In this blog post we are going to show how to run a Hugging Face model locally using Apple's MLX-LM framework, connect it to the Agno agentic framework, and extend the agent's capabilities with custom tools. Agno is a modern and extensible agentic framework for Python. It is designed to be model-agnostic, allowing to seamlessly switch between different LLM providers and models. Medium post can be found here.

Environment variables are key-value pairs that exist in the operating system's environment and can be accessed by programs running on that system. They are a way to configure applications without hardcoding values directly into the source code. Python provides several ways to work with environment variables through the os module. Two common methods are os.environ and os.getenv. While they might seem similar, they have different use cases and behaviors. Let's explore the differences and when to use each method effectively.

I have always believed that mathematics is about thinking rather than memorizing. The trigonometric identities were among the things we were told to memorize at school, and not only I struggled with that but I also actively rebelled against this approach. For me, mathematics is fundamentally about having a minimal but sufficient set of definitions and axioms, understanding them deeply, and then deriving everything else from these foundations. During one of my Complex Analysis lectures in my undergraduate Applied Mathematics studies, I was introduced to Euler's formula. When I discovered how it could be used to derive some of the most important trigonometric identities, I was more than relieved. I could finally derive the identities easily whenever I needed them, instead of relying on rote memorization, or panicking about my failure to memorize them.

Nanonets has released Nanonets-OCR-s, a state-of-the-art small 3B image-to-markdown OCR model that goes far beyond traditional text extraction. The model is available on Hugging Face and integrated with their docext tool for immediate use. Medium post can be found here.

Building intelligent applications that can interact with real-world data requires more than just Large Language Models (LLMs), it requires the ability to call external functions and tools. Tool calling transforms a conversational LLM into an agent that can execute code, query APIs, and perform tasks. In this blog post, we are going to create a job search assistant using the MLX-LM Server, connect it to the OpenAI client , and utilise the Qwen3-8B model’s tool‐calling abilities. We are going to build a tool that scrapes job postings from DEV.BG, a popular Bulgarian job board, and provides intelligent responses about available positions.