03 Update a vector store

This workflow demonstrates how you can update an existing vector storewithout recreating embeddings for the entire knowledge base.

The workflow reads an existing vector store and imports new data, converting only the new content into embeddings. It then combines the original and new embeddings and writes them back to the vector store.

Import the saved vector store

Import new documents and create embeddings only for the new content

Extract existing embeddings into a table

Write the combined embeddings back to a vector store

🔁 No need to recreate embeddings – existing embeddings stored in the table column will be reused. Simply select the column with the embeddings in the node configuration.

Savedvector store
Model Reader
Updateddocuments
PDF Parser
Write back tothe vector store
FAISS Vector Store Creator
Model Writer
Combine old andnew embeddings
Concatenate
Same model that createdthe existing vector store
OpenAI Embedding Model Selector
Create embeddings for the new documentsTemporarily store in the table
Text Embedder
Text chunking
Text Chunker
OpenAI Authenticator
OpenAI key
Credentials Configuration
Document Data Extractor
Extract data andvectors to a table
Vector Store Data Extractor

Nodes

Extensions

Links