Python Hub Weekly Digest for 2025-03-09
This week in Python, the most popular topics included Xonsh, a Python-powered shell, and a tutorial on web scraping and data preparation with PyFlyde. A video on using uv for Python project layout for production apps and FastRTC, a tool to turn any Python function into a real-time audio and video stream, also gained attention. Articles covered a range of topics, from the evolution of Python and managing Python monorepos to creating music with Python and building AI agents for JupyterLab. New projects introduced include opendbc, a Python API for cars, and Minions, a protocol for small on-device models to collaborate with cloud models. Have a great week and happy coding!
💖 Most Popular
Xonsh – A Python-powered shell
AI Engineering Goes Visual: Web Scraping & Data Prep with PyFlyde
In part one of two part tutorial, we explore how to use Flyde, a visual programming tool, to build a web scraper that feeds into a Retrieval Augmentation Generation (RAG) system. We will cover the process of scraping web content and storing it locally, setting the stage for more advanced AI engineering tasks.
My 2025 uv-based Python Project Layout for Production Apps
This video demonstrates how to use uv, a modern Python tool, to transform a simple script into a production-ready application with an optimal project structure. It guides viewers through creating a well-organized FastAPI project, leveraging advanced packaging features like dependency groups, resulting in a scalable and easily shareable codebase.
FastRTC
Turn any python function into a real-time audio and video stream over WebRTC or WebSockets.
The Best Way to Use Text Embeddings Portably is With Parquet and Polars
This article advocates using Parquet files, paired with the Polars library in Python, as an efficient and portable alternative to vector databases for storing and querying text embeddings in smaller projects. It demonstrates how this approach enables fast similarity searches and metadata filtering, outperforming traditional methods like CSV files and offering a balance between simplicity...
📖 Articles
Robyn – "Batman Inspired" Python Web Framework Built with Rust
How Python Evolves: From PEP to Feature
Explore how Python gets new functionality and how PEPs have a huge role in this, by looking at PEP484 and its interesting backstory.
Cracking the Python Monorepo
This article explores using uv and Dagger to manage Python monorepos, aiming to solve challenges like slow CI/CD pipelines and complex dependency management. It details setting up a monorepo with these tools, achieving efficient end-to-end pipeline caching and easy local/CI execution.
Creating drum machines, effects, sequencers, samplers, and synths with Python
I created a subreddit where I post scripts showing how to make drum machines, sequencers, samplers, and synths with Python.
Programming with LLM Agents in 2025
In this video, Sentdex offers tips and tricks for building with modern LLM agents, using OpenHands as an example but applicable to other web-based UIs and agents. It covers practical advice for leveraging LLMs effectively, with resources including the OpenHands GitHub repository, a neural networks book, and community channels for support and discussion.
Ditching Django Admin for FastHTML with HTMX
Building a Django Admin-like dashboard with automatic refresh using FastHTML and HTMX.
Part 2 - Building an LLM RAG with PyFlyde & LangChain
Python as a Second Language Empathy
CCTV_YOLO
Fast Real-time Object Detection with High-Res Output
Probly – Spreadsheets, Python, and AI in the browser
Design of Everyday APIs
What makes a good API for a library? How can you design an API that is delightful to use? This post walks through the principles of what goes into user-centered design and how best to apply those principles when writing a Python library for fellow developers.
Building AI Agents for JupyterLab using Notebook Intelligence
This post introduces Notebook Intelligence (NBI), an AI coding assistant framework for JupyterLab, and demonstrates how to build AI Agents using NBI's tool-calling capabilities with examples of tools for geo-coordinate lookup, map generation, notebook creation, and sharing. It provides a guide for developers to integrate custom AI-powered functionalities into JupyterLab's Copilot Chat.
Visualizing F1 race results using Polars and Great Tables
Programmatically create an F1 race analysis table using the FastF1 API, the Great Tables library, and Polars.
How to add an object in Django Admin with a bookmarklet
The article explains how to create a Django Admin bookmarklet that pre-fills form fields when adding new items, like bookmarks, using a web page's URL, title, and selected text. It walks through setting up the JavaScript snippet, customizing it for your Django app, and adding it as a browser bookmark for quick data entry.
An Introduction to Flask-SQLAlchemy in Python
This article introduces Flask-SQLAlchemy, a Flask extension simplifying database management by integrating SQLAlchemy with Flask applications. It explains SQLAlchemy's core components (Core and ORM) and demonstrates how Flask-SQLAlchemy streamlines configuration, session management, and querying in Flask-based projects with a practical example.
Python micro event loop library (~250 LOC)
volcengine / verl
verl: Volcano Engine Reinforcement Learning for LLMs
Python Hub Weekly Digest for 2025-03-02
⚙️ Projects
opendbc
A Python API for your car.
KTransformers
A Flexible Framework for Experiencing Cutting-edge LLM Inference Optimizations.
TinyCompiler
Writing a compiler in a week-end.
SkyThought
Train your own O1 preview model within $450.
FlashMLA
FlashMLA is an efficient MLA decoding kernel for Hopper GPUs, optimized for variable-length sequences serving.
Codegen
A python library for manipulating codebases.
Minions
Minions is a communication protocol that enables small on-device models to collaborate with frontier models in the cloud. By only reading long contexts locally, we can reduce cloud costs with minimal or no quality degradation.
logis
Turn your git commit history into a scientific log.
Craw4LLM
CRAW4LLM is an efficient web crawling method that prioritizes webpages based on their potential influence on LLM pretraining, replacing traditional graph-connectivity-based priorities. By crawling only 21% of URLs, it achieves the same downstream performance as previous methods, significantly reducing data waste and website burden.
vlmrun-hub
A hub for various industry-specific schemas to be used with VLMs.
AIBrix
Cost-efficient and pluggable Infrastructure components for GenAI inference.
👾 Reddits
Anyone used UV package manager in production
Why not just plot everything in numpy?! P.2.
Project by Ruslan Keba. Since 2012. Powered by Python. Made in 🇺🇦Ukraine.