Projects

### Overview: This repository is a comprehensive reading roadmap for anyone interested in deep learning. It provides a structured approach to learning from the foundational concepts to the latest advancements in the field. ### Key Features: - **Deep Learning History and Basics**: Introduces the history of deep learning, starting with the basics and moving through the milestones. - **Deep Belief Network (DBN)**: Explains the concept and its significance. - **ImageNet Evolution**: Tracks the evolution of deep learning from inception to the present. - **AlexNet and VGGNet**: Highlights the breakthroughs in image recognition and speech recognition. - **ResNet and Deep Residual Learning**: Discusses the advancements in deep neural networks. - **Speech Recognition Evolution**: Covers the development of deep learning in speech recognition. ### Tech Stack: Python ### Why It Matters: This roadmap serves as a practical guide for anyone looking to dive into the world of deep learning. It provides a systematic way to understand the evolution of the field, from foundational concepts to the latest research papers, ensuring that readers can stay updated with the most recent developments in deep learning.

### Overview: TARS is an open-source multimodal AI agent stack, currently shipping two projects: Agent TARS and UI-TARS-desktop. It connects cutting-edge AI models and agent infrastructures, aiming to provide a workflow closer to human-like task completion through advanced multimodal LLMs and seamless integration with real-world tools. ### Key Features: - **Agent TARS**: A general multimodal AI agent stack that integrates GUI Agent and Vision capabilities, designed for use in terminals, computers, browsers, and products. It includes a CLI and a Web UI for easy access. - **UI-TARS Desktop**: A desktop application based on the UI-TARS model, offering a native GUI Agent experience. It supports local and remote computer and browser operators. ### Tech Stack: TypeScript ### Why it Matters: TARS aims to streamline AI agent development and integration, making it easier to deploy and manage multimodal AI solutions in various environments. Its focus on real-world tool integration and human-like task completion makes it particularly valuable for developers and businesses looking to enhance their AI capabilities.

Overview: AirLLM is a project that enables the efficient inference of large language models (LLMs) on single 4GB GPUs, significantly reducing memory usage. It supports running models like Kimi K3 (2.8T), Qwen3.8-Flash-Next (125B), and DeepSeek-V3 (671B) on limited VRAM. Key features: - Dramatically reduces inference memory usage for large models. - Supports training of large models on small VRAM. - Enables running of the largest open-source model, Kimi K3 (2.8T), on under 4GB of VRAM. Tech stack: Jupyter Notebook Why it matters: AirLLM allows developers to run and train large language models on hardware with limited VRAM, democratizing access to cutting-edge AI technology.

Overview: PageIndex is a document indexing system that uses a hierarchical tree index and LLM reasoning to retrieve information from long and complex documents, offering a more accurate and context-aware retrieval method compared to traditional vector-based approaches. It is designed to work without vector databases or document chunking, making it ideal for professional documents like financial reports and legal documents. Key features: - Vectorless, reasoning-based retrieval - Context-aware and explainable search - Local mode for indexing and retrieval - Hierarchical tree indexing for documents Tech stack: Python Why it matters: PageIndex provides a more accurate and context-aware retrieval method for complex documents, replacing the need for vector databases and chunking with a hierarchical tree index and LLM reasoning.

### Overview: LangExtract is a Python library designed to extract structured information from unstructured text documents using LLMs, ensuring precise source grounding and interactive visualization. ### Key Features: - **Precise Source Grounding:** Maps every extraction to its exact location in the source text, enabling easy traceability and verification. - **Reliable Structured Outputs:** Enforces a consistent output schema based on user-defined examples, leveraging controlled generation in supported models. - **Optimized for Long Documents:** Overcomes the "needle-in-a-haystack" challenge of large document extraction by using optimized text chunking, parallel processing, and multiple passes. - **Interactive Visualization:** Instantly generates a self-contained, interactive HTML file to visualize and review thousands of extracted entities in their original context. - **Flexible LLM Support:** Supports various LLMs, including cloud-based models like Gemini and local open-source models via the built-in Ollama interface. - **Adaptable to Any Domain:** Defines extraction tasks for any domain using just a few examples, making LangExtract highly adaptable without requiring fine-tuning. - **Leverages LLM World Knowledge:** Influences how the extraction task may utilize LLM knowledge, improving the accuracy of the extracted data. ### Tech Stack: Python ### Why it Matters: LangExtract is essential for anyone working with unstructured text documents, such as clinical notes or reports, who needs to extract structured information in a reliable and precise manner. Its ability to map every extraction to its exact location in the source text, combined with its interactive visualization feature, makes it a

Overview: This repository contains code released by Google Research, including datasets and source files under various licenses. Key features: - Datasets and source files released under CC BY 4.0 International license. - Source files released under Apache 2.0 license. - Subdirectories for easy downloading and cloning. Tech stack: Jupyter Notebook Why it matters: Google Research releases open-source code and datasets, making it accessible to the developer community. Repository: google-research/google-research GitHub stars: 38,931

Overview: Tesseract.js is a pure JavaScript library designed to extract text from images in over 100 languages, providing a powerful OCR solution for developers. Key features: - Supports over 100 languages - Works in the browser and Node.js - Integrates seamlessly with webpack, esm, and CDN - Provides a simple and efficient way to recognize text in images Tech stack: JavaScript Why it matters: Tesseract.js simplifies the process of extracting text from images, making it easier for developers to integrate OCR functionality into their projects, especially in environments where traditional OCR libraries are not suitable or available.

Overview: The 500 AI Agents Projects is a curated collection of AI agent use cases across various industries, showcasing practical applications and providing open-source projects for implementation. Key features: - A curated collection of 500+ AI agent projects - Production examples, tutorials, and working code - Spanning major frameworks (LangGraph, CrewAI, AutoGen, Agno) and industries (Healthcare, Finance, Education, Cybersecurity, etc.) - Illustrating how AI agents are transforming sectors Tech stack: Python Why it matters: This repository provides a comprehensive overview of AI agent projects, offering practical insights and real-world implementations for developers, researchers, and teams looking to evaluate and implement AI agent frameworks and technologies. Repository: ashishpatel26/500-AI-Agents-Projects GitHub stars: 38,761 GitHub description: The curated collection of AI agent use cases across various industries, showcasing practical applications and open-source projects for implementation.

Overview: Langchain-Chatchat（原Langchain-ChatGLM）是一个基于Langchain与ChatGLM等语言模型的RAG与Agent应用的开源项目，旨在提供一套对中文场景和开源模型支持友好的离线私有部署的知识库问答解决方案。 Key features: - 支持多种开源模型和模型部署框架（如GLM-4、Qwen2、Llama3等） - 可离线私有部署，无需网络连接 - 支持多种开源模型和向量数据库 - 能够调用OpenAI GPT API - 实现原理图和视频介绍 Tech stack: Python Why it matters: 本项目提供了一种基于本地知识库的问答解决方案，适用于对中文场景和开源模型有需求的应用场景，特别适合需要离线部署和快速响应的场景。

Overview: Anthropic's Interactive Prompt Engineering Tutorial is a comprehensive course designed to teach users how to create effective prompts for Claude, their language model. The course includes nine chapters with exercises and an appendix for advanced techniques, allowing learners to practice and experiment with prompt engineering. Key features: - Comprehensive step-by-step guide to prompt engineering - Interactive "Example Playground" for experimenting with prompts - Structured into beginner, intermediate, and advanced levels - Covers common failure modes and techniques to address them Tech stack: Jupyter Notebook Why it matters: This tutorial provides a practical and interactive approach to mastering prompt engineering for Claude, enhancing user interaction and understanding of the model's capabilities.

### Overview: Prompt Optimizer is an AI prompt optimization tool designed to enhance the quality of AI-generated prompts and outputs. It supports various methods for prompt creation and optimization, including manual writing, template usage, local imports, and integration with the Prompt Garden. The tool helps users craft more effective and structured prompts, leading to better AI results. ### Key Features: - **Prompt Optimization**: Enhances the clarity and structure of prompts to improve AI output quality. - **Prompt Templates**: Allows for the reuse of prompts across different contexts, such as marketplaces and role-based conversations. - **Variable Optimization**: Enables the optimization of variables within prompts to create more contextually relevant and transactional responses. ### Tech Stack: The project is built using TypeScript, a modern programming language that supports type safety and is well-suited for AI applications. ### Why It Matters: Prompt Optimizer significantly improves the effectiveness and relevance of AI-generated prompts, making them more useful and impactful in various applications. By optimizing prompts, it ensures that AI outputs are more aligned with the intended goals and context, leading to better user experiences and outcomes. ### Repository: The project is hosted on GitHub, with 38,366 stars, indicating its popularity and wide adoption. It is also available as a Chrome extension, Docker deployment, and as a web application, making it versatile for different use cases. ### Primary Language: The primary language used in the project is TypeScript, a modern programming language that is well-suited for AI applications. ### GitHub Stars: The project has received 38,366 stars on GitHub

Overview: This repository offers a wealth of production-ready projects, tutorials, and real-world applications in AI engineering, covering LLMs, RAGs, and AI agents. Key features: - 93+ Projects - In-depth tutorials on LLMs, RAG, Agents, and more - Real-world AI agent applications - Examples for implementation, adaptation, and scaling Tech stack: Jupyter Notebook Why it matters: This resource is essential for anyone looking to learn and build with AI, providing a comprehensive guide for beginners, practitioners, and researchers alike. Repository: patchy631/ai-engineering-hub GitHub stars: 38,350

Overview: CopilotKit is a comprehensive SDK designed for building full-stack agentic applications, Generative UI, and chat applications. It simplifies the integration of agents into various channels, including React, Angular, Vue, React Native, and Slack, Microsoft Teams, and more. Key features: - **Chat UI**: A customizable chat interface that supports message streaming, tool calls, and human-in-the-loop workflows. - **Shared State**: Enables shared state across different platforms and channels. - **Human-in-the-Loop Workflows**: Facilitates workflows where agents can learn and adapt based on real user interactions. Tech stack: TypeScript Why it matters: CopilotKit simplifies the integration of agents into various channels, making it easier to build full-stack agentic applications and Generative UI. It is the company behind the AG-UI Protocol, adopted by Google, LangChain, AWS, Microsoft, Mastra, PydanticAI, and more. Repository: CopilotKit/CopilotKit Primary language: TypeScript GitHub stars: 38082 GitHub description: The Frontend Stack for Agents & Generative UI. React, Angular, Mobile, Slack, and more. Makers of the AG-UI Protocol

### Overview: llmfit is a terminal tool designed to find the best performing Large Language Models (LLMs) that can run on your hardware. It analyzes your system's CPU, RAM, GPU(s), VRAM, and accelerator configuration to recommend models across popular quantizations. ### Key Features: - **Hardware Auto-Detection**: Detects CPU cores, system RAM, available discrete/integrated GPUs, VRAM, and unified memory architecture (NVIDIA CUDA, Apple Silicon, AMD ROCm, Intel OneAPI). - **Model Compatibility Engine**: Analyzes model parameter counts, context lengths, and quantization formats (GGUF, AWQ, GPTQ, EXL2) to project memory footprints and tokens-per-second performance. - **Interactive TUI & Web Dashboard**: Choose between a lightweight, zero-dependency terminal interface or a feature-rich web dashboard. - **REST API Endpoint**: Exposes standard HTTP JSON endpoints (`/api/v1/system`, `/api/v1/models`) for integration into orchestrators, dashboards, and automated deployment pipelines. - **Multi-Platform Support**: macOS (Apple Silicon & Intel), Linux (x86_64 & ARM64), and Windows (x86_64). - **Hundreds of models & providers. One command to find what runs on your hardware.** ### Tech Stack: Rust ### Why It Matters: llmfit helps developers find the optimal LLM models for their hardware, ensuring better performance and resource efficiency. It simplifies the process of selecting models that fit within the constraints of their system

Overview: 这是一门由吴恩达教授编写的机器学习课程的个人笔记，涵盖了监督学习、无监督学习和机器学习最佳实践等内容，旨在帮助学习者掌握机器学习的核心技术和应用。 Key features: - 监督学习（参数/非参数算法，支持向量机，核函数，神经网络） - 无监督学习（聚类，降维，推荐系统，深度学习推荐） - 机器学习最佳实践（偏差/方差理论；机器学习和人工智能创新过程） Tech stack: HTML Why it matters: 本课程通过详细的案例研究和实用技术，帮助学习者掌握机器学习的核心技术和应用，是人工智能领域中非常普遍和重要的内容。 Repository: fengdu78/Coursera-ML-AndrewNg-Notes Primary language: HTML GitHub stars: 37,975 GitHub description: 吴恩达老师的机器学习课程个人笔记 README excerpt: 课程目录：[目录](index.html) 课程地址：<https://www.coursera.org/course/ml> Machine Learning(机器学习)是研究计算机怎样模拟或实现人类的学习行为，以获取新的知识或技能，重新组织已有的知识结构使之不断改善自身的性能。它是人工智能的核心，是使计算机具有智能的根本途径，其应用遍及人工智能的各个领域，它主要使用归纳、综合而不是演译。在过去的十年中，机器学习帮助我们自动驾驶汽车，有效的语音识别，有效的网络搜索，并极大地提高了人类基因组的认识。机器

### Overview Khoj is a personal AI app designed to enhance your capabilities. It seamlessly integrates with any local or online language model, enabling you to access answers from the internet and your own documents. Khoj can be used from your browser, Obsidian, Emacs, desktop, phone, or WhatsApp. It supports creating custom agents with specific knowledge, personas, chat models, and tools, automating repetitive research, and generating images and playing messages. Khoj is self-hostable and open-source, always available. ### Key Features - **Chat with any LLM**: Connects to various language models like LLaMA, Qwen, and others. - **Internet and Document Answers**: Fetches information from the web and your documents. - **Custom Agents**: Create agents with specific knowledge and personas. - **Automation**: Automate repetitive research tasks. - **Personal Newsletters**: Receive personalized newsletters. - **Smart Notifications**: Get smart notifications for your inbox. - **Advanced Search**: Use advanced semantic search for quick document access. - **Image Generation**: Generate images using Khoj. - **Talk Out Loud**: Enable voice-based interactions. - **Self-Hostable**: Run Khoj privately on your computer or on our cloud app. - **Enterprise**: Available as a cloud service, on-premises, or hybrid solution. ### Tech Stack Khoj is built using Python and leverages various open-source tools and libraries for its functionality. ### Why It Matters Khoj offers a powerful tool for anyone looking to enhance their productivity and knowledge acquisition

### Overview: GFPGAN is a powerful tool designed for real-world face restoration, leveraging the capabilities of StyleGAN2 to enhance and restore faces in practical applications. ### Key Features: - **Practical Algorithm for Real-world Face Restoration**: GFPGAN aims to provide practical solutions for face restoration in real-world scenarios. - **Blind Face Restoration**: It can restore faces even when the input quality is poor, making it highly versatile. - **Model Zoo**: GFPGAN includes a model zoo with different versions, each offering slightly different levels of detail and performance. ### Tech Stack: - Python ### Why It Matters: GFPGAN is essential for anyone working on face restoration, as it provides a practical solution for restoring faces in various real-world conditions, including low-quality and high-quality inputs. This tool is particularly useful in fields such as digital media, gaming, and facial recognition systems. ### Repository: - **Repository**: [TencentARC/GFPGAN](https://github.com/TencentARC/GFPGAN) - **Stars**: 37,701 - **Description**: GFPGAN aims at developing Practical Algorithms for Real-world Face Restoration. ### GitHub Actions: - **PyLint**: Ensures code quality with static analysis. - **Publish-Pip**: Automatically publishes the package to PyPI for easy installation. ### Online Demos: - **Replicate**: Offers an online demo for real-time face restoration. - **Huggingface Gradio**: Provides a Gradio web demo for interactive face restoration. - **Colab Demo**: A Colab

Overview: Open Assistant is a chat-based assistant designed to understand tasks, interact with third-party systems, and retrieve information dynamically. It aims to democratize access to large language models, aiming to revolutionize language innovation in the same way stable-diffusion has done for art and images. Key features: - Task Understanding - System Interaction - Dynamic Information Retrieval Tech stack: Python Why it matters: By providing everyone with access to a powerful language model, Open Assistant aims to enhance language capabilities and potentially transform various fields through improved communication and information retrieval. Repository: LAION-AI/Open-Assistant GitHub stars: 37,374 GitHub description: OpenAssistant is a chat-based assistant that understands tasks, can interact with third-party systems, and retrieve information dynamically to do so.

### Overview: Vane is an AI-powered answering engine that runs entirely on your own hardware, combining knowledge from the internet with support for local LLMs and cloud providers, delivering accurate answers with cited sources while keeping your searches completely private. ### Key Features: - **Support for all major AI providers** - Use local LLMs through Ollama or connect to OpenAI, Anthropic Claude, Google Gemini, Groq, and more. Mix and match models based on your needs. - **Smart search modes** - Choose Speed Mode for quick answers, Balanced Mode for everyday searches, or Quality Mode for deep research. - **Pick your sources** - Search the web, discussions, or academic papers. More sources and integrations are in progress. - **Widgets** - Helpful UI cards that show up when relevant, like weather, calculations, stock prices, and other quick lookups. - **Web search powered by SearxNG** - Access multiple search engines while keeping your identity private. Support for Tavily and Exa coming soon for even better results. - **Image and video search** - Find visual content alongside text results. Search isn't limited to just articles anymore. - **File uploads** - Upload documents and ask questions about them. PDFs, text files, images - Vane understands them all. - **Search specific domains** - Limit your search to specific websites when you know where to look. Perfect for technical documentation or research papers. - **Smart suggestions** - Get intelligent search suggestions as you type, helping you formulate better queries. - **Discover

### Overview: MediaPipe Solutions is a cross-platform, customizable library for on-device machine learning, designed to simplify the integration and deployment of AI and ML features in applications. It provides a suite of solutions for vision, text, and audio tasks, and includes tools for customization and evaluation. ### Key Features: - **MediaPipe Tasks**: Cross-platform APIs for deploying solutions. - **MediaPipe Models**: Pre-trained models for use with each solution. - **MediaPipe Model Maker**: Customize models for solutions with your data. - **MediaPipe Studio**: Visualize, evaluate, and benchmark solutions in the browser. - **Legacy Solutions**: End support for legacy solutions as of March 1, 2023. ### Tech Stack: C++ ### Why it Matters: MediaPipe Solutions simplifies the integration and deployment of AI and ML features in applications, making it easier for developers to add cutting-edge technology to their products without the complexity of custom development.