Projects

Overview: This repository is a collection of 161+ specialized Claude Code subagents designed for various development tasks, covering 10 categories. Key features: - Language specialists - Infrastructure & DevOps - Security & Compliance - User Interface Design - Natural Language Processing - Web Scraping & Data Extraction - Machine Learning & AI - Database Management - Testing & Quality Assurance - Deployment & Monitoring - Miscellaneous & Utilities Tech stack: Shell Why it matters: This comprehensive collection of subagents is essential for developers looking to streamline their workflow with tailored AI solutions.

### Overview: Distilly is a tool designed to distill the experience, judgment, voice, and ways of working of individuals into reusable skills for AI agents and bots. Formerly known as Colleague Skill, it helps organizations capture and share knowledge in a structured format. ### Key Features: - **Source-Grounded Person Profiles**: Converts materials into a portable, source-grounded Person Profile. - **Installable Skills**: Available as Agent Skills that can be installed and used by compatible bots. - **Supports Multiple Types**: Suitable for colleagues, partners, family, old friends, idols, public figures, and fictional characters. - **Reusable Output**: Provides a source-grounded Person Profile that can be used as an Agent Skill. ### Tech Stack: Python ### Why It Matters: Distilly simplifies the process of capturing and sharing knowledge, making it easier for organizations to document and utilize the experiences and insights of their colleagues and other individuals. This tool helps in creating a consistent and shareable representation of people's experiences, making it easier for agents and bots to interact with and learn from these insights.

Overview: Stagehand is a powerful SDK designed to extract structured data from any web site using the browser as a tool. It simplifies the process of interacting with websites and automating tasks by leveraging the browser's capabilities. Key features: - Extract structured data from web sites - Interact with any site on the web - Support for multiple programming languages (TypeScript, Python, Go) - Seamless integration with Claude Code, Codex, Eve, and more - Easy setup and deployment Tech stack: TypeScript Why it matters: Stagehand simplifies web scraping and automation tasks, making it easier for developers to extract valuable data from various websites without needing to write complex code. Its versatility across multiple programming languages and its integration with popular AI models make it a valuable tool for developers working on web applications and APIs. Repository: browserbase/stagehand Primary language: TypeScript GitHub stars: 25,719 GitHub description: The SDK to extract data and interact with any site on the web. Get started with Claude Code, Codex, Eve, and more. README excerpt: <div id="toc" align="center" style="margin-bottom: 0;"> <ul style="list-style: none; margin: 0; padding: 0;"> <a href="https://stagehand.dev"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/browserbase/stagehand/main/media/dark_logo.png" /> <img alt="Stagehand" src="

Overview: This project provides a collection of Jupyter notebooks that serve as a comprehensive guide to machine learning fundamentals, based on the book "Hands-on Machine Learning with Scikit-Learn and TensorFlow" by Aurélien Géron. The notebooks cover various machine learning algorithms and techniques, including regression, classification, clustering, and more. Key features: - Machine Learning Fundamentals - Scikit-Learn and TensorFlow Integration - Hands-on Exercises and Solutions - Updated for the Third Edition of the Book Tech stack: Jupyter Notebook Why it matters: This project is essential for anyone looking to learn and understand the practical applications of machine learning using Python. The notebooks provide a practical approach to learning by including exercises and solutions, making it a valuable resource for both beginners and advanced learners. Repository: ageron/handson-ml Primary language: Jupyter Notebook GitHub stars: 25603 GitHub description: This project aims to teach the fundamentals of machine learning in Python. It contains the example code and solutions to the exercises in the book "Hands-on Machine Learning with Scikit-Learn and TensorFlow." The repository is deprecated and can be found at https://github.com/ageron/handson-ml3 or https://github.com/ageron/handson-mlp, and there are also translations available.

### Overview: A simple yet effective implementation of Vision Transformer (ViT) in PyTorch, designed to achieve state-of-the-art performance in vision classification tasks using a single transformer encoder. ### Key Features: - **Single Transformer Encoder**: Utilizes a single transformer encoder to achieve state-of-the-art results. - **Efficient Attention Mechanism**: Optimized attention mechanism for faster inference. - **Pretrained Models**: Available pretrained models for easy integration. - **Flexible Configuration**: Supports various configurations such as different patch sizes, image sizes, and more. ### Tech Stack: PyTorch ### Why It Matters: This implementation simplifies the process of achieving state-of-the-art performance in vision classification tasks by leveraging a single transformer encoder, making it accessible and efficient for developers and researchers alike. ### Repository: [lucidrains/vit-pytorch](https://github.com/lucidrains/vit-pytorch) ### GitHub Stars: 25553 ### Table of Contents: - **Vision Transformer - Pytorch** - Installation - Usage - Parameters - Simple ViT - NaViT - Distillation - Deep ViT - CaiT - Token-to-Token ViT - CCT - Cross ViT - PiT - LeViT - CViT - Twins SVT - CrossFormer - RegionViT - ScalableViT - SepViT

### Overview Language serves as an interface for Large Language Models (LLMs) to connect various AI models for solving complex AI tasks! ### Key Features - **Easytool**: Simplifies tool usage with easy-to-understand instructions. - **TaskBench**: Evaluates the automation capabilities of LLMs. - **Light LangChain Version**: Supports the OpenAI service on Azure and GPT-4. - **CLI Mode**: Offers a lightweight experience without local model deployment. - **Gradio Demo**: Provides a user-friendly interface for evaluating model performance. ### Tech Stack Python ### Why It Matters JARVIS aims to explore artificial general intelligence (AGI) and share cutting-edge research with the AI community, aiming to improve the efficiency and effectiveness of AI tasks.

### Overview: WaveFunctionCollapse is a program that generates bitmaps by mimicking the principles of quantum mechanics. It operates by initializing an output bitmap in a superposition of colors from the input bitmap, then observing and propagating the effects of the collapse to produce a final, localized pattern. ### Key Features: - **Local Similarity**: The output bitmap contains only the patterns present in the input, ensuring local similarity. - **Entropy Minimization**: The algorithm chooses the region with the lowest entropy for collapse, ensuring efficient use of information. - **Observation-Propagation Cycle**: The program alternates between observing and propagating the effects of the collapse, gradually collapsing the wave function. ### Tech Stack: - **Programming Languages**: C#, Python, Kotlin, Rust, Julia, Go, Haxe, Java, Clojure, Free Pascal, Dart, p5.js, Unity, Unreal Engine 5, Godot 4, Houdini. ### Why It Matters: - **Practical Applications**: WaveFunctionCollapse has been implemented in various programming languages, demonstrating its versatility and effectiveness. - **Algorithmic Insights**: The program provides insights into how quantum mechanics can be simulated in a deterministic way, offering a unique approach to bitmap generation. - **Real-World Applications**: The algorithm has been used in various game and level generation projects, showcasing its practical utility in creating visually appealing and locally similar patterns. ### GitHub Stars: - **GitHub Stars**: 25,405 ### Repository: - **Repository**: mxgmn/WaveFunctionCollapse ### How to Build: -

### Overview: The fastai book, published as Jupyter Notebooks, serves as an introduction to deep learning and fastai, a layered API for deep learning. It is used for a MOOC and forms the basis of a book available for purchase, offering practical examples and explanations for learning deep learning. ### Key features: - **Jupyter Notebooks**: The entire content is presented as Jupyter Notebooks. - **Introduction to Deep Learning and fastai**: Covers the basics of deep learning and fastai. - **Practical Examples**: Includes practical examples and explanations to help users learn effectively. ### Tech stack: - **Python**: The primary programming language used. - **Jupyter Notebook**: Used for the presentation and execution of the notebooks. ### Why it matters: - **Practical Learning**: The notebooks provide practical examples and explanations, making it easier for users to learn deep learning. - **Community Contribution**: The materials are freely available and contributed by the community, making it accessible to a wide audience. - **MOOC Foundation**: The notebooks form the basis of a book that is currently available for purchase, providing a comprehensive resource for deep learning learners. ### GitHub Stars: - **25,392**: The repository has received a significant number of stars, indicating its popularity and usefulness in the community. ### Repository: - **GitHub Repository**: The repository is hosted on GitHub, accessible at <https://github.com/fastai/fastbook>. ### Primary Language: - **Jupyter Notebook**: The primary language used in the notebooks. ### Why it Matters: -

### Overview: This repository contains PyTorch implementations of CycleGAN and pix2pix, two state-of-the-art image-to-image translation models. It supports both paired and unpaired training and provides improved results using pre-trained StableDiffusion-Turbo models. The code is compatible with PyTorch 2.4+ and supports DDP for single-machine multiple-GPU training. ### Key Features: - **CycleGAN**: Implements Cycle-Consistent Adversarial Networks for image-to-image translation. - **pix2pix**: Supports Conditional Adversarial Networks for image-to-image translation. ### Tech Stack: - Python 3.11 - PyTorch 2.4 ### Why It Matters: This code offers comparable or better results to the original Torch implementations, making it a valuable resource for researchers and developers looking to improve image-to-image translation tasks.

### Overview: 本项目旨在分享大模型相关技术原理以及实战经验，涵盖大模型工程化、大模型应用落地等方面。通过分享从6B到65B的训练教程，从全量微调到高效微调（LoRA，QLoRA，P-Tuning v2），再到RLHF（基于人工反馈的强化学习），展示了大模型训练的全过程。 ### Key features: - 全量微调 - 高效微调（LoRA, QLoRA, P-Tuning v2） - RLHF（基于人工反馈的强化学习） - 分布式训练并行技术 - 分布式训练网络通信 - LLM训练优化技术 - LLM对齐技术 ### Tech stack: Python, PyTorch, Hugging Face Transformers, TensorFlow, DGL, PyTorch Lightning, Transformers, Hugging Face, FastAPI, Docker, Kubernetes, Git, GitHub Actions, AWS, Google Cloud, Azure, Docker, Kubernetes, Git, GitHub Actions, AWS, Google Cloud, Azure, Docker, Kubernetes, Git, GitHub Actions, AWS, Google Cloud, Azure, Docker, Kubernetes, Git, GitHub Actions, AWS, Google Cloud, Azure, Docker, Kubernetes, Git, GitHub Actions, AWS, Google Cloud, Azure, Docker, Kubernetes, Git, GitHub Actions, AWS, Google Cloud, Azure, Docker, Kubernetes, Git, GitHub Actions, AWS, Google Cloud, Azure, Docker, Kubernetes, Git, GitHub Actions, AWS, Google Cloud, Azure, Docker

Overview: Letta is a platform for building stateful agents that can learn and improve over time. Key features: - Advanced memory capabilities - Learning and self-improvement - Active development and community support Tech stack: Node.js Why it matters: Letta provides a robust solution for creating intelligent agents that can adapt and evolve with experience, making them invaluable in various applications. Repository: letta-ai/letta Primary language: Node.js GitHub stars: 25,174

### Overview: Activepieces is an open-source AI automation platform designed for developers and non-developers alike. It offers a type-safe pieces framework written in TypeScript, enabling seamless integration with large language models like Claude Desktop, Cursor, and Windsurf. ### Key Features: - **Type-safe Pieces Framework**: Pieces are npm packages written in TypeScript, ensuring full customization and hot reloading for local development. - **Community Contributions**: 60% of pieces are contributed by the community, making it a large open-source MCP toolkit. - **Native AI Pieces**: Native AI pieces allow experimentation with various providers or creation of custom agents using the AI SDK. - **Extensible and Secure**: Supports loops, branches, auto retries, HTTP, and more. Secure by design, self-hosted and network-gapped. - **Human-in-the-Loop**: Supports human input triggers like Chat Interface and Form Interface. - **Versioned Flows**: Flows are fully versioned, ensuring data integrity. - **200+ Pieces**: Check out the full list of supported integrations on the website. ### Tech Stack: TypeScript ### Why it Matters: Activepieces offers an intuitive interface, great experience, and a large open-source MCP toolkit, making it an excellent choice for developers and non-developers alike. The platform supports integrations with Google Sheets, OpenAI, Discord, RSS, and over 200 other services, expanding continuously with community contributions.

### Overview: LLaVA is a state-of-the-art large language and vision assistant that leverages GPT-4-level capabilities and beyond. It excels in visual instruction tuning, enabling powerful visual reasoning and application across various domains. ### Key Features: - **Visual Instruction Tuning**: Fine-tuned towards GPT-4-level capabilities. - **LLaVA-NeXT**: Stronger models with support for LLMs like LLaVA-34B and Qwen-1.5. - **LLaVA-NeXT Video**: Strong on video tasks with zero-shot modality transfer. - **LMMs-Eval**: Highly efficient evaluation pipeline for LMMs. ### Tech Stack: Python ### Why It Matters: LLaVA is a groundbreaking project that pushes the boundaries of language and vision capabilities, making it a pivotal tool for researchers and developers alike.

### Overview: LabelImg is a graphical image annotation tool written in Python, utilizing Qt for its user interface. It supports annotations in PASCAL VOC, YOLO, and CreateML formats, making it a versatile tool for image data labeling. ### Key Features: - **Graphical Interface**: Easy-to-use interface for image annotation. - **Annotation Formats**: Supports PASCAL VOC, YOLO, and CreateML formats for various image annotation tasks. - **Python Integration**: Built with Python, leveraging Qt for a robust and user-friendly experience. ### Tech Stack: - **Python**: The primary programming language used for development. - **Qt**: Framework for creating graphical user interfaces. ### Why It Matters: LabelImg is a valuable tool for developers and data scientists looking to annotate images efficiently. Its compatibility with multiple annotation formats and ease of use make it a preferred choice for various applications requiring image data labeling.

### Overview: Pascal Editor is an open-source, local-first 3D building editor built with React Three Fiber and WebGPU. It supports running the editor in the browser or from the command line, and connects AI agents through MCP. ### Key features: - Local-first development workflow - Supports React Three Fiber and WebGPU for rendering - Runs in the browser or from the command line - Connects AI agents through MCP - Built with TypeScript ### Tech stack: TypeScript ### Why it matters: Pascal Editor offers a practical and human-friendly approach to 3D building design, making it accessible to both humans and AI agents. Its local-first development and AI integration capabilities make it a valuable tool for facility management, home services, architecture, and infrastructure.

### Overview: This repository contains examples of popular machine learning algorithms implemented in Python, with mathematics behind them explained. Each algorithm has an interactive Jupyter Notebook demo that allows you to play with training data, algorithm configurations, and see results, charts, and predictions right in your browser. The focus is on understanding the mathematics behind each algorithm, rather than using third-party libraries. ### Key Features: - 📗 Math | Linear Regression: Theory and links for further readings. - ⚙️ Code | Linear Regression: Implementation example. - ▶️ Demo | Univariate Linear Regression: Predict country happiness score by economy GDP. - ▶️ Demo | Multivariate Linear Regression: Predict country happiness score by economy GDP and freedom index. - ▶️ Demo | Non-linear Regression: Use linear regression with polynomial and sinusoid features to predict non-linear dependencies. ### Tech Stack: Python ### Why it Matters: The purpose of this repository is not to implement machine learning algorithms by using third-party library one-liners, but rather to practice implementing these algorithms from scratch and gain a better understanding of the mathematics behind each algorithm. This approach helps in developing a deeper understanding of the underlying concepts and improving one's ability to implement and interpret machine learning models.

Overview: Dolt is a Git-like SQL database that allows developers to manage and version their data just like they do with Git. Key features: - Git-like command line interface for version control - SQL-based data manipulation - Importing CSV files and committing changes - Pushing and pulling changes to a remote repository - Version control files (Git) and versions tables (Dolt) Tech stack: Go Why it matters: Dolt provides a powerful tool for managing and versioning data, making it an essential tool for developers working with structured data. Repository: dolthub/dolt GitHub stars: 24662 GitHub description: Dolt – Git for Data README excerpt: <img height="100" src="./images/Dolt-Logo@3x.svg"/>

### Overview: WhisperX is an automatic speech recognition (ASR) tool that provides word-level timestamps and speaker diarization for fast and accurate transcription of audio meetings. It leverages Whisper, an ASR model developed by OpenAI, and improves upon its timestamp accuracy and performance. ### Key Features: - **Batched Inference**: WhisperX achieves 70x real-time transcription using Whisper large-v2, with a requirement of <8GB GPU memory for large-v2 with beam_size=5. - **Word-Level Timestamps**: Utilizes wav2vec2 alignment to provide accurate word-level timestamps. - **Multispeaker ASR**: Uses speaker diarization from PyAudioAnote-Audio for improved speaker identification and accuracy. - **VAD Preprocessing**: Reduces hallucination and batching without compromising WER, ensuring high-quality transcription. ### Tech Stack: Whisper, PyAudioAnote-Audio, and faster-whisper ### Why It Matters: WhisperX offers a significant improvement in transcription accuracy and speed, making it an essential tool for meeting transcriptions and other audio-based communication tasks. Its ability to provide word-level timestamps and speaker diarization makes it highly valuable for applications requiring precise and accurate transcriptions.

Overview: This repository contains 50+ tutorials and implementations for Generative AI Agent techniques, covering a wide range from basic conversational bots to complex multi-agent systems. Key features: - Tutorials and implementations for Generative AI Agents - A diverse set of agent types including conversational bots, multi-agent systems, and more - Comprehensive documentation and examples Tech stack: Jupyter Notebook Why it matters: This repository serves as an essential resource for developers looking to learn and implement Generative AI Agents in their projects, making it a valuable asset for anyone interested in AI and automation.

### Overview: NocoBase is an open-source AI + no-code platform designed to build business systems quickly and reliably. Unlike traditional methods, it leverages AI to work on top of existing infrastructure, providing a WYSIWYG no-code interface for both speed and reliability. ### Key Features: - **Collaborative Development**: Coding agents can integrate with existing infrastructure, while people can use a WYSIWYG interface, allowing for efficient collaboration. - **AI Integration**: AI agents can handle setup, development, migration, and release processes end-to-end, ensuring seamless integration. - **WYSIWYG Interface**: People can build and modify systems visually, without needing AI skills. - **AI Employees**: AI employees can be integrated into business workflows, providing real-time business context and enabling intelligent decision-making. ### Tech Stack: TypeScript ### Why It Matters: NocoBase offers a scalable, reliable, and efficient way to build business systems, combining the speed of no-code interfaces with the reliability of AI-driven development. This approach ensures that businesses can quickly and effectively implement new systems without the need for extensive coding expertise. Repository: nocobase/nocobase GitHub Stars: 24,607