Skip to content

Homelab update: a local RAG setup

The basic configuration of my homelab is now complete. The current stack consists of:

  • Ollama, serving local LLMs such as qwen3:30b-a3b, which runs on CPU alone;
  • Qdrant as the vector database;
  • Flowise as the orchestration layer, chosen in order to assess low-code tools for the visual construction and management of LLM pipelines.

The workflow

Flowise RAG workflow

The objective is an entirely local retrieval-augmented generation (RAG) setup.

As a test, a research paper was uploaded as a PDF and subsequently queried through the agent. The setup produced usable answers on the first attempt and required comparatively little configuration effort.