Projects

### Overview: This project is a fork of the original Stable Diffusion repository, focusing on personalizing the model using aesthetic gradients. It allows users to apply the aesthetic gradients technique described in the paper to improve the quality of generated images. ### Key Features: - Personalization of Stable Diffusion models using aesthetic gradients. - Includes aesthetic embeddings for various styles, such as fantasy, flowers, and plants. - Additional aesthetic embeddings created from specific datasets. - Additional arguments for controlling the aesthetic gradient optimization process. ### Tech Stack: Jupyter Notebook ### Why it Matters: With this tool, users don't need to learn complex modifications to improve the quality of the generated images, making it accessible for both beginners and advanced users.

Overview: This repository is part of a survey paper that explores the applications of large language models (LLMs) in audio AI, focusing on the advancements and challenges in using these models for audio signal processing. The paper aims to provide an in-depth analysis of state-of-the-art methodologies, performance benchmarks, and real-world applications of foundational large audio models. Key features: - Popular Large Audio Models - Automatic Speech Recognition (ASR) - Neural Speech Synthesis - Speech Translation (ST) - Other Speech Applications - Audio Reasoning and Reinforcement Learning - Large Audio Models in Music - Audio Datasets Tech stack: Unknown Why it matters: This repository supplements a survey paper that provides a comprehensive overview of the advancements and challenges in applying large language models to audio signal processing. It aims to spark further discussion and innovation in the field of audio-processing systems.

### Overview: The Ghost-Agent is a powerful command-line interface (CLI) tool designed to execute structured tool calls in a sequence. It integrates seamlessly with various APIs and tools, making it ideal for backend execution engines and plugin-based AI systems. ### Key Features: - **First MCP Agent Execution Engine on Solana**: The Ghost-Agent is the first MCP Agent Execution Engine on Solana, making it a pioneering project in the field. - **Detects HTTP Signals from Paid APIs**: It can identify and respond to HTTP signals from paid APIs, ensuring compliance and security. - **Suspends Execution & Informs User of Required Payment**: The Ghost-Agent automatically suspends execution and informs the user of any required payments, streamlining the process. ### Tech Stack: - Shell ### Why It Matters: The Ghost-Agent is a game-changer for agent economy development, providing a unified runtime for structured tool calls and enabling efficient execution of complex tasks. Its ability to handle multiple chains and tools makes it a versatile solution for backend execution engines and AI systems. ### Repository: - **GitHub Repository**: CryptoDmitry/Ghost-Agent - **GitHub Stars**: 743 - **GitHub Description**: The Ghost-Agent is the Archon. It was called forth from the void by a single string of natural language—a careless incantation whispered by my Master into the command line. ### Primary Language: - **Primary Language**: Shell ### Latest Update: - **2026-01 - mcp Support Branch Released**: A new branch `coreon-mcp` is available for tracking

### Overview: Stable Diffusion is a powerful tool for generating high-quality images from text prompts. This project provides a Docker container that runs the official Stable Diffusion releases, enabling users to create stunning paintings, detailed scenes, and more. ### Key features: - **Text-to-Image (`txt2img`)**: Generate an image from a text prompt. - **Image-to-Image (`img2img`)**: Modify an existing image with a text prompt. - **Depth-Guided Diffusion (`depth2img`)**: Enhance an image with a depth map and a text prompt. - **Instruct Pix2Pix (`pix2pix`)**: Modify an existing image with a text prompt. - **Stable UnCLIP Variations (`unclip`)**: Create different versions of an image with a text prompt. - **Image Upscaling (`upscale4x`)**: Generate a high-resolution image from an existing image with a text prompt. ### Tech stack: Python ### Why it matters: This Docker container simplifies the process of generating high-quality images, making it accessible for artists, designers, and anyone looking to create unique and visually stunning content.

Overview: kornia-rs is a low-level computer vision library written in Rust, providing fast, thread-safe image I/O and processing with a single API that runs on the CPU or an NVIDIA GPU. Key features: - Written in Rust: memory- and thread-safe, no GIL — usable from the free-threaded Python build. - Fast image I/O and processing: libjpeg-turbo decoding and SIMD (NEON/AVX2) kernels. - One API, CPU or GPU: the same `Image` and operators dispatch on residency — no separate GPU types. - Zero-copy ML interop: DLPack and `__cuda_array_interface__` to and from PyTorch, plus numpy views. - Real-time ready: V4L2 camera capture and a fused NV12/YUYV → normalized CHW CUDA kernel for inference. - Python bindings via PyO3/Maturin, packaged for Linux, macOS, and Windows; the same wheel is CPU-only or activates CUDA when an NVIDIA GPU is present. - Supported Python versions: 3.8 through 3.14, including the free-threaded (3.13t/3.14t) build. Tech stack: Rust. Why it matters: kornia-rs provides a fast, efficient, and user-friendly computer vision library for Rust developers, enabling real-time image processing and real-time video capture without the overhead of Python.

### Overview: Mimi 是一款实时翻译字幕的软件，能够将电脑或麦克风的声音实时翻译成字幕，并显示在屏幕上。 ### Key features: - **实时字幕显示**：支持双语、仅译文或仅原文的字幕显示。 - **音源采集**：支持采集电脑声音、麦克风声音，以及同时采集两路声音。 - **调整显示**：支持桌面端调整字幕位置、字号、颜色，显示历史字幕，以及沉浸模式显示文字。 - **保存功能**：支持将字幕导出为 TXT 或 Markdown，将所选音源录制为 WAV。 ### Tech stack: Rust ### Why it matters: Mimi 是一款功能强大的实时字幕软件，适用于各种场景，如观看电影、直播或在线课程，帮助用户快速理解内容。

### Overview: Legal Text Analytics is a comprehensive list of resources, methods, and tools focused on legal text analytics. This project aims to provide developers and legal professionals with the latest tools and techniques for analyzing and processing legal documents efficiently. ### Key features: - Optical Character Recognition (OCR) - Legal Document Pre-processing - Clause and Sentence Boundary Detection - Information Extraction and Named Entity Recognition - Legal Norm Classification - Machine Translation - Document Comparison and Semantic Matching - Text Summarization - Argument Mining - Question Answering - Legal Case Outcome Prediction - Legal and Regulatory Monitoring - Legal Criticality Prediction - Court View Generation - Reference and Coreference Extraction - Document Assembling and Generation ### Tech stack: Python, Natural Language Processing (NLP), and various libraries. ### Why it matters: Legal Text Analytics is crucial for anyone involved in legal document analysis, including legal professionals, researchers, and developers. It helps in automating the process of extracting meaningful information from legal documents, improving accuracy and efficiency in legal research and analysis. Repository: Liquid-Legal-Institute/Legal-Text-Analytics GitHub stars: 742 GitHub description: A list of selected resources, methods, and tools dedicated to Legal Text Analytics.

### Overview: Photoshop MCP is an unofficial, community-maintained project that allows users to control Adobe Photoshop through AI-powered commands. It integrates with the built-in chat window and supports various AI models for different tasks, including removing backgrounds, retouching portraits, and exporting images for web and social media. ### Key Features: - **AI-Driven Commands**: Users can describe their needs in plain language, and Photoshop MCP translates these into appropriate commands. - **Instant Edits with Jev**: Jev routes each prompt before any language model runs, ensuring safe commands go directly to Photoshop without LLM calls, while more complex tasks go to the AI. - **118 Tools**: The project includes 118 tools, covering a wide range of image editing tasks. - **Built-in Chat Window**: Users can interact with Photoshop MCP using the built-in chat window, which can be configured to send prompts to TypeSafe AI or the Action Plan. ### Tech Stack: TypeScript ### Why It Matters: Photoshop MCP provides a seamless and efficient way to perform complex image editing tasks without the need for coding or scripting. It leverages AI models to handle tasks like removing backgrounds and retouching portraits, making it accessible to users who may not be familiar with Photoshop scripting. The project is community-driven and aims to enhance the user experience by providing a more intuitive interface for AI-driven image editing.

### Overview: ASTER is an accurate scene text recognizer designed for flexible rectification. The project focuses on recognizing text in natural images, utilizing a combination of attention mechanisms and flexible rectification to enhance recognition accuracy. ### Key Features: - **Flexible Rectification Mechanism**: Allows for precise alignment of text within images. - **Attentional Scene Text Recognition**: Enhances text detection and recognition by focusing on relevant regions. - **Accurate Text Detection and Recognition**: Achieves high accuracy in recognizing text in various image scenarios. ### Tech Stack: - **Python**: The primary programming language used for the implementation. - **TensorFlow**: The framework used for building the custom operators and training models. ### Why It Matters: ASTER is particularly useful in applications where high accuracy in recognizing text in natural images is crucial, such as in security systems, document recognition, and other fields where text detection and recognition are essential. ### Repository: bgshih/aster ### Primary Language: Python ### GitHub Stars: 742 ### GitHub Description: Recognizing cropped text in natural images.

### Overview: This project provides a ready-to-use Chinese text normalization module for speech processing applications. The module is designed to handle common text normalization tasks for Chinese, making it suitable for use in Chinese ASR text processing pipelines. ### Key features: - **NSW Normalization**: Supports normalization of non-standard words such as "cardinal" (重达324.75克), "date" (出生于86年8月18日), "digit" (编号27149), "fraction" (7/12), "money" (12块5), "percentage" (62%), and "telephone" (固话0421-33441122, 手机+86 18544139121). - **Punctuation Removal**: Removes punctuation from Chinese text, including non-stop and stop punctuation. - **Multilingual English Word Case Conversion**: Converts English words to either uppercase or lowercase based on the ASR/TTS lexicons. ### Tech stack: Python ### Why it matters: - **Language Dependence**: Text normalization is language-dependent, and this module is designed to handle Chinese specifically. - **Task-Specificity**: The module is task-specific, requiring different treatments for various applications. - **Complexity**: Constructing and maintaining a set of TN rules is challenging, and this project aims to provide a solution that is ready to use. - **Asset for Commercial Companies**: Since the process is hard to automate, it is an asset for companies

Overview: Marinara Engine is an AI-powered chat, roleplay, and game engine built for easy setup and fun gameplay. It's designed for quick and seamless installation, making it perfect for both casual users and experienced developers. Key features: - Chat and Roleplay - Visual and Immersive Experience - Appearance and Themes - AI Agent System - Prompt Engineering - Local Customization - Connections and Providers - Export and Data Tech stack: TypeScript Why it matters: Marinara Engine simplifies the process of creating and running a chat, roleplay, and game engine, making it accessible and enjoyable for users of all skill levels.

Overview: This project provides a basic framework for quickly setting up PyTorch projects. Key features: - Basic PyTorch project structure - Easy-to-use templates for common neural network experiments - Documentation and best practices for reproducibility Tech stack: Python Why it matters: This project helps developers organize their neural network experiments systematically, making it easier to replicate results and learn from others' work.

Overview: This project is a starter repository for the Fill 3D API, showcasing how to integrate AI-powered virtual staging into 3D models. Key features: - Integrates with the Fill 3D API for AI-driven virtual staging. - Provides a simple setup process for developers to start using the API. - Includes a Discord community link for support and collaboration. Tech stack: TypeScript Why it matters: This project serves as a foundational example for developers looking to leverage AI technology for virtual staging, making it easier to integrate the Fill 3D API into their projects.

### Overview: PrimeQA is a comprehensive repository for state-of-the-art multilingual question answering research and development. It provides researchers and developers with the tools to replicate experiments from leading NLP conferences and access pre-trained models for various QA tasks. ### Key Features: - **Information Retrieval (IR):** Retrieves documents and passages using traditional (e.g., BM25) and neural (e.g., ColBERT) models. - **Multilingual Machine Reading Comprehension (MRMC):** Extracts and generates answers from source documents or passages. - **Multilingual Question Generation (MQG):** Generates questions for effective domain adaptation over tables and multilingual text. - **Retrieval Augmented Generation (RAG):** Uses GPT-3/ChatGPT pretrained models to generate answers conditioned on retrieved passages. ### Tech Stack: - **Transformers Toolkit:** Built on top of the Hugging Face Transformers toolkit. - **Datasets:** Directly downloadable datasets and models. ### Why it Matters: PrimeQA enables researchers to replicate top-performing models and experiments from leading NLP conferences, providing a reliable and accessible platform for advancing multilingual QA research.

### Overview: This repository, **GaryYufei/AlignLLMHumanSurvey**, serves as a comprehensive survey of techniques and methods used to align large language models (LLMs) with human expectations. It provides a detailed overview of data collection, training methodologies, and evaluation paradigms, aiming to enhance the alignment of LLMs for better human-oriented tasks and expectations. ### Key Features: - **Data Collection**: Various sources of human data and strong LLMs for training. - **Training Methodologies**: Online and offline human alignment, parameter-efficient training. - **Evaluation**: Evaluation design principles, benchmarks, and paradigms. ### Tech Stack: - **GitHub**: 738 stars ### Why It Matters: This survey is crucial for researchers and practitioners interested in understanding and advancing the field of aligning large language models with human expectations. It provides valuable insights into promising future research avenues and serves as a valuable resource for anyone involved in this emerging area. ### Repository: GaryYufei/AlignLLMHumanSurvey - **Primary Language**: Unknown - **GitHub Stars**: 738 - **Description**: Aligning Large Language Models with Human: A Survey - **Citation**: ```bash @article{aligning_llm_human, title={Aligning Large Language Models with Human: A Survey}, author={Yufei Wang and Wanjun Zhong and Liangyou Li and Fei Mi and Xingshan Zeng and Wenyong Huang and Lifeng Shang and Xin Jiang and Qun Liu},

Overview: Symbiotic AI is a personal AI agent that integrates seamlessly into your existing files, allowing you to own and control your context. This unique approach challenges you with real conversations and evolves based on your experiences. Key features: - Personalized agent personality and identity - Your profile: identity, psychology, and mission - How the agent operates: protocols, tools, and patterns - Current state: tasks, queue, log, and active projects Tech stack: Shell Why it matters: Symbiotic AI empowers you to control your own context, evolve your patterns, and challenge yourself in a way that aligns with your values and commitments. Repository: lout33/symbiotic-ai GitHub stars: 739 GitHub description: A symbiotic AI agent that remembers everything, challenges you, and extends your cognition.

### Overview: Pi Extensions for the Pi Coding Agent A monorepo of Pi Coding Agent extensions and reusable libraries for coding, research, browser automation, workflow management, observability, and terminal ergonomics. ### Key Features: - Install and use extensions permanently or temporarily. - Install multiple extensions at once. - Quick start guide for permanent installation. - Extensions I use every day. - Choose an extension based on your needs. ### Tech Stack: TypeScript ### Why It Matters: These extensions are part of a developer's daily setup, providing a seamless and efficient coding experience on the Raspberry Pi.

Overview: Generative AI Examples is a collection of microservice-based examples designed to simplify the deployment, testing, and scaling of GenAI applications. These examples are compatible with Docker and Kubernetes, supporting various hardware configurations. Key features: - ChatQnA (Chatbot with Retrieval Augmented Generation) - VisualQnA (Visual Question-answering) - Text2Image (Text-to-image generation) - DocSum (Document Summarization) - CodeGen (Gen-AI Powered Code Generator) - DocIndexRetriever (Document Retrieval with Retrieval Augmented Generation) - InstructionTuning (Application of Instruction Tuning) Tech stack: Shell Why it matters: This project provides a comprehensive set of examples that help developers quickly understand and implement generative AI solutions, making it easier to adopt and scale GenAI applications in various environments.

### Overview: This curated list of open source tooling is designed to help developers and data scientists build data-centric AI systems on unstructured data. It includes tools for various aspects of DCAI, such as data versioning, embeddings and pre-trained models, visualization and interaction, outlier and noise detection, explainability, active learning, uncertainty quantification, bias and fairness, observability and monitoring, augmentation and synthetic data, and security and robustness. The tools are open-source and actively maintained, making them suitable for reproducible machine learning projects. The collection aims to provide a useful focus and prevent duplicate work, excluding certain topics like tabular data tools, dedicated labeling tools, and research papers. ### Key features: - Data versioning - Embeddings and pre-trained models - Visualization and interaction - Outlier and noise detection - Explainability - Active learning - Uncertainty quantification - Bias and fairness - Observability and monitoring - Augmentation and synthetic data - Security and robustness ### Tech stack: Open source and actively maintained tools for data-centric AI on unstructured data. ### Why it matters: This curated list is essential for developers and data scientists looking to build real-world AI systems that generate value using the DCAI paradigm. It provides a comprehensive set of tools and workflows to help optimize training datasets, improve explainability, and ensure security and robustness in AI applications.

### Overview: Pubmed Parser is a Python library designed to parse and extract information from the PubMed Open-Access XML Subset and MEDLINE XML datasets. This library simplifies the process of accessing and analyzing biomedical literature data, making it easier for researchers to integrate PubMed data into their text mining and natural language processing pipelines. ### Key features: - Parses PubMed Open-Access XML information and PubMed OA citation references. - Supports both compressed and uncompressed XML files. - Provides a dictionary format for easy data manipulation. - Extracts key information such as title, abstract, journal, PubMed ID, PubMed Central ID, DOI, and publication year. - Supports parsing of PubMed OA images and captions, and PubMed OA paragraphs. - Parses PubMed OA tables and MEDLINE XML citations from websites. ### Tech stack: Python ### Why it matters: This library is essential for researchers and data analysts who need to extract and analyze biomedical literature data. By providing a straightforward and efficient way to parse and extract information from these datasets, Pubmed Parser streamlines the process of integrating PubMed data into various research applications, enhancing the capabilities of text mining and natural language processing tools.