Projects

### Key features: - **Natural Language Processing (NLP):** - Sentiment analysis - Natural language understanding - **Robustness, Ethics, Biases, and Trustworthiness:** - Evaluations across various NLP tasks - Assessing model robustness, ethics, biases, and trustworthiness - **Social Science:** - Evaluations in fields like language modeling, ethics, and biases - **Natural Science and Engineering:** - Assessing model performance in scientific and engineering applications - **Medical Applications:** - Evaluations in healthcare and medical applications - **Agent Applications:** - Assessing model performance in agent-based systems - **Other Applications:** - Evaluations in various other domains ### Tech stack: Python, Markdown, Git ### Why it matters: This repository serves as an official resource for the survey paper "A Survey on Evaluation of Large Language Models," providing up-to-date information and contributions from leading researchers in the field.

### Overview: Cache-Augmented Generation (CAG) is a novel approach to enhance language models by bypassing real-time retrieval, significantly reducing latency and improving reliability. ### Key Features: - **Reduced Latency**: CAG eliminates the need for real-time retrieval, enabling faster inference. - **Improved Reliability**: Minimizes retrieval errors while maintaining context relevance. - **Simplified Design**: Provides a streamlined, retrieval-free alternative to RAG, achieving comparable or superior results with lower complexity. ### Tech Stack: Python ### Why It Matters: CAG offers a practical and scalable alternative to traditional RAG, addressing the limitations of RAG while maintaining comparable performance with lower complexity.

### Overview: MemoryOS aims to enhance the capabilities of AI agents by providing a memory operating system that improves their performance in long-term memory management, achieving significant improvements in F1 and BLEU-1 scores on the LoCoMo benchmark. ### Key Features: - **TOP Performance in Memory Management**: The model outperforms state-of-the-art results in long-term memory benchmarks, boosting F1 scores by 49.11% and BLEU-1 by 46.18%. - **Plug-and-Play Memory Management Architecture**: Enables seamless integration of pluggable memory modules, including storage engines, update strategies, and retrieval algorithms. - **Agent Workflow Create with Ease**: Injects long-term memory capabilities into various AI applications by calling modular tools provided by the MCP Server. - **Universal LLM Support**: Seamlessly integrates with a wide range of language models (e.g., OpenAI, Deepseek, Qwen ...) ### Why it Matters: MemoryOS offers a comprehensive and efficient memory management solution for AI agents, enhancing their performance in long-term memory tasks, thereby improving their coherence, personalization, and context-aware interactions.

Overview: Rivet Agents is a developer-focused project that provides a robust infrastructure for running agents using the Rivet framework. It ensures agents are durable and can survive restarts, making them reliable for various applications. Key features: - Durable process per agent - Memory that survives restarts - Support for running agents as Rivet Actors - Ability to run coding agents in sandboxes - Comprehensive documentation and project management tools Tech stack: TypeScript Why it matters: Rivet Agents offers a reliable and scalable solution for developers working with Rivet, ensuring their agents are resilient and easily maintainable. Repository: rivet-dev/agents Primary language: TypeScript GitHub stars: 1609

Overview: DELTA is a deep learning-based platform designed for natural language and speech processing, offering easy, fast, and reliable solutions for academia and industry use cases. Key features: - Easy-to-use: One command for training NLP and speech models, including text classification, named entity recognition, question and answering, text summarization, etc. - Easy-to-deploy: Integrates data processing and features extraction into a model graph, uniform I/O interfaces, and no changes for new models. - Easy-to-develop: Modularized components, reliable and fully-tested modules. Tech stack: TensorFlow and Python 3. Why it matters: DELTA simplifies the development and deployment of natural language processing and speech models, providing a user-friendly and efficient solution for both academic and industrial applications.

### Overview: 本项目提供了一套用于生成炫酷 HTML 动画的 AI Prompts，涵盖动画效果、3D 可视化、PPT 风格演示、UI 美化等多个类别。 ### Key features: - **PPT 风格演示**: 科普、技术讲解、视频配套演示。 - **流程图 / 原理演示**: 流程图、概念图、对比图、时序图、AI 模型可视化等。 - **网络协议可视化**: TCP/IP、IPv4、以太帧、路由、DHCP 等。 - **动态架构图**: 系统架构、流程图、时序图、数据流图、状态机，带流动动画 + 多格式导出。 - **学霸笔记**: 手写笔记本风格的精美 HTML 学习笔记，两种模板风格。 - **卡片剧场**: 侧边栏叙事 + 3D 卡片轮播，协议流程 / 产品特性 / 分步讲解的叙事感演示。 - **视频分镜演示**: 电影级"一个镜头一个 HTML"演示动画：29 种风格轮换、镜头推拉、WebAudio 音效、角色表情吐槽，全屏录屏即成片。 - **手机系统 UI 演示**: 一台"真手机"的电影化编排录屏：锁屏通知、聊天、设置页、App 界面逐镜头呈现，HyperOS 级手感动

### Overview: Face Mask Detection is a project that leverages computer vision and deep learning techniques using OpenCV and TensorFlow/Keras to detect face masks in both static images and real-time video streams. The system is designed to be accurate and computationally efficient, making it suitable for deployment in various applications, including transportation and manufacturing. ### Key Features: - **Real-time Detection**: The system can accurately detect face masks in both static images and real-time video streams. - **MobileNetV2 Architecture**: Utilizes MobileNetV2 for efficient computation, making the model suitable for deployment on embedded systems. - **OpenCV Integration**: Integrates OpenCV for image processing and analysis. - **Keras/TensorFlow Framework**: Employs Keras and TensorFlow for deep learning models. ### Tech Stack: - OpenCV - Keras/TensorFlow ### Why It Matters: The absence of large datasets of "with_mask" images has made this task challenging. The project aims to address this by leveraging accurate and efficient models, making it crucial for ensuring safety in various applications, including transportation and manufacturing. ### Repository: The repository is hosted on GitHub and is currently at 1605 stars, indicating a significant interest and engagement from the developer community. ### Primary Language: The project is primarily written in Jupyter Notebook, leveraging Python for its computational capabilities. ### GitHub Description: Face Mask Detection is a system built with OpenCV, Keras/TensorFlow using Deep Learning and Computer Vision concepts. It is designed to detect face masks in static images and real-time video streams

Overview: Apache OpenNLP is a machine learning-based toolkit for natural language text processing, written entirely in Java. Key features: - Tokenization - Sentence segmentation - Part-of-speech tagging - Named entity extraction - Chunking - Parsing - Coreference resolution - Language detection - Stopword filtering Tech stack: Java Why it matters: Apache OpenNLP provides a comprehensive set of tools for natural language processing tasks, making it a valuable resource for developers working on text processing applications.

Overview: 将 WorkBuddy 桌面应用中的模型自动接入 DeepSeek Harness，实现零配置的无缝集成。 Key features: - 自动将 WorkBuddy 模型接入 DeepSeek Harness - 分别显示 WorkBuddy 和 WorkBuddy AI 模型组 - 支持图片输入和推理档位检测 - 显示剩余积分和模型优惠 - 企业账号和费率比例显示 - 徽章展示促销信息 Tech stack: TypeScript Why it matters: 提升工作效率，简化模型接入流程，实现零配置的深度学习模型集成。

### Overview: Fast AutoAugment is an efficient implementation of the Fast AutoAugment algorithm, which uses a more efficient search strategy to speed up the augmentation process while maintaining comparable performance. ### Key Features: - **Efficient Search Strategy**: Fast AutoAugment employs a more efficient search strategy based on density matching to significantly reduce the search time. - **Order of Magnitude Speedup**: It achieves speeds up to 1428 times faster than AutoAugment while maintaining comparable performance. - **Versatile Models**: The algorithm is compatible with various models and datasets, including Wide-ResNet, Shake-Shake, PyramidNet, and ResNet. - **Cross-Dataset Performance**: It performs well across different datasets, including CIFAR-10, CIFAR-100, and ImageNet. ### Tech Stack: Python ### Why It Matters: Fast AutoAugment offers a significant improvement in performance and speed, making it an essential tool for researchers and practitioners working with image augmentation techniques in deep learning.

Overview: This project is a plug-and-play library designed to enable agents to call MCP and UTCP tools via code execution, offering a more efficient approach to tool calling compared to traditional methods. Key features: - Code-Mode Library: A library that simplifies the process of calling MCP and UTCP tools through code execution. - TypeScript Support: The library is built with TypeScript, ensuring compatibility and ease of integration. - Performance Improvement: It claims to provide a 67% to 88% performance improvement over traditional methods, as validated by a benchmark study. Tech stack: TypeScript Why it matters: The library is a significant advancement in the field of AI and automation, offering a more efficient and streamlined way to interact with various tools, potentially reducing operational costs and improving overall efficiency. Repository: universal-tool-calling-protocol/code-mode GitHub stars: 1603 GitHub description: 🔌 Plug-and-play library to enable agents to call MCP and UTCP tools via code execution.

### Overview xLLM is a high-performance inference engine designed specifically for LLM, VLM, DiT, and REC models, optimized for diverse AI accelerators. It is hosted in the OpenAtom Foundation and is licensed under the Apache 2.0 license. The repository has received 1,601 GitHub stars, making it a valuable resource for developers looking to enhance their AI applications. ### Key Features - **Efficient Inference**: Utilizes advanced features to deliver high-throughput and low-latency inference. - **Versatile Hardware Support**: Optimized for various AI accelerators, including GPUs, TPUs, and Ascend processors. - **Flexible Deployment**: Supports deployment on multiple platforms, including Linux, Windows, and macOS. ### Tech Stack The core technology stack of xLLM includes C++ for the backend implementation and Python for the frontend interface. ### Why It Matters xLLM is crucial for developers and organizations looking to deploy large language models (LLMs) and related AI models efficiently. By leveraging its optimized performance and hardware support, it enables enterprises to achieve better resource utilization and reduced operational costs, making it an essential tool for AI development and deployment. ### Repository The repository is maintained by xLLM-AI and can be found at [xLLM-AI/xllm](https://github.com/xLLM-AI/xllm). ### Primary Language The primary language used in the repository is C++. ### GitHub Stars The repository has received 1,601 GitHub stars, indicating its popularity and utility

### Overview: ExtractThinker is a powerful Python library designed for developers working with Document Intelligence. It simplifies the process of turning documents into structured data, leveraging LLMs for efficient and flexible document workflows. ### Key Features: - **ORM-style Interaction**: ExtractThinker provides a way to interact with documents similar to working with ORM models. - **Document Loaders**: Supports various document types like Text, PDF, Images, Tables, and Spreadsheets. - **Pydantic Contracts**: Defines schemas for document fields to ensure data integrity. - **Classification and Splitting**: Classifies documents and splits them into manageable parts. - **Model Routing**: Configurable routing for different LLM models. - **Local Entity Masking**: Masks entities locally to prevent sensitive information leakage. - **Page Events**: Tracks page events for better document understanding. - **PyMuPDF/Camelot/Tabula/Adobe Loaders**: Supports multiple document loaders for a variety of document types. ### Tech Stack: - Python 3.9–3.13 ### Why It Matters: ExtractThinker simplifies the process of turning raw documents into structured data, making it easier for developers to work with complex documents efficiently. The library's flexibility and powerful features make it a valuable tool for any developer dealing with document intelligence. ### Repository: - **Repository**: `https://github.com/enoch3712/ExtractThinker` - **Stars**: 1602 - **Description**: ExtractThinker is an open-source Python library for turning documents into typed data. It offers

### Overview DEEIX Chat is an open-source, deployable AI platform designed for enterprise-level access to multiple model providers. It offers a unified interface for model routing, multimodal chat, files, tools, billing, identity management, and operational controls. The system is lightweight, efficient, and scalable, ensuring stable and unified access to various model providers. ### Key Features - **Conversations**: A multimodal chat interface for daily use, including streaming, branches, retries, and rich rendering. - **Models and Routing**: A platform-model layer for upstream channels, real models, route bindings, and capability configuration. - **Protocols and Adaptation**: Support for OpenAI, Anthropic, Google/Gemini, xAI, OpenRouter, and OpenAI-compatible protocols. - **Files and Retrieval**: File upload, preview, extraction, OCR, storage quota, full-context injection, chunking, embeddings, and semantic retrieval. - **Tool Ecosystem**: MCP servers and provider-native tools with discovery, enablement, user selection, execution limits, result rendering, and tool-call traceability. - **Context and Memory**: Message windows, token budgets, summary compression, conversation memory, long-term memory, and RAG evidence records. - **Billing and Payments**: Model pricing, per-call tool pricing, subscriptions, top-ups, balances, usage ledgers, Stripe Checkout, EPay, and webhook validation. - **Identity and Security**: Local accounts, session management, HttpOnly refresh cookies, 2FA/TOTP, trusted devices, SSO/OIDC/OAuth,

### Overview: This repository is a curated list of 100+ resources designed to help aspiring Generative AI Data Scientists build and deploy applications using Large Language Models (LLMs) and cloud-based solutions. ### Key features: - 🚀🚀 100+ Free Resources On Generative AI for Data Scientists - Contributing: Welcome contributions to improve the list - Table of Contents: Contains links to various resources ### Tech stack: Python ### Why it matters: This repository serves as a comprehensive guide for individuals looking to enhance their skills in creating and deploying generative AI applications, focusing on the intersection of AI and data science.

Overview: GHOST is a new one-shot face swap approach that addresses the limitations of existing methods by providing high-fidelity results for image and video domains. Key features: - High-fidelity image and video face swaps - Easy-to-use API and software - Ethical statement against misuse - Colab demo and local Jupyter notebook Tech stack: Python Why it matters: GHOST offers a more robust and user-friendly solution for face swapping, addressing the concerns of researchers and the public about the ethical implications of deepfake technology.

Overview: DataProfiler is a Python library designed to extract schema, statistics, and sensitive data from datasets, making data analysis and monitoring easier. Key features: - Automatically formats and loads files into a DataFrame - Identifies schema, statistics, entities, and sensitive data - Supports CSV, AVRO, Parquet, JSON, and URL data formats - Provides a compact report format for quick access Tech stack: Python Why it matters: Simplifies data analysis and monitoring by automating the process of extracting schema and statistics from datasets, making it easier for developers to work with and understand their data. Repository: capitalone/DataProfiler Primary language: Python GitHub stars: 1595 GitHub description: What's in your data? Extract schema, statistics and entities from datasets

### Overview: MiniMax-AI's MiniMax-MCP server is an official implementation of the MiniMax Model Context Protocol (MCP), designed to enable seamless interaction with powerful Text to Speech, image generation, and video generation APIs. ### Key Features: - **Text to Speech**: Converts text into spoken words. - **Image Generation**: Creates images based on textual descriptions. - **Video Generation**: Produces videos from textual inputs. ### Tech Stack: Python ### Why It Matters: This server is crucial for developers and content creators who need to integrate advanced AI functionalities into their projects, providing a robust and efficient solution for generating human-like text, images, and videos.

### Overview: Steel is an open-source browser API designed specifically for AI agents, making it easy to build AI applications that can effectively interact with the web. ### Key features: - **Steel Browser**: A browser API built for AI agents. - **Autonomous Web Agents**: AI agents that autonomously navigate and interact with the web through a user-friendly interface. - **AI Web Automation Tools**: Tools and frameworks for automating web interactions. - **AI Web Scrapers/Crawlers**: Tools for web scraping and crawling. - **Web Search & Query Tools**: Tools for searching and querying web content. - **Benchmarks & Research**: Resources for benchmarking and research in AI web agents. - **Tutorials & Guides**: Resources for learning how to implement AI agents. - **Archive**: A collection of resources and tools. - **Interested in implementing Steel?**: Resources for getting started. - **Join the Community**: Join the community for support and collaboration. - **Contributing**: Resources for contributing to the project. ### Tech stack: Python ### Why it matters: Steel is an open-source browser API designed specifically for AI agents, making it easy to build AI applications that can effectively interact with the web.

### Overview: Infinity is a cutting-edge visual synthesis framework that leverages bitwise auto-regressive modeling to generate high-resolution and photorealistic images. This project is a significant advancement in the field of visual autoregressive models, achieving state-of-the-art performance on various benchmarks. ### Key Features: - **Bitwise AutoRegressive Modeling**: Uses a bitwise token prediction framework with an infinite-vocabulary tokenizer and classifier for powerful scaling capabilities. - **Infinite Vocabulary Tokenizer**: Enables the model to scale vocabulary size to infinity while maintaining high performance. - **Infinite Transformer**: Utilizes an infinite transformer to further enhance the model's capabilities. - **Generative Performance**: Outperforms top-tier diffusion models like SD3-Medium and SDXL in terms of accuracy and efficiency. - **Benchmark Improvements**: Achieves a win rate of 66% in the GenEval benchmark, surpassing SD3-Medium by 0.11 points and SDXL by 0.11 points. ### Tech Stack: - **Python**: The primary programming language used for implementing the project. ### Why It Matters: Infinity represents a significant leap forward in visual autoregressive modeling, offering unparalleled performance and scalability. Its ability to generate high-resolution images with photorealistic details makes it a valuable tool for various applications, including art, design, and scientific visualization. ### Repository: - **GitHub Repository**: [FoundationVision/Infinity](https://github.com/FoundationVision/Infinity) - **GitHub Stars**: 1593 - **GitHub Description**: [CVPR 202