We are discussing at the concept of an MCP server. We look at a simple example of how to create an MCP server, how to add tools to it, and how to connect it to applications like Claude Desktop or even other Python scripts.

This blog posts discusses the difference between two architectures, namely Tools for LLMs vs LLM-powered tools. It looks at the differences of each and when to use which architecture. It also looks at how tools for LLMs should be organized and why docstrings are so important for them.

We look at combining different RAG options into an agent. We explore a bit how to put together a basic agent setup with LangChain or LlamaIndex. The goal is not a complete tutorial, but more an outlook at how it can be done.

We look at the difference between Standard RAG and Graph RAG and how one can implement them.

We are looking at four Python packages that look at how to implement agents in a system.

We are taking a look at four packages that handle specialised problems and tasks related to LLM calls, namely Instructor, Outlines, Mirascope, and Guidance.

We look at two specialised LLM frameworks. DSPy and LangChain. One focuses on taking prompt engineering out of the developers ' hands and the other changes a usually linear workflow into a conditional graph.

We are taking a look at two frameworks for LLM-powered applications. We start with LlamaIndex and then proceed to Haystack. Looking at what they excel and specialise in, in contrast to good old LangChain.

We return to the Python Library LangChain and look at how to implement RAG workflow capabilities to our previous scripts.

This post is a very small introduction into the workings of the Python LLM "control" package LangChain. We show the basic functionality and how to build a small pipeline for a simple documentation request script.