Local RAG with Langflow, Milvus, and LM Studio

Had a lot of fun playing around with Langflow, Milvus, and LM Studio this week.

At first, Langflow may seem like just another drag-and-drop, no-code AI builder. But, there's one feature that makes this platform stand out - the Code button that comes with every pre-packaged (and custom) drag-and-drop component. This turns a no-code black-box into an open book where the user can not only view all the inner workings, but also edit the underlying code for their specific use cases.

This is, by far, my favorite feature in Langflow. It makes the quick and painless method of building AI systems (i.e. vibe coding) more tenable and illuminating. I learned, I built (super quickly, btw), I played, and it was just really, really fun.

Langflow is built on top of LangChain, so much of what can be done in this powerhouse of a library can be spun-up and tested in Langflow in a flash. It also uses Docling by default for processing many different types of documents, so you know your documents will be taken good care of (checkout my Docling tutorial to see just how powerful this library is).

To help others start testing out Langflow, I added a tutorial showcasing how to build a simple RAG system using the built-in Milvus and LM Studio components. In this guide, I also demonstrate how to edit some of these default components to obtain behaviors that I was already familiar with from using LangChain.

You can also checkout Langflow's RAG tutorial, Milvus's Langflow tutorial, my Medium post, and my Github repo including different Langflow templates to test.

Next up, building our own components from scratch! Stay tuned for more posts!

You can also checkout my current projects or dive into in-depth tutorials covering various aspects of AI .