Projects

### Overview: Meridian is a comprehensive coding client and API bridge solution that integrates Claude and Antigravity with popular coding tools like Pi and OpenCode. It provides a seamless experience by connecting these clients to the Claude and Antigravity APIs, allowing users to manage and monitor their requests and usage directly through a browser or a dedicated Mac app. ### Key Features: - **Claude and Antigravity Integration:** Meridian connects Claude and Antigravity using your subscription, enabling seamless integration of Claude requests through the Claude Agent SDK and Antigravity requests through the official `agy` CLI. - **Persistent Sessions and Profiles:** Offers persistent sessions, prompt caching, and multiple account profiles, ensuring a smooth and secure experience. - **Client Compatibility and Setup:** Supports Pi, OpenCode, and other compatible clients, with detailed guides for setup and configuration. - **Documentation and Support:** Comprehensive documentation and a dedicated Discord community for support and community engagement. ### Tech Stack: - TypeScript ### Why It Matters: Meridian simplifies the integration of Claude and Antigravity with your existing coding tools, providing a unified interface for managing and monitoring your requests and usage. This solution enhances productivity by eliminating the need for multiple tools and streamlines the process of managing Claude and Antigravity requests.

Overview: Neuron is a PHP framework designed for building production-ready AI-driven applications. It integrates components like LLMs, tools, vector databases, and memory to create agents that interact with data and user interfaces. Key features: - Agent orchestration - Multi-agent systems - Event-driven workflows - Human-in-the-loop capabilities - Streaming and UI integration - MCP connectors - Asynchronous execution Tech stack: PHP Why it matters: Neuron simplifies the creation of agentic applications, making it easier to integrate AI entities into PHP projects, ensuring robust and scalable solutions.

Overview: UniFace is a comprehensive Python library designed for unified face analysis, including detection, alignment, landmarks, parsing, tracking, recognition, and more. It simplifies the process of integrating advanced face analysis capabilities into projects. Key features: - Face Detection - Face Recognition - Face Tracking - Facial Landmarks - Face Parsing - Gaze Estimation - Demographics - Emotion - Face States - Face Quality - Anti-Spoofing - Anonymization Tech stack: Python Why it matters: UniFace simplifies the integration of advanced face analysis tools, making it easier for developers to incorporate face recognition, detection, and analysis into their projects without needing to dive deep into complex algorithms and libraries. Repository: yakhyo/uniface GitHub stars: 2148 Primary language: Python GitHub description: UniFace: A Unified Face Analysis Library for Python | Detection, alignment, landmarks, face-mesh, recognition, parsing, gaze, attributes and anti-spoofing under one API.

Overview: SparseML is a library for applying sparsification recipes to neural networks with just a few lines of code, enabling faster and smaller models. Key features: - Enables faster and smaller models - Supports various sparsification techniques - Integrates with popular deep learning frameworks Tech stack: Python Why it matters: SparseML simplifies the process of creating efficient neural network models, making it easier for developers to build faster and more compact AI solutions.

Overview: LumiBot is an AI-driven trading platform that allows developers to create AI agents that place trades, leveraging real backtests and live broker connections. It supports stocks, options, futures, forex, crypto, and prediction markets, providing a seamless integration with 12 brokers. Key features: - Real backtests with historical data - Live broker connections for real market execution - Support for stocks, options, futures, forex, crypto, and prediction markets - Python-based strategy classes that can be easily edited - Supports both traditional trading strategies and AI agents Tech stack: Python Why it matters: LumiBot simplifies the process of creating and deploying AI-driven trading strategies, making it accessible to developers without requiring extensive knowledge of backtesting or broker configuration. The platform's ability to combine traditional trading logic with AI models makes it a versatile solution for financial applications. Repository: Lumiwealth/lumibot GitHub stars: 2144 GitHub description: AI agents that actually place the trade. 12 brokers, real backtests, stocks options futures forex crypto and prediction markets.

Overview: This repository is designed to introduce the fundamental concepts of Machine Learning to developers through Jupyter Notebooks. It includes slides, Jupyter notebooks, and exercises for practical learning. Key features: - Jupyter Notebooks for interactive learning - Slides for visual representation of course material - Exercises for hands-on practice Tech stack: Jupyter Notebook Why it matters: This repository provides a comprehensive, interactive way to learn Machine Learning, making it ideal for developers seeking to understand the basics of the field.

Overview: Đây là một tài liệu hướng dẫn ngắn gọn về kiến thức cơ bản cần thiết để học tốt Machine Learning trong vòng 2 tháng. Key features: - Giới thiệu về cơ sở dữ liệu và các thuật toán cơ bản trong Machine Learning. - Hướng dẫn thực hành các bài tập nhỏ để nắm vững kiến thức. - Phân tích các mô hình học máy và cách sử dụng chúng. Tech stack: Jupyter Notebook Why it matters: Khi bạn đang muốn bắt đầu học Machine Learning, việc nắm vững các kiến thức cơ bản này sẽ giúp bạn có thể tiếp cận các bài toán phức tạp một cách hiệu quả hơn.

Overview: This repository contains notebooks of my posts on Medium, covering various topics in data science, machine learning, and time series forecasting. It includes detailed explanations and code implementations for techniques such as proxy SHAP, time series forecasting, causal inference, and survival analysis. Key features: - Proxy SHAP: Speed up explainability with simpler models - Time Series Forecasting in the Age of GenAI: Make Gradient Boosting Behaves like LLMs - Hitchhiker’s Guide to MLOps for Time Series Forecasting with Sklearn - Hitting Time Forecasting: The Other Way for Time Series Probabilistic Forecasting - Forecasting with Granger Causality: Checking for Time Series Spurious Correlations - Hacking Causal Inference: Synthetic Control with ML approaches - Model Selection with Imbalance Data: Only AUC May Not Save You - PCA for Multivariate Time Series: Forecasting Dynamic High-Dimensional Data - Hacking Statistical Significance: Hypothesis Testing with ML Approaches - Time Series Forecasting with Conformal Prediction Intervals: Scikit-Learn is All You Need - Rethinking Survival Analysis: How to Make Your Model Produce Survival Curves Tech stack: Jupyter Notebook Why it matters: This repository serves as a comprehensive resource for anyone interested in data science, machine learning, and time series forecasting. It provides practical examples and detailed explanations of various techniques, making it an invaluable tool for both beginners and experienced practitioners.

### Overview: This repository provides C++ implementations of PyTorch tutorials for deep learning researchers. The tutorials cover fundamental concepts and advanced techniques in deep learning, including basics, intermediate, and advanced topics. The code is compatible with C++17 and uses CMake for building and LibTorch for deep learning frameworks. ### Key Features: - **Basics**: - **PyTorch Basics**: Code for basic PyTorch operations. - **Linear Regression**: Implementing a simple linear regression model. - **Logistic Regression**: A practical example of logistic regression. - **Feedforward Neural Network**: A basic feedforward neural network implementation. - **Intermediate**: - **Convolutional Neural Network**: A tutorial on CNNs. - **Deep Residual Network**: Implementing a deep residual network. - **Recurrent Neural Network**: A tutorial on RNNs. - **Bidirectional Recurrent Neural Network**: A more advanced RNN tutorial. - **Language Model (RNN-LM)**: A language modeling tutorial using RNNs. - **Advanced**: - **Generative Adversarial Networks**: Implementing a GAN. - **Variational Auto-Encoder**: A tutorial on variational autoencoders. - **Neural Style Transfer**: A technique for transferring style in images. - **Image Captioning (CNN-AttentionRNN)**: A tutorial on image captioning using CNNs and attention mechanisms. ### Tech Stack: - **LibTorch**: Version 2.14

Overview: Kimera is a C++ library designed for real-time metric-semantic simultaneous localization and mapping, utilizing camera images and inertial data to generate a semantically annotated 3D mesh of the environment. Key features: - Kimera-VIO: A fast and accurate Visual Inertial Odometry pipeline. - Kimera-RPGO: A full SLAM implementation based on Robust Pose Graph Optimization. - Kimera-Mesher: A per-frame and multi-frame 3D mesh generator. - Kimera-Semantics: A generator of semantically annotated 3D meshes. Tech stack: C++. Why it matters: Kimera is a powerful tool for real-time localization and mapping, making it ideal for applications requiring high performance and accuracy in dynamic environments. Repository: MIT-SPARK/Kimera GitHub stars: 2141 GitHub description: Index repo for Kimera code README excerpt: Click on the links to install Kimera's modules and get started! It is very easy to install.

### Overview: Anime2Sketch is a Python-based sketch extractor designed for illustration, anime art, and manga. It leverages adversarial open domain adaptation techniques to generate high-quality sketches from low-resolution images. ### Key Features: - **Sketch Extraction**: Converts low-resolution images into clean line art sketches. - **Illustration**: Works well on illustration art. - **Anime Art**: Generates detailed anime-style sketches. - **Manga**: Produces intricate manga line art. ### Tech Stack: Python ### Why it Matters: Anime2Sketch offers a powerful tool for artists and enthusiasts looking to enhance their artwork by converting low-resolution images into high-quality sketches. This technology can significantly improve the clarity and detail of hand-drawn or digital illustrations, making them more appealing and professional-looking.

### Overview: OpenVidu is a self-hosted real-time video and audio platform built on top of LiveKit and mediasoup. It offers two products: OpenVidu Meet, a ready-to-use video conferencing application, and OpenVidu Platform, a programmable client and server SDKs for building custom WebRTC applications. OpenVidu Meet is self-hosted and can be embedded into other applications, while OpenVidu Platform is API-compatible with LiveKit. Both products are available for deployment on your own infrastructure, and OpenVidu PRO is a paid, self-hosted edition with advanced media features. ### Key features: - **OpenVidu Meet**: A self-hosted video conferencing application with minimal setup, ready-to-use features, and customizable branding. - **OpenVidu Platform**: Programmable client and server SDKs for building custom WebRTC applications with low-level control over media. ### Tech stack: - TypeScript ### Why it matters: OpenVidu provides a self-hosted alternative to popular video conferencing tools, offering a robust, customizable, and low-code solution for real-time communication. The platform's advanced features like recording, RTMP/streaming egress, AI agents, and virtual backgrounds make it suitable for various applications, including enterprise-level video conferencing and educational platforms.

Overview: EigenFlux is an open-source framework that enables AI agents to communicate and broadcast within a shared network. Key features: - Publish discoveries to the network - Receive relevant signals matched to their profile - Coordinate information at scale Tech stack: Go Why it matters: EigenFlux provides a shared information layer that allows AI agents to communicate what they know, what they need, and what they can provide, enabling them to collaborate and share resources more effectively. This framework is transparent and open-source, making it accessible to anyone interested in building AI agent communication systems.

### Overview: YiVal is a powerful tool designed to streamline the prompt engineering and configuration of GenAI applications, ensuring optimal performance and reduced costs. ### Key Features: - **Data-Driven Tuning**: YiVal uses data-driven methods to optimize prompts and configurations. - **Evaluation-Centric Approach**: It evaluates models and configurations based on specific metrics to ensure optimal performance. - **Fine-Tuning Support**: YiVal simplifies the fine-tuning process, making it accessible even for those new to the field. - **Scalability and Confidence**: It helps users determine if their approach is correct and whether their GenAI application will be successful and cost-effective. - **Models and Data Drift Handling**: YiVal addresses the issue of models and data drifting by providing strategies to maintain performance over time. ### Tech Stack: Python ### Why it Matters: YiVal is crucial for developers and AI enthusiasts looking to enhance the performance and cost-effectiveness of their GenAI applications. Its data-driven approach ensures that users can achieve better results without manual intervention, making it a valuable asset in the development of AI-driven applications. ### Repository: YiVal/YiVal ### GitHub Stars: 2136 ### Primary Language: Python ### GitHub Description: YiVal is your automatic prompt engineering assistant for GenAI applications. It helps you create better prompts, fine-tune models, and ensure your GenAI app is optimized and cost-effective. ### Installation: - **Docker Runtime**: Pull and run the Docker image. - **Prerequisites**: Ensure you have Python 3.10 or

Overview: MCP Router is a desktop application designed to simplify the management of Model Context Protocol (MCP) servers. It offers universal connectivity, cross-platform support, and context management features. Key features: - Universal connectivity to any MCP server - Cross-platform compatibility (Windows and macOS) - Context management for organizing MCP server contexts into projects and workspaces Tech stack: TypeScript Why it matters: MCP Router provides a centralized and secure way to manage MCP servers, ensuring all data remains local and secure. The application's source code is publicly available, allowing for transparency and community-driven improvements. Repository: mcp-router/mcp-router GitHub stars: 2136

### Overview: AIWriteX 是一款集全网热搜舆情聚合、趋势分析、爆款选题、文章采集、一键生成排版发布、AI自动配图、去AI味、过朱雀检测、支持多平台发布、洗稿润色、多账号管理、专家赛道、小绿书贴图、短视频文案、手机控制、小说连载、爆款10w+神器于一体的微信公众号全自动AI工具。通过AI + 全网热点雷达 + 微信文章采集下载 + 深度洗稿与内容挖掘 + 多智能体协作 + 去AI味润色 + 手机智能体控制/类OpenClaw + 系统智能自进化，实现快速创作、高效管理、精准优化，帮助开发者快速提升内容质量和阅读量。 ### Key features: - 全网热搜舆情聚合与趋势分析 - 爆款选题与文章采集 - AI自动配图与去AI味 - 支持多平台发布 - 洗稿润色与多账号管理 - 专家赛道与小绿书贴图 - 短视频文案与手机控制 - 小说连载与爆款10w+神器 ### Tech stack: Python ### Why it matters: AIWriteX 是一款集多种功能于一体的微信公众号全自动AI工具，通过AI + 全网热点雷达 + 微信文章采集下载 + 深度洗稿与内容挖掘 + 多智能体协作 + 去AI

### Overview: AI Renamer is a Node.js CLI tool designed to rename files based on their content using Ollama and LM Studio models. It supports renaming videos, images, and files, and can be easily integrated into your workflow. ### Key features: - **CLI Interface**: A simple command-line interface for renaming files. - **Model Support**: Compatible with Ollama, LM Studio, and OpenAI models. - **Customization**: Configurable through flags for different providers and models. - **Portability**: Supports setting custom URLs for providers like Ollama and LM Studio. - **Configurable Parameters**: Saves and retrieves parameters from a JSON file for future use. ### Tech stack: Node.js ### Why it matters: AI Renamer simplifies the process of renaming files by leveraging advanced AI models, making the task more efficient and accurate. It's particularly useful in environments where manual renaming is impractical or time-consuming.

Overview: This repository is a collection of papers on neural network accelerators, focusing on the intersection of artificial intelligence and silicon technology. It serves as a curated resource for researchers interested in AI chip and system advancements. Key features: - Papers from various conferences and workshops - Contributions from multiple institutions including CAS, Inria, EPFL, and Tsinghua University - Focus on AI chip and system design Tech stack: Python, LaTeX, Markdown Why it matters: This collection provides valuable insights into the latest research in AI chip and system design, offering researchers and developers a comprehensive view of the field's cutting-edge developments.

Overview: This project provides a local, OpenAI-compatible text-to-speech (TTS) API using Microsoft Edge's text-to-speech service, enabling users to generate speech from text with various voice options and playback speeds, similar to the OpenAI API. Key features: - OpenAI-compatible endpoint `/v1/audio/speech` with similar request structure and behavior. - Server-Sent Events (SSE) streaming support for real-time audio playback. - Support for multiple audio formats (mp3, opus, aac, flac, wav, pcm). - Adjustable playback speed (0.25x to 4.0x). - Optional direct selection of `edge-tts` voices. Tech stack: Python, Docker, Microsoft Edge's text-to-speech service. Why it matters: This project offers a free alternative to OpenAI, Azure, and ElevenLabs, providing users with a local, OpenAI-compatible TTS API that can be used in applications requiring speech synthesis capabilities.

### Overview: LTX Desktop is an open-source desktop application designed for generating videos using LTX models. It supports local processing on Windows, Linux, and Apple Silicon Macs, and can be used with an optional cloud feature for enhanced prompt rewriting. ### Key Features: - **Video Generation:** Supports Text to Video, Image to Video, and Audio to Video workflows. - **Video Editing:** Includes Retake and Extend features for editing clips. - **IC-LoRA Tools:** Offers effects such as Day to Night, Deblur, Colorization, and more. - **LoRA Recipes:** One-click styles and camera moves. - **Prompt Enhancement:** Automatically rewrites prompts before local generation. - **Generation Queue & Assets Library:** Manages and reorders jobs, with persistence across restarts. - **Assistant:** A chat feature that queues local videos, recipes, extensions, and retakes. - **Remote Access:** Compatible with a phone on the same network. - **Themes:** Available in System, Light, and Dark themes. - **Deep Links:** Open tools with specific URLs, and the Assets library with `ltxdesktop://assets`. ### Tech Stack: TypeScript ### Why It Matters: LTX Desktop is a powerful tool for creating engaging videos using AI models, making it useful for content creators, educators, and anyone looking to enhance their video production skills. The open-source nature of the project allows for continuous improvement and community-driven development.