Projects

Overview: GraphJin is a compiler and runtime system designed to give AI agents a smart, auditable way to work across databases, files, APIs, and code. Key features: - One governed graph over various systems - Built-in agent with built-in agent, one call away - Smart discovery before action - Guarded action, not raw access - Operational awareness - Standing questions, not just answers - Durable memory for saved queries, fragments, and workflows Tech stack: Go Why it matters: GraphJin simplifies the process of integrating AI agents into existing systems by providing a unified graph-based approach to access and manage data, code, and operations. It ensures that agents can operate safely and audibly, without needing to guess at permissions or access to sensitive information. Repository: dosco/graphjin Primary language: Go GitHub stars: 3216 GitHub description: One governed graph for AI agents — GraphQL + MCP over your databases, files, APIs, and code README excerpt: # GraphJin — One Governed Graph for Your AI Agents [](https://github.com/dosco/graphjin/blob/master/LICENSE) [](https://www.npmjs.com/package/graphjin) [![Docker Pulls](https://img.shields.io/docker/pulls/dos

Overview: AGiXT is a comprehensive AI automation platform designed to bridge the gap between artificial intelligence and real-world applications. It offers 40+ built-in extensions, multi-provider support, enterprise-grade features, and natural language control, making it a central nervous system for digital and physical environments. Key features: - 40+ built-in extensions - Multi-provider support - Enterprise-Ready features - Natural Language Control - Workflow Automation - Real-Time Integration Tech stack: Python Why it matters: AGiXT simplifies the integration and management of AI solutions, making it a versatile tool for developers and businesses looking to automate complex systems through natural language interactions.

### Overview: LLM Guard is a comprehensive tool designed to enhance the security of Large Language Models (LLMs). ### Key features: - **Prompt Scanners**: Anonymize, BanCode, BanCompetitors, BanSubstrings, BanTopics, Code, Gibberish, InvisibleText, Language, PromptInjection, Regex, Secrets, Sentiment, TokenLimit, and Tox

### Overview: CyberStrike is an open-source AI-powered offensive security harness designed for automated penetration testing. It leverages cutting-edge AI and machine learning to automate security assessments, providing a comprehensive suite of tools and skills. ### Key Features: - **Automated Penetration Testing**: Turn any LLM subscription into a red team agent with 13+ specialized agents, 7,600+ security skills, and 120+ OWASP test cases. - **150+ AI Providers**: Access to 150+ AI providers, 5,300+ models, and 56+ built-in tools. - **56+ Built-in Tools**: Includes 56+ built-in tools for various security tasks. - **176+ MCP Tools**: Utilizes 176+ MCP tools for more advanced security operations. - **Web UI**: A web-based interface for remote access and management. - **Bolt**: Remote tool execution capabilities. - **MCP Ecosystem**: A robust ecosystem for post-exploitation and further security tasks. ### Tech Stack: TypeScript ### Why It Matters: CyberStrike empowers security professionals to automate their security assessments, making it easier and more efficient to identify vulnerabilities and improve security posture. With its comprehensive suite of tools and skills, CyberStrike transforms the way security is conducted, providing a powerful and versatile solution for red team operations.

Overview: JoePenna/Dreambooth-Stable-Diffusion is an implementation of Dreambooth, a method for training Stable Diffusion models to generate images of specific subjects, objects, and styles. The project focuses on improving the training of faces, objects, and styles, and has been adapted for use on various platforms including Vast.AI, Google Colab, and local PCs. Key features: - Improved training of faces, objects, and styles - Adapted for use on Vast.AI, Google Colab, and local PCs - Configuration file and command line reference Tech stack: Jupyter Notebook Why it matters: This project provides a tool for filmmakers and concept artists to train their own likeness into a Stable Diffusion model, making it easier to generate the seed of an initial idea and communicate visually.

### Overview: Pyramid Flow is an efficient autoregressive video generation method that leverages flow matching to generate high-quality 10-second videos at 768p resolution and 24 FPS. It is trained on open-source datasets and supports image-to-video generation. ### Key features: - **Flow Matching**: Generates high-quality videos by interpolating between different resolutions and noise levels. - **End-to-End Optimization**: Uses a single DiT model for efficient training and generation. - **Training Efficiency**: Capable of generating videos at 768p resolution and 24 FPS using only open-source datasets. ### Tech stack: Python ### Why it matters: Pyramid Flow offers a significant improvement in computational efficiency compared to existing video diffusion models, making it suitable for real-time applications and low-resource environments.

### Overview: QualityScaler is a Windows app powered by AI to enhance, upscale, and de-noise photographs and videos. ### Key features: - **Image and Video Upscaling**: Enhance and upscale images and videos. - **Multiple GPUs Support**: Utilizes multiple GPUs for optimal performance. - **Compatible Formats**: Supports various image and video formats. - **Automatic Image Tiling**: Avoids GPU VRAM limitations by tiling images. - **Resize Before Upscaling**: Resize images before upscaling. - **Interpolation**: Provides interpolation between the original and upscaled files. - **Video Upscaling**: Supports video upscaling with features like STOP & RESUME. - **Privacy-Focused**: Runs entirely on your PC without internet connection. ### Tech stack: - **Backend**: Written in Python using libraries like PyTorch, ONNX, onnxconverter-common, onnxruntime-directml, CustomTkinter, OpenCV, and PyInstaller. - **Frontend**: Built with CustomTkinter for a user-friendly interface. ### Why it matters: QualityScaler is designed to be a powerful tool for enhancing and upscaling images and videos, making it ideal for those looking to improve their digital media without the need for an internet connection or additional hardware. ### Repository: Repository: [Djdefrag/QualityScaler](https://github.com/Djdefrag/QualityScaler) ### GitHub Stars: GitHub Stars: 3210 ### Primary Language: Primary Language: Python ### How to Use QualityScaler: - **GUI Interface**: Accessible through

### Overview: LSP-AI is an open-source language server designed to enhance software development environments by integrating AI-powered functionality. It is particularly useful for software engineers who prefer to use familiar tools like VS Code, NeoVim, Emacs, Helix, and Sublime. LSP-AI offers features such as in-editor chatting with LLMs and code completions, making it a valuable tool for improving productivity and collaboration. ### Key Features: - **In-Editor Chatting**: Chat directly in your codebase with your favorite local or hosted models. - **Custom Actions**: Create custom actions for code refactoring, code completions, and more. - **Code Completions**: An alternative to Github Copilot, offering faster completions with backend support. ### Tech Stack: Rust, leveraging the Language Server Protocol (LSP). ### Why It Matters: LSP-AI aims to bridge the gap between advanced AI-powered development tools and the open-source community, providing a unified backend for editors that can benefit from the latest AI advancements. This approach centralizes development efforts and makes it easier for developers to integrate cutting-edge AI features into their workflows, enhancing productivity and collaboration without the need for proprietary solutions. ### GitHub Stars: 32,100 ### Repository: - **Repository**: SilasMarvin/lsp-ai - **Primary Language**: Rust - **GitHub Stars**: 32,100 ### Documentation: - **Wiki**: https://github.com/SilasMarvin/lsp-ai/wiki ### The Case for LSP-A

Overview: TanStack AI is a type-safe, provider-agnostic TypeScript AI SDK designed for building streaming chat, tool calling, agents, and multimodal apps across various frameworks and providers. Key features: - Streaming chat - Tool calling - Agents - Structured outputs - Multimodal prompts and responses Tech stack: TypeScript Why it matters: TanStack AI simplifies the development of AI-powered applications by providing a composable, provider-agnostic SDK that supports multiple frameworks and providers, making it easy to integrate AI into existing projects.

Overview: WebDataset is a high-performance Python-based I/O system designed for large and small deep learning problems, offering strong support for PyTorch. It supports reading from local disk, cloud object stores, and concatenated MsgPack/CBOR sources, with a focus on achieving high I/O rates and efficient use of object stores. Key features: - WebDataset format files are tar archives containing training samples. - Files within each tar file share the same basename when stripped of extensions. - Tar files are numbered sequentially, from `something-000000.tar` to `something-012345.tar`. - Supports PyTorch `IterableDataset` and can be used with Julia and Golang libraries. Tech stack: Python Why it matters: WebDataset is optimized for large-scale deep learning, providing high I/O rates, efficient use of object stores, and ease of access and processing through its native file formats, making it ideal for training and inference tasks. Repository: webdataset/webdataset Primary language: Python GitHub stars: 3211 GitHub description: A high-performance Python-based I/O system for large (and small) deep learning problems, with strong support for PyTorch.

Overview: Attu is an AI-native management tool designed specifically for Milvus vector databases, providing a comprehensive interface for managing and interacting with these databases. Key features: - Connects to multiple Milvus clusters from a single instance. - Allows browsing collections, running vector searches, managing backups, monitoring health, and chatting with an AI agent. - Supports external collections and snapshots, text and schema evolution, search and analytics, minhash deduplication, and collection and data management. Tech stack: Attu is built using Node.js, with a frontend built with React and a backend API. Why it matters: Attu simplifies the management and interaction with Milvus databases, making it easier for developers and data scientists to work with vector databases efficiently and effectively.

Overview: PentestAgent is an AI agent framework designed for black-box security testing, offering support for bug bounty, red-team, and penetration testing workflows. It utilizes various models like Anthropic and OpenAI for tasks such as vulnerability discovery and exploitation. Key features: - Supports multiple models including Anthropic and OpenAI - Compatible with various pentesting tools like nmap, netcat, curl, and metasploit - Can be run in a Docker container for isolated execution - Provides different modes: Assist, Agent, Crew, and Interact Tech stack: Python Why it matters: PentestAgent simplifies complex security testing by automating tasks and providing insights through AI-driven assistance, making it an essential tool for security professionals. Repository: GH05TCREW/pentestagent GitHub stars: 3202 GitHub description: PentestAgent is an AI agent framework for black-box security testing, supporting bug bounty, red-team, and penetration testing workflows. It utilizes various models like Anthropic and OpenAI for tasks such as vulnerability discovery and exploitation. Key features: - Supports multiple models including Anthropic and OpenAI - Compatible with various pentesting tools like nmap, netcat, curl, and metasploit - Can be run in a Docker container for isolated execution - Provides different modes: Assist, Agent, Crew, and Interact Tech stack: Python Why it matters: PentestAgent simplifies complex security testing by automating tasks and providing insights through AI-driven assistance, making it an essential tool for security professionals.

Overview: InternGPT is an open-source platform that allows users to interact with ChatGPT using pointing devices, enhancing communication efficiency and accuracy in vision-centric tasks. Key features: - DragGAN-based interactive demonstration - ImageBind integration for image-based interactions - Support for multimodal chat like GPT-4 - SAM integration for image editing - Advanced control mechanism for large language models Tech stack: Python Why it matters: InternGPT offers a unique and efficient way to interact with ChatGPT, improving communication and task accuracy in complex visual scenarios. Repository: OpenGVLab/InternGPT GitHub stars: 3203 GitHub description: InternGPT (iGPT) is an open source demo platform where you can easily showcase your AI models. It supports DragGAN, ChatGPT, ImageBind, multimodal chat like GPT-4, SAM, interactive image editing, etc. Try it at igpt.opengvlab.com. README excerpt: [[中文文档]](README_CN.md)

### Overview: This repository contains a comprehensive reading list of 400+ papers on Time Series Forecasting/Prediction (TSF) and Spatio-Temporal Forecasting/Prediction (STF). The papers are categorized by the type of model used, and the repository includes the latest papers from arXiv. ### Key Features: - **Reading List of Papers**: Over 400 papers on TSF and STF. - **Categorized by Model Type**: Papers are categorized according to the type of model used, such as univariate, multivariate, and spatio-temporal forecasting. - **Latest Papers from arXiv**: The repository includes the latest papers from arXiv, ensuring up-to-date information. - **Univariate Forecasting**: Predicts one variable with its history. - **Multivariate Forecasting**: Predicts multiple variables with their histories. - **Spatio-Temporal Forecasting**: Adds a spatial dimension to time series forecasting, often used in traffic and weather forecasting. ### Tech Stack: The repository uses Markdown, GitHub, and LaTeX for documentation and rendering. ### Why It Matters: This repository is a valuable resource for developers and researchers interested in time series forecasting and spatio-temporal forecasting. It provides a clear overview of the current state of the field, including the latest research and methodologies.

Overview: AlphaGPT 是一个基于深度强化学习的开源自动因子工厂项目，旨在通过自动化因子生成提高金融投资效率。 Key features: - 自动因子生成 - 高效的因子管理 - 实时市场分析 Tech stack: Python Why it matters: AlphaGPT 的出现为金融行业提供了自动化因子生成的解决方案，有望显著提高投资效率和决策质量。 Repository: imbue-bit/AlphaGPT Primary language: Python GitHub stars: 3199

Overview: textlint is a pluggable linter designed for natural language text. Key features: - No bundled rules - Supports Markdown and plain text by default - Supports HTML and other file formats via plugins - Allows installing rules globally or locally - Provides a quick tour guide - Supports configuration via `.textlintrc.json` Tech stack: TypeScript Why it matters: textlint is a useful tool for developers to lint their natural language text, ensuring consistent formatting and quality. Repository: textlint/textlint GitHub stars: 3202

Overview: This repository contains Jupyter notebook tutorials designed to help students understand the "theory" from the Deep Learning course taught at the University of Amsterdam (MSc AI), Fall 2023. The notebooks cover various topics such as optimization techniques, transformers, and graph neural networks, and provide one-to-one translations to JAX+Flax for alternative framework implementation. Key features: - Jupyter notebook tutorials - Covers various deep learning topics - One-to-one translations to JAX+Flax - Designed for understanding theory and teaching PyTorch framework - Presented in the first hour of group tutorial sessions - Available on GitHub, RTD website, and YouTube playlist - Integrates with official PyTorch Lightning documentation Tech stack: Jupyter Notebook Why it matters: These notebooks serve as supplementary material for the course, helping students understand the theoretical concepts and learn practical implementation details of PyTorch, including PyTorch Lightning. The tutorials are designed to be accessible on a variety of devices and can be run locally on a CPU, on Google Colab for GPU support, or on the Snellius cluster for training purposes.

Overview: Loopy is a library of practical AI-agent loops designed to help developers automate and improve their workflows. It includes a website for browsing and copying loops, and a companion skill Loopy that guides the agent through the process of finding, adapting, and designing loops. Key features: - A public catalog for browsing and copying loops - A guided workflow for discovering, finding, and adapting loops - A way to interview developers about their goals and craft new loops Tech stack: JavaScript Why it matters: Loopy streamlines the process of automating tasks, making workflows more efficient and reliable. It's particularly useful for developers looking to improve their productivity and consistency in their projects.

### Overview: Eidos is a single-file relational spreadsheet format designed for simplicity, ease of use, and compatibility with both local and cloud-based applications. It is built on SQLite and supports both desktop and mobile environments, making it accessible to developers and users alike. ### Key Features: - **Single-File Relational Spreadsheet:** Eidos allows you to store and manage data in a single file, making it easy to share and collaborate. - **SQLite Backend:** Built on SQLite for efficient data storage and retrieval. - **Desktop App:** Eidos Lite is a desktop application for working with Eidos files and ordinary files in a local folder. - **Android and iOS Apps:** Android and iOS apps bring local Eidos files and Markdown editing capabilities to mobile devices. - **Open Source:** The project is open-source under the AGPL v3 license, ensuring flexibility and community engagement. ### Tech Stack: - TypeScript ### Why It Matters: Eidos is particularly useful for developers who need a lightweight, efficient way to manage and share data across different platforms and devices. Its simplicity and compatibility with both local and cloud-based applications make it a versatile tool for various development and collaboration scenarios.

Overview: The Compute Library is a collection of low-level machine learning functions optimized for Arm Cortex-A, Neoverse, and Mali GPUs, providing superior performance and support for new Arm technologies. Key features: - Open source software available under a permissive MIT license - Over 100 machine learning functions for CPU and GPU - Multiple convolution algorithms (GeMM, Winograd, FFT, Direct, and indirect-GeMM) - Support for multiple data types: FP32, FP16, INT8, UINT8, BFLOAT16 - Micro-architecture optimization for key ML primitives - Highly configurable build options enabling lightweight binaries - Advanced optimization techniques such as kernel fusion, Fast math enablement, and texture utilization - Device and workload specific tuning using OpenCL tuner and GeMM optimized heuristics Tech stack: C++ Why it matters: The Compute Library offers superior performance compared to other open-source alternatives and supports new Arm technologies like SVE2, making it a valuable tool for developers working on Arm-based systems.