GitHub Weekly Top 10 Trends (22 Jun 2025)

This week’s GitHub Top 10 Trends highlight the incredible pace of innovation and collaboration within the developer community. From powerful new tools and frameworks to creative solutions in AI, robotics, and web technologies, developers around the world are sharing and refining projects that push the boundaries of what’s possible. The trending repositories reflect a strong focus on accessibility, performance, and making complex technologies easier for everyone to use. It’s exciting to see such a diverse range of ideas gaining traction and inspiring developers to build, learn, and contribute together.

For past weekly trending, please view Weekly Tags.

Jan

Jan is a ChatGPT-alternative desktop application that empowers users to run large language models (LLMs) locally on their own devices, ensuring full privacy and control over their data. Developed with the Tauri framework, Jan features a Rust backend and a React frontend, offering a seamless and efficient user experience. Its modular, extension-based architecture enables flexible AI inference, easy model management, and robust chat functionality, making it a powerful tool for anyone seeking a private and customizable conversational AI solution.

Data Engineering Handbook

Data Engineering Handbook is a comprehensive educational resource aimed at helping individuals develop proficiency in data engineering. This repository acts as a centralized hub, offering curated learning materials, hands-on projects, interview preparation resources, and a structured bootcamp curriculum. It is suitable for both beginners seeking to enter the field and experienced professionals who wish to enhance their skills or stay updated with the latest industry trends.

Anthropic Cookbook

Anthropic Cookbook is a comprehensive repository of practical code examples and guides for developing applications with Claude, Anthropic’s AI assistant. This project offers a wide range of copy-and-paste code snippets, complete implementations, and best practices organized into four key areas: core AI skills, tool integration, third-party service integration, and advanced techniques. Covering topics from basic prompt engineering to sophisticated multi-agent systems, the cookbook places special emphasis on production-ready solutions for retrieval augmented generation (RAG), text-to-SQL generation, and tool-augmented workflows, making it an invaluable resource for developers working with Claude.

Awesome LLM Apps

Awesome LLM Apps is a comprehensive repository showcasing practical applications built with Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and AI Agents. This collection demonstrates how to implement a diverse array of LLM-powered solutions using models from leading providers such as OpenAI, Anthropic, and Google, as well as open-source alternatives like DeepSeek, Qwen, and Llama. The repository serves as a valuable resource for developers seeking inspiration and guidance on leveraging cutting-edge AI technologies in real-world applications.

all-rag-techniques

all-rag-techniques is a comprehensive repository offering a technical introduction to a wide range of Retrieval-Augmented Generation (RAG) implementations. Designed as a practical reference, this project features self-contained Jupyter notebooks that demonstrate diverse RAG techniques with an emphasis on simplicity and educational value. Unlike solutions built with specialized frameworks such as LangChain or FAISS, all-rag-techniques showcases how to implement RAG systems from scratch using basic Python libraries. Each notebook provides complete implementation code, step-by-step explanations, evaluation methods, and result visualizations, making this repository an ideal resource for learners and practitioners seeking a deeper understanding of RAG concepts and methods.

zapret-discord-youtube

zapret-discord-youtube is a Windows network traffic manipulation tool designed to bypass Deep Packet Inspection (DPI) restrictions imposed by ISPs or network administrators on Discord and YouTube services. Leveraging the WinDivert driver to intercept network packets at the kernel level, the system applies various desynchronization techniques to evade DPI detection while preserving service functionality. As a user-friendly distribution of the zapret project, zapret-discord-youtube offers multiple pre-configured bypass strategies and comprehensive service management through intuitive batch script interfaces, making it accessible for users seeking to restore access to restricted platforms.

VGGT

VGGT (Visual Geometry Grounded Transformer) is a computer vision system designed for fast and accurate 3D scene reconstruction from images. This repository introduces a feed-forward neural network capable of inferring all essential 3D attributes of a scene, such as camera poses, depth maps, point maps, and 3D point tracks, from one or multiple input views within seconds. VGGT offers an efficient and unified approach for extracting comprehensive 3D information from visual data, making it a valuable resource for researchers and practitioners in computer vision and 3D scene understanding.

LeRobot

LeRobot is a state-of-the-art machine learning framework for real-world robotics, developed in PyTorch. This repository offers a comprehensive suite of models, datasets, and tools specifically tailored for robotics applications, with a strong emphasis on imitation learning and reinforcement learning. By supporting the sharing of datasets and pretrained models through the Hugging Face ecosystem, LeRobot aims to lower the barrier to entry for robotics research and development, making advanced robotics techniques more accessible to the broader machine learning community.

YouTube Transcript API

YouTube Transcript API is a lightweight and efficient Python library designed for retrieving transcripts and subtitles from YouTube videos without the need for Selenium or a headless browser. The library provides a comprehensive set of features, allowing developers to retrieve and list available transcripts for a video, filter transcripts by language, translate them into different languages, and format the output in various formats such as JSON, SRT, and WebVTT. Additionally, YouTube Transcript API supports advanced functionalities like bypassing IP blocks through proxy configurations and accessing age-restricted videos using cookie authentication, making it a versatile tool for transcript retrieval tasks.

FrankenPHP

FrankenPHP is a cutting-edge PHP application server that integrates the PHP runtime directly into a Go application, leveraging the Caddy web server as its foundation. By embedding PHP’s Zend Engine within Go using CGO, FrankenPHP eliminates the traditional process boundary between web server and PHP interpreter, resulting in high performance and full PHP compatibility. It serves as both a modern replacement for PHP-FPM/Apache setups and a Go library for embedding PHP into custom applications. Key use cases include high-performance PHP serving with worker mode, self-contained deployment via static binaries, seamless embedding of PHP functionality in Go applications using net/http, and support for advanced web server features such as HTTP/2, HTTP/3, and automatic HTTPS for PHP apps.