Projects

### Overview: CCCC is a powerful coordination tool for coding agents, offering durable collaboration state, explicit message status, and seamless integration with Web and IM access. With a single install command, CCCC brings together multiple coding agents into a persistent, coordinated system, ensuring reliable and durable event history, reliable messaging, and a unified control plane. ### Key Features: - **Durable Event History**: Messages and collaboration events are recorded in an append-only ledger (`ledger.jsonl`) for replay and audit. - **Reliable Messaging**: Messages and delivery facts are sent and delivered separately, ensuring runtime handoffs do not pretend messages were read. - **Unified Control Plane**: All communication tools (Web UI, CLI, MCP, IM bridges) operate on the same daemon, ensuring state fragmentation. - **Multi-Runtime Orchestration**: Mix supported coding-agent runtimes in one Group, with `custom` for other command-line agents. - **CCCC Connect**: Connect your own instances, selected Groups across member accounts, or two Groups directly without an account. - **Workspace Tools**: Browse and edit files, inspect Git changes, pin documents in Presentation, and operate tiled native terminals. - **Voice Workflows**: A global Voice Secretary handles speech, documents, and composer drafts for all Groups in a shared conversation. - **Role-Based Coordination**: Foreman and peer model for role-based coordination. ### Tech Stack: Rust 1.88+ ### Why It Matters: CCCC simplifies the coordination of multiple coding agents by providing a single install command, durable collaboration state, and seamless integration with Web

### Overview: Awesome-AI-Memory 是一个集中式、持续更新的 AI 记忆知识库，系统性整理了与大模型记忆（LLM Memory）与智能体记忆（Agent Memory）相关的前沿研究、工程框架、系统设计、评测基准与真实应用实践。 ### Key features: - 收录智能体的长期、情节、语义、共享及经验形成的流程记忆； - 收录语言模型中的存储、保持、寻址、召回与遗忘机制； - 收录针对这些机制的评测、安全研究及综述。 ### Tech stack: Python ### Why it matters: 为突破上下文窗口的固有限制，通过引入参数之外的外部、可持久化且可控的记忆结构，使大模型能够在生成过程中存储、检索、压缩和管理历史信息，从而在有限上下文中持续利用长期经验，实现跨会话的一致性与连续推理能力。

### Overview: Pashov Audit Group Skills is a collection of AI-powered tools designed to enhance security audits of Solidity code. These tools are built by the Pashov Audit Group and leverage cutting-edge AI technologies to identify and mitigate security vulnerabilities. ### Key Features: - **Fuzzing (fizz)**: Generates a comprehensive fuzz suite for Foundry or Hardhat projects, covering a wide range of security vulnerabilities. - **Solidity Auditor**: Conducts an AI security audit of Solidity code, identifying hundreds of critical and high-level vulnerabilities. - **X-ray**: A tool for performing a thorough security audit of Solidity code, focusing on identifying potential security issues. ### Tech Stack: The project is built using JavaScript and leverages various AI platforms and tools, including Claude Code, Cursor, Codex, and GitHub Copilot. ### Why It Matters: These tools are essential for developers looking to enhance the security of their Solidity projects. By automating the security audit process, they help identify vulnerabilities that might otherwise go unnoticed, ensuring the integrity and security of the blockchain applications they develop. ### GitHub Stars: The repository has received 1,282 stars from the developer community, indicating a high level of interest and engagement in the project. ### Repository: The repository is hosted on GitHub and can be found at [pashov/skills](https://github.com/pashov/skills). ### Primary Language: The primary language used in the project is JavaScript. ### Installation, Run, and Update: To use these tools, developers can install the repository and run the following

### Overview: Join a time-traveling adventure where you meet history’s legends while learning Generative AI technologies! 🚀 ### Key features: - **Generative AI Technologies**: Dive into the world of Generative AI and understand its potential for your applications. - **Epic Time-Travel Stories**: Engage with iconic figures like Leonardo da Vinci, Ada Lovelace, and Montezuma, each lesson bringing a unique tale to life. - **Companion App**: Interact with historical figures using Generative AI technologies, enhancing your learning experience with audio tags. - **Accessibility**: Read it, hear it—audio tags bring it to life. ### Tech stack: JavaScript ### Why it matters: This course is designed for developers looking to integrate Generative AI into their JavaScript applications. It provides a fun and interactive way to learn about Generative AI and its applications.

Overview: Prompty is a markdown file format designed for creating, managing, debugging, and evaluating Large Language Model (LLM) prompts. It simplifies the process by aligning generated model types and conformance vectors across different runtimes. Key features: - **Markdown File Format (.prompty)**: A simple markdown file format for LLM prompts. - **Shared `.prompty` Format**: Prompty's shared format ensures consistency across different runtimes. - **Python, TypeScript, Rust, and C# Support**: Prompty supports multiple programming languages. - **OpenAI Provider**: Prompty's `openai` provider can target OpenAI-compatible control planes or self-hosted model servers. - **Tool Loop and Runtime Controls**: Prompty introduces a tool loop for executing model-requested tools and runtime controls for better tool management and runtime management. - **Observable Agents**: Every render, parse, model call, tool call, and final result can flow into traces, enhancing observability. Tech stack: Prompty is written in Rust. Why it matters: Prompty simplifies the process of creating, managing, and evaluating LLM prompts, making it easier for developers to enhance the observability, understandability, and portability of their AI applications. Repository: The repository is hosted on GitHub at [microsoft/prompty](https://github.com/microsoft/prompty). GitHub stars: The repository has received 1,285 stars from the developer community. GitHub description: Prompty is a markdown file format designed for creating, managing, debugging, and evaluating Large Language Model (LL

Overview: This repository hosts articles, lectures, and tutorials on computer vision and OpenMMLab, providing resources for learners to understand algorithms and master the OpenMMLab toolboxes. Key features: - Articles and tutorials on computer vision and OpenMMLab - Lectures on various computer vision topics - Code examples for visualizing and understanding OpenMMLab functionalities Tech stack: Jupyter Notebook Why it matters: This repository serves as a comprehensive resource for computer vision enthusiasts and learners, offering practical insights into the latest algorithms and tools used in the field of computer vision.

Overview: Transformers4Rec is a flexible and efficient library for sequential and session-based recommendation, designed to work with PyTorch. Key features: - Modular building blocks for custom architectures - Access to over 64 Transformer architectures from Hugging Face Transformers - Integration with PyTorch Tech stack: Python Why it matters: Transformers4Rec simplifies the process of implementing and evaluating sequential and session-based recommendation tasks, leveraging the power of transformer architectures and integrating seamlessly with PyTorch.

Overview: This Python SDK allows developers to easily remove backgrounds from images using either local weights or the cloud-based API, providing both free and paid options for different use cases. Key features: - Local background removal using open weights - Cloud-based background removal with improved quality and no local GPU requirements - Supports both offline and online processing - Batch processing for efficient image processing Tech stack: Python Why it matters: This SDK simplifies background removal for developers, offering a balance of free and paid options based on the user's needs, making it suitable for a wide range of applications from small projects to large-scale product development. Repository: withoutbg/withoutbg-python Primary language: Python GitHub stars: 1284 GitHub description: Python SDK for local and cloud background removal (pip install withoutbg)

Overview: Spotlight is an interactive tool designed to help developers quickly explore and understand unstructured datasets from their existing dataframes. It supports various data types including images, audio, text, videos, time-series, and geometric data. Key features: - Interactive visualizations - Data enrichments such as embeddings and predictions - Support for most unstructured data types - Easy integration with existing dataframes - Documentation, playbooks, and API reference available Tech stack: TypeScript Why it matters: Spotlight simplifies the process of understanding complex unstructured data by providing intuitive visualizations and leveraging data enrichments, making it a valuable tool for machine learning and engineering teams. Repository: Renumics/spotlight GitHub stars: 1282 GitHub description: Interactively explore unstructured datasets from your dataframe.

Overview: Search2AI is a self-hosted search gateway designed for AI agents and clients, offering unified search, news, and page extraction with provider fallbacks. It supports REST API, MCP server, and TypeScript library. Key features: - Unified search, news, and page extraction - Configurable providers for various services - Fallback and caching mechanisms - Flexible integration options - Optional chat proxy for model integration Tech stack: TypeScript Why it matters: Search2AI provides a reliable and scalable solution for integrating search and page extraction services, making it ideal for AI agents and clients that need to access and utilize these services. Repository: fatwang2/search2ai Primary language: TypeScript GitHub stars: 1282 GitHub description: Self-hosted search gateway for AI agents and AI clients. Unified search, news and page extraction with provider fallback. REST API, MCP server and TypeScript library.

Overview: hackingBuddyGPT is an open-source framework designed to help ethical hackers use large language models (LLMs) in 50 lines of code or less. It provides essential groundwork for LLM-driven security testing, including LLM connectivity, target connectors, capability/tool wiring, and structured logging. Key features: - MinimalPrivEscLinux: A minimal 20-line Linux privilege-escalation experiment. - MinimalToolCallPrivEscLinux: A tool-calling experiment that maintains a real chat history and drives the target through function calling. - PrivEscLinux: A full-featured strategy-based Linux privilege-escalation experiment with optional retrieval-augmented generation. Tech stack: Python. Why it matters: hackingBuddyGPT is a go-to framework for security researchers and pen-testers who want to use LLMs for security testing, providing a structured way to express experiments in a few dozen lines of code. Repository: ipa-lab/hackingBuddyGPT Primary language: Python GitHub stars: 1282 GitHub description: Helping Ethical Hackers use LLMs in 50 Lines of Code or less. GitHub description: "Helping Ethical Hackers use LLMs in 50 Lines of Code or less…" README excerpt: <div align="center"> <img src="https://github.com/ipa-lab/hackingBuddyGPT/blob/main/docs/hackingbuddy-rounded.png?raw=true" width="96" alt="hackingBuddyGPT logo"> <

### Overview: TrustRAG is a configurable and modular Retrieval-Augmented Generation (RAG) framework designed to provide reliable input and trusted output, ensuring high-quality and trustworthy results in retrieval-based question-answering scenarios. ### Key Features: - **Reliable Input, Trusted Output**: Ensures users obtain high-quality and trustworthy results. - **Configurable and Modular Design**: Allows users to flexibly adjust and optimize each component according to specific needs. - **Deep Information Search and Processing**: Utilizes layered queries, recursive iteration, and intelligent decision-making through deep research. - **Intent Understanding**: Parses user queries into sub-queries for precise understanding. - **Processing Condition Judgment**: Determines whether to continue execution based on token budget and action depth. - **Recursive Execution Steps**: Performs information retrieval, model reasoning, and context processing tasks. - **Action Type Determination**: Decides the next type of action to execute, affecting the context and continuously updating the system state. - **Result Feedback**: Performs corresponding tasks and returns results to the user. ### Tech Stack: Python ### Why it Matters: TrustRAG provides a reliable and trustworthy solution for retrieval-based question-answering scenarios, ensuring users can obtain high-quality and trustworthy results.

### Overview: This repository is an awesome list of attention mechanisms used in computer vision, including plug-and-play modules. It aims to provide a comprehensive overview of attention techniques and their applications in various computer vision tasks. ### Key features: - **Attention Mechanism**: Various attention mechanisms such as Squeeze and Excitation Network, Global Second-order Pooling Convolutional Networks, Neural Architecture Search for Lightweight Non-Local Networks, and others. - **Plug and Play Module**: A collection of pre-trained modules that can be easily integrated into existing projects. - **Vision Transformer**: A module specifically designed for vision tasks, leveraging transformers for better performance. ### Tech stack: Python ### Why it matters: This repository serves as an invaluable resource for developers looking to enhance their computer vision projects with attention mechanisms. It provides a curated list of attention techniques and modules, making it easier to implement advanced features in their projects.

### Overview: 基于医药知识图谱的智能问答系统，是一个基于Python模块REfO实现的知识库问答初级系统。该系统能够解析输入的自然语言问题生成SPARQL查询，进一步请求后台基于TDB知识库的Apache Jena Fuseki服务，从而得到问题的结果。 ### Key features: - SPARQL查询生成 - 疾病症状、疾病用药、药品查询功能 - 提供疾病症状、疾病用药、药品查询等功能 - 支持一问一答式的对话 - 支持查询知识库有的数据，不包含的数据则查询不到 - 页面UI设计简陋 ### Tech stack: JavaScript ### Why it matters: 基于医药知识图谱的智能问答系统旨在通过解析自然语言问题并生成SPARQL查询，从而从TDB知识库中获取所需信息，为用户提供便捷的医药知识查询服务。

### Overview: This repository contains my paper reading notes on deep learning and machine learning, inspired by Denny Britz's and Daniel Takeshi's projects. It serves as a minimalistic webpage with notes on various papers and topics related to autonomous driving and deep learning. ### Key features: - **GitHub Stars:** 1279 - **Repository:** patrick-llgc/Learning-Deep-Learning - **Primary Language:** Jupyter Notebook ### Tech Stack: Jupyter Notebook ### Why it matters: This repository serves as a resource for those new to deep learning in computer vision, offering a structured way to learn from the latest research papers in the field.

Overview: Chubby Skills 是一套包含 14 个 AI 技能的工具，旨在帮助内容创作者和知识工作者整理和管理他们的信息流。这些技能包括图像/视频转文字、本地文档管理、RSS/YouTube 订阅调度、每日情报简报以及一个知识库服务器。通过这些技能，用户可以轻松地保存和整理他们的内容，并在写作时找到原文和来源。 Key features: - 图像/视频转文字 - 本地文档管理 - RSS/YouTube 订阅调度 - 每日情报简报 - 知识库 MCP 服务 Tech stack: Python Why it matters: 通过这些技能，内容创作者和知识工作者能够更高效地整理和管理他们的信息流，提高写作效率，并确保他们的内容来源的准确性。 Repository: https://github.com/chubbyguan/chubbyskills GitHub stars: 1275 GitHub description: 把中文全渠道内容（抖音 / B站 / 小红书 / 公众号 / X / 播客）采集进个人知识库的 14 个 AI Skill：图文存图、视频转文字稿、字幕优先免 GPU、RSS/YouTube 订阅调度与每日情报简报，附带知识库 MCP server。｜ Ingest Chinese content into your personal knowledge base — image/video routing, subtitle-first transcription, feed subscriptions with daily digests, and a KB MCP server.

### Overview: Basaran is an open-source alternative to the OpenAI text completion API, providing a compatible streaming API for Hugging Face Transformers-based text generation models. It supports various decoding strategies, encoder-decoder models, and real-time progress tracking. ### Key Features: - **Streaming Generation**: Uses various decoding strategies for efficient text generation. - **Support for Decoder-Only and Encoder-Decoder Models**: Ensures compatibility with different model architectures. - **Detokenizer**: Handles surrogates and whitespace for seamless text processing. - **Multi-GPU Support**: Optional quantization for improved performance. - **Real-Time Progress**: Utilizes server-sent events for real-time progress tracking. - **OpenAI API and Client Libraries Compatibility**: Ensures seamless integration with existing OpenAI tools. - **Fancy Web-Based Playground**: Provides a user-friendly interface for experimenting with Basaran. ### Tech Stack: Python ### Why It Matters: Basaran allows developers to replace OpenAI's service with the latest open-source model, powering their applications without modifying a single line of code. This opens up new possibilities for large language models, making them more accessible and efficient.

### Overview: This repository is a curated list of research papers, projects, and resources related to the application of **Agentic AI / AI agents for healthcare**, including medical image analysis, EHR manipulation, counseling, drug discovery, patient dialogue, and healthcare administration. AI agents refer to artificial intelligence systems that can autonomously perform tasks, make decisions, and interact with their environment, often through the use of large language models (LLMs), multi-agent systems, and tool integrations. ### Key features: - Comprehensive conceptual framework - Quantitative analysis of recent academic literature - Detailed overview of top data modalities, technologies, and application domains ### Tech stack: Unknown ### Why it matters: This repository serves as a valuable resource for researchers, developers, and healthcare professionals interested in the latest advancements in the field of AI agents for healthcare. It provides a holistic view of the current landscape, technological underpinnings, and application hotspots, making it an essential tool for staying updated and informed about the latest developments in this rapidly evolving field.

### Overview: CityGaussian is a series of works focused on high-quality large-scale scene reconstruction using Gaussian splatting. This project aims to provide efficient and geometrically accurate reconstruction for large-scale scenes, with real-time high-quality rendering capabilities. ### Key features: - **CityGaussianV2**: Efficient and geometrically accurate reconstruction for large-scale scenes. - **CityGaussian**: Real-time high-quality large-scale scene rendering with Gaussians. - **2DGS-style mesh extraction and large-scale scene geometric performance evaluation**. - **Trajectory aligned rendering and mesh video generation with floater removal**. - **Joint pose and 3DGS optimization for imperfect COLMAP results**. - **Features of Gaussian Lightning**: Web viewer, MipSplatting, AbsGS, StopThePop. ### Tech stack: Jupyter Notebook ### Why it matters: This project offers efficient and accurate reconstruction methods for large-scale scenes, making it valuable for applications requiring high-quality rendering and reconstruction of complex environments.

### Overview: AlpaSim is an open-source autonomous vehicle simulation platform designed for development and testing of end-to-end AV policies. It provides a modular and extensible testbed for realistic sensor data, vehicle dynamics, and traffic scenarios, enabling users to validate algorithms, evaluate safety, benchmark performance, and debug complex behaviors. ### Key Features: - **Sensor Fidelity**: Pluggable renderer service with NuRec and OmniDreams support, high-fidelity camera feeds, and stateful video-model rendering. - **Research Hackability**: Python-based implementation, modular grpc interface, extensive configuration options, and debugging tools. - **Horizontal Scalability**: Microservices architecture, support for distributed computing, and multi-node deployments. ### Tech Stack: Python ### Why It Matters: AlpaSim is essential for researchers and developers looking to test and validate autonomous driving algorithms in a controlled and realistic environment. Its modular design and high-fidelity simulation capabilities make it a valuable tool for both algorithm validation and safety analysis in autonomous driving research.