I have a local weaviate server running. I ingest my data in it and can use it for similarity search usecases

The process is

1. Define an embeddings model using langchain
2. Ingest the data in the weaviate local client. This includes passing the data through the embedding model
3. Use it in similarity search

However, whenever I take down the service with docker compose down or restart the VM I have to do all three steps again. Is that entirely necessary? Can I do better? Presently it takes me about 2 hours to get started