Projects

### Overview: CARLA is an open-source simulator designed for autonomous driving research. It supports the development, training, and validation of autonomous driving systems. CARLA features open-source code, open digital assets, and flexible sensor specifications, making it a valuable tool for researchers and developers. ### Key Features: - **Open-Source Code and Protocols**: Supports the development, training, and validation of autonomous driving systems. - **Open Digital Assets**: Urban layouts, buildings, and vehicles created for autonomous driving research. - **Flexible Sensor Suite**: Supports various sensor types and configurations. - **Simulation Platform**: Supports flexible specification of sensor suites and environmental conditions. ### Tech Stack: C++ ### Why it Matters: CARLA is essential for autonomous driving research, providing a robust and flexible platform for developing, testing, and validating autonomous driving systems. Its open-source nature and open digital assets make it a valuable resource for researchers and developers alike. ### Repository: carla-simulator/carla ### GitHub Stars: 14,503 ### Recommended System: - **Intel i7 gen 9th - 11th / Intel i9 gen 9th - 11th / AMD Ryzen 7 / AMD Ryzen 9** - **+32 Gb RAM memory** - **NVIDIA RTX 3070/3080/3090 / NVIDIA RTX 4090 / NVIDIA RTX 5090 or better** - **16 Gb or more VRAM** - **Ubuntu 22.0

Overview: Dlib is a modern C++ toolkit providing machine learning algorithms and tools for real-world applications. It is designed to be used in C++ to solve complex problems. Key features: - Machine learning algorithms - Data analysis tools - Modern C++ toolkit Tech stack: C++ Why it matters: Dlib simplifies the process of creating applications that integrate machine learning and data analysis, making it accessible to developers without requiring deep knowledge of complex algorithms and frameworks.

Overview: Flair is a powerful, state-of-the-art NLP library that simplifies the application of advanced models to natural language tasks. Key features: - State-of-the-art NLP models for various tasks including named entity recognition (NER), sentiment analysis, part-of-speech tagging, biomedical text support, and sense disambiguation. - Simple interfaces for combining different word and document embeddings. - Built on PyTorch, making it easy to train custom models and experiment with new approaches. Tech stack: Python Why it matters: Flair provides a straightforward and efficient way to integrate cutting-edge NLP techniques into projects, making it accessible for developers without deep expertise in machine learning.

### Overview: This project provides a TensorFlow implementation of the face recognition algorithm described in the paper "FaceNet: A Unified Embedding for Face Recognition and Clustering". It uses ideas from the paper "Deep Face Recognition" by the Visual Geometry Group at Oxford. ### Key features: - Face recognition using Tensorflow - Pre-trained models trained on CASIA-WebFace and VGGFace2 datasets - Compatibility with Tensorflow r1.7 under Ubuntu 14.04 with Python 2.7 and Python 3.5 - Updated to Tensorflow r1.0 with Continuous Integration using Travis-CI ### Tech stack: Python ### Why it matters: This project is valuable for developers looking to implement face recognition systems using TensorFlow, as it provides a pre-trained model and compatibility with popular frameworks.

Overview: Deeplearning4J is a comprehensive suite of tools designed for deploying and training deep learning models on JVM-based systems. It supports a wide range of languages and formats, including Keras, TensorFlow, and ONNX. Key features: - Supports importing Keras models from h5, TensorFlow models (1.0.0-beta7), and ONNX models. - Modular and lightweight C++ library for math operations. - Java-based math library atop the core C++ library. - SameDiff: Automatic differentiation framework for deep learning. - DataVec: ETL for machine learning data in various formats. - LibND4J: C++ library underpinning the JVM. - Python4J: Bundled cpython execution for JVM. Tech stack: Java Why it matters: Deeplearning4J simplifies the process of building and deploying deep learning models, making it accessible to developers using JVM-based languages like Java, Scala, Kotlin, and more. It supports various languages and formats, including Keras, TensorFlow, and ONNX, making it a versatile tool for deep learning projects.

### Overview: OpenCLIP is an open-source implementation of CLIP, a state-of-the-art visual-textual similarity model. It provides a comprehensive set of tools and resources for developers and researchers to integrate CLIP into their projects. ### Key features: - **CLIP Model**: Implements the CLIP model architecture. - **Coca Model**: Enhances CLIP with a contrastive learning component. - **NaFlex CLIP**: Variable-resolution image towers with token-budget batching. - **NaFlex CLAP**: Audio-text contrastive training with variable-duration audio. - **NaFlex GenLIP / GenLAP**: Generative image/audio captioning with prefix-LM attention. - **Modern Text Tower**: RoPE, SwiGLU/ReLU², RMSNorm, masked pooling. - **Variable-Length Text**: Pads captions to the per-batch max. - **Hugging Face ModernBERT Text Towers**: Supports modern text towers. - **MaMMUT**: Single text decoder with two passes. - **CoCa v2 Configs**: Corrected attentional pooling and CLS/pad attention mask. ### Tech Stack: Python ### Why it Matters: OpenCLIP offers a robust and versatile toolkit for integrating CLIP into various applications, making it a valuable resource for developers and researchers in the fields of computer vision and natural language processing.

Overview: 本项目是一个面向小白开发者的大模型应用开发教程，旨在通过一个课程帮助初学者掌握大模型开发的基础入门技能。 Key features: - LLM 开发入门 - 如何调用大模型 API - 知识库搭建 - 构建 RAG 应用 - 验证迭代 Tech stack: Jupyter Notebook Why it matters: 本项目通过实践教学，帮助开发者快速入门大模型开发，减少学习门槛，提高学习效率。

### Overview: Semantica is a graph-native infrastructure designed for context and accountable AI systems. It provides a developer-first, knowledge infrastructure that allows for the ingestion of enterprise data, extraction of what matters, and building context graphs and knowledge graphs for deterministic reasoning and decision intelligence. Built with open source, governed, and zero vendor lock-in, Semantica supports polyglot graph storage, RDF, LPG, and W3C standards, making it interoperable and scalable. ### Key Features: - **Context Management**: Manages the context of AI systems. - **Knowledge Modeling**: Models and extracts relevant knowledge from data. - **Deterministic Reasoning**: Runs graph analytics and causal reasoning with full decision provenance. - **Ontology Management**: Supports ontology management for context and accountability. - **Decision Intelligence**: Provides end-to-end traceability for decision-making processes. ### Tech Stack: - Python ### Why It Matters: Semantica is designed for high-stakes, regulated domains, ensuring accountability and trustworthiness in AI systems. It offers a developer-first approach, open source, and interoperability, making it an alternative to expensive enterprise platforms.

Overview: AX is a high-throughput, declarative orchestrator designed to run billions of autonomous agent workloads in a cluster. It simplifies the management of agent tasks, including sandboxed execution, warm Git repos, and model specifications. Key features: - Declarative task management - Isolated sandbox execution with CPU/memory limits - Pre-wired Git repos and warm start of workspaces - Model specifications for LLMs and credentials - Pause and resume agents - SSH access to running agents Tech stack: Go Why it matters: AX provides a set of small primitives that handle the complexities of running autonomous agents declaratively, making it easier to manage and scale agent workloads in a Kubernetes cluster. Repository: google/ax GitHub stars: 13,902 GitHub description: Google's open agentic orchestration runtime

Overview: LitGPT is a collection of 20+ high-performance language models designed for enterprise-scale training, fine-tuning, and deployment. It includes recipes for pretraining, fine-tuning, and deploying models at scale. Key features: - From scratch implementations - No abstractions - Beginner-friendly - Flash attention - FSDP - LoRA, QLoRA, Adapter - Reduce GPU memory (fp4/8/16/32) - 1-1000+ GPUs/TPUs - 20+ LLMs - Recipes (YAML) Tech stack: Python Why it matters: LitGPT provides optimized, enterprise-ready language models that are fast, minimal, and performant, making them suitable for large-scale applications and production environments. Repository: Lightning-AI/litgpt GitHub stars: 13714 GitHub description: 20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.

### Overview: OpenCreator is an open-source AI workspace designed for creators, offering a unified platform for video translation, creation of videos, images, voiceovers, and more. It combines the power of Codex CLI with a stable local runtime, a visual workspace, and a desktop host, making it suitable for use on various platforms. ### Key Features: - **Content Workspace**: Use visual tools and creation templates for video translation, downloading, image and video generation, voiceovers, article and social post writing, short-video scripting, and stick figure animation. - **Agent Conversation**: Start and guide creative or development tasks in natural language, organize conversations by project, keep runs working in the background, and manage approvals, attachments, files, Skills, MCP, schedules, notifications, memory, and diagnostics from one place. ### Tech Stack: - TypeScript ### Why It Matters: OpenCreator provides a streamlined workflow for creators, allowing them to focus on their creative and development work without the need to reinvent the wheel. The integration of Codex CLI ensures that the AI capabilities are seamlessly integrated into the workspace, making it a valuable tool for anyone looking to enhance their content creation and development processes. Repository: krillinai/OpenCreator GitHub Stars: 13,625

### Overview: EverOS is a Python library and local-first memory runtime designed for AI agents and makers. It stores conversations, files, and agent trajectories as readable Markdown, ensuring local SQLite and LanceDB indexes for fast retrieval and continuous evolution across apps, tools, and workflows. ### Key Features: - **Markdown Source of Truth**: EverOS uses Markdown to store conversations, files, and agent trajectories, making them easily editable and version-controlled. - **Direct File Editing**: Users can directly edit Markdown files, and the changes are synced via a cascade watcher. - **Local Three-Part Stack**: It combines Markdown, SQLite, and LanceDB for a local, efficient memory layer without the need for managed services or vector databases. - **User & Agent Tracks**: Separate user and agent tracks for distinct surfaces, allowing for clear separation of user and agent interactions. - **Orthogonal Retrieval**: Search functionality by user, agent, app, project, and session IDs for precise and flexible access. - **Knowledge Wiki**: Editable, source-backed Markdown knowledge pages with taxonomy, CRUD APIs, and topic search capabilities. - **Offline Memory Evolution**: Continuous evolution of offline memory to refine profiles and skills between sessions, enhancing long-term memory retention. ### Tech Stack: Python ### Why It Matters: EverOS provides a robust, local-first memory layer that integrates seamlessly with various AI platforms, ensuring that agents and workflows have a consistent and up-to-date memory source. This makes EverOS a valuable tool for developers and makers looking to enhance the memory capabilities of their AI agents and workflows. ### Repository: EverMind

### Overview: 本项目提供了一个全面的AI学习路线图，涵盖了从基础的Python和数学到深度学习、计算机视觉和自然语言处理等热门领域的实战案例和项目。通过这些内容，旨在帮助初学者快速入门AI并进行实战项目，同时提供配套教材和丰富的学习资源。 ### Key Features: - **必备基础技能**: - **必备Python基础**: 通过视频课程快速入门Python编程。 - **必备数学基础**: 理解基本数学概念，如高数、线性代数和概率论。 - **必备Python工具包**: 熟悉Numpy、Pandas、Matplotlib和Seaborn等常用工具包。 - **机器学习**: - **机器学习算法**: 介绍各种机器学习算法及其应用。 - **机器学习算法实验分析**: 实战分析算法性能。 - **机器学习算法代码复现**: 学习并复现机器学习算法代码。 - **机器学习经典案例实战**: 实战多个机器学习经典案例。 - **机器学习实战集锦**: 合集多个机器学习项目。 - **数据分析与挖掘**: - **数据挖掘实战**: 实战数据挖掘技术。 - **数据挖掘竞赛优胜解决方案**: 学习竞赛中优秀解决方案。 - **数据分析实战**: 实战数据分析方法。 - **深度学习**: - **深度学习必备算法**: 介绍深度学习中的关键算法。 - **深度学习必备工具**:

### Overview: This script aims to remove all AI features from Windows 11 22H2, including Copilot, Recall, and other AI-driven components, to improve user experience, privacy, and security. ### Key Features: - **Disable Registry Keys**: - Disable Copilot - Disable Recall - Disable Input Insights and typing data harvesting - Disable AI Experiments Program in Paint - Remove AI Fabric Service - Disable AI Actions - Disable Voice Access - Disable AI Voice Effects - Disable AI in Settings Search - Disable Gaming Copilot - Disable Copilot in All Office Apps - Disable AI Features in Photos App - Disable Click to Do in Snipping Tool - **Prevent Reinstall of AI Packages**: - Installs custom Windows Update package to prevent reinstall of AI packages in the CBS (Component-Based Servicing) store - **Disable Copilot Policies**: - Disables policies related to Copilot and Recall in IntegratedServicesRegionPolicySet.json - **Remove AI Appx Packages**: - Removes all AI appx packages including `Nonremovable` packages and WindowsWorkload - **Remove Recall Optional Feature** - **Remove AI Packages in CBS**: - This will remove hidden and locked AI packages in the CBS (Component-Based Servicing) store - **Remove AI Files**: - Full system cleanup removing all remaining AI installers, registry keys, and package files - **Hide AI Components**:

Overview: This repository highlights the top AI papers of each week, providing a curated list to keep you updated on the latest advancements in the field. Key features: - Weekly AI paper summaries - Easy-to-read format - Comprehensive coverage of AI research Tech stack: Markdown, GitHub Pages Why it matters: Stay informed about the cutting-edge developments in AI by following this repository.

### Overview: LangChain4j is an idiomatic, open-source Java library designed to simplify the integration of Large Language Models (LLMs) into Java applications. It provides a unified API for popular LLM providers and embedding stores, making it easy to experiment with different models without rewriting code. The library integrates seamlessly with enterprise Java frameworks like Quarkus and Spring Boot, offering a comprehensive toolbox for building LLM-powered applications. ### Key Features: - **Unified APIs:** LangChain4j offers a single interface for integrating with various LLM providers and embedding stores, reducing the need to learn and implement specific APIs. - **Comprehensive Toolbox:** The library includes practical code for common tasks such as prompt templating, chat memory management, and function calling. It also provides a wide variety of patterns and abstractions for building LLM-powered applications, including Agents and RAG. - **Examples:** The library comes with numerous examples that demonstrate how to build various LLM-powered applications, providing inspiration and enabling quick development. ### Tech Stack: Java ### Why it Matters: LangChain4j simplifies the integration of LLMs into Java applications, making it easier to experiment with different models and build LLM-powered applications. It integrates seamlessly with enterprise frameworks, ensuring compatibility with existing Java infrastructure. The library is designed with best practices in mind, including type safety, POJOs, and annotations, making it idiomatic and easy to use.

Overview: The AI Gateway is a fast, reliable, and secure solution for integrating language models into applications. It supports over 1,600 language models and offers features like automatic retries, fallbacks, load balancing, and guardrails. Key features: - Blazing fast (<1ms latency) with a tiny footprint (122kb) - Battle-tested with over 10 billion tokens processed daily - Enterprise-ready with enhanced security, scalability, and custom deployments Tech stack: TypeScript Why it matters: The AI Gateway simplifies integrating AI models into applications, providing fast, reliable, and secure routing to multiple language models, vision, audio, and image models. Repository: Portkey-AI/gateway Primary language: TypeScript GitHub stars: 132,120 GitHub description: A blazing fast AI Gateway with integrated guardrails. Route to 1,600+ LLMs, 50+ AI Guardrails with 1 fast & friendly API.

### Overview: Voice-Pro is a powerful AI-powered web application designed for creators, researchers, and multilingual professionals. It combines YouTube video processing, speech recognition, translation, and text-to-speech functionalities into a single, user-friendly tool. ### Key Features: - **Speech Recognition**: Uses Whisper, Faster-Whisper, and Whisper-Timestamped for top-tier accuracy. - **Voice Cloning**: Offers F5-TTS, E2-TTS, CosyVoice (including Fun-CosyVoice3 for Korean and additional languages), and Edge-TTS. - **Translation**: Supports instant translation for 10 languages. - **YouTube Processing**: Integrates YouTube video downloading and audio extraction. - **Audio Separation**: Separates vocals from background audio for improved clarity. ### Tech Stack: Voice-Pro leverages Whisper, Edge-TTS, Gradio, CUDA, Faster-Whisper, and other cutting-edge technologies for seamless integration and performance. ### Why It Matters: Voice-Pro simplifies the process of creating high-quality multimedia content by combining essential AI tools into a single, intuitive interface. Whether you're a content creator, researcher, or multilingual professional, Voice-Pro streamlines your workflow, ensuring that every project is of the highest quality.

Overview: LEANN is a groundbreaking vector database designed to democratize personal AI, transforming your laptop into a powerful RAG system that can index and search through millions of documents while achieving **97% storage savings** compared to traditional solutions without any accuracy loss. Key features: - RAG Everything: Index and search through millions of documents on your personal device. - Graph-based selective recomputation: High-degree preserving pruning for efficient indexing. - On-demand computing: Embeddings are computed only when needed, significantly reducing storage. - Personal data manager: Semantic search for file system, emails, browser history, chat history, and agent memory. - Live data integration: RAG on live data from any platform, including Slack, Twitter, and external knowledge bases. - Code integration: Transform your development workflow with Claude Code. Tech stack: Python Why it matters: LEANN offers a **97% storage savings** while maintaining **100% privacy** and **97% accuracy**, making it an essential tool for anyone looking to harness the power of AI without compromising on performance or security. Repository: StarTrail-org/LEANN GitHub stars: 13032

### Overview: txtai is an all-in-one AI framework designed for semantic search, LLM orchestration, and language model workflows. It provides a comprehensive solution for embedding creation, vector search, and integration with language models, making it a versatile tool for developers and AI enthusiasts alike. ### Key Features: - 🔎 Vector search with SQL, object storage, topic modeling, graph analysis, and multimodal indexing. - 📄 Create embeddings for text, documents, audio, images, and video. - 💡 Pipelines powered by language models that run LLM prompts, question-answering, labeling, transcription, translation, summarization, and more. - 🍽️ Workflows to join pipelines together and aggregate business logic. txtai processes can be simple microservices or multi-model workflows. - 🤖 Agents that connect embeddings, pipelines, workflows, and other agents to autonomously solve complex problems. - 🔋 Batteries included with defaults to get up and running fast. - ☁️ Run local or scale out with container orchestration. ### Tech Stack: Python 3.11+, Hugging Face Transformers, Sentence Transformers, FastAPI. ### Why It Matters: txtai is a comprehensive solution for embedding creation, vector search, and integration with language models, making it a versatile tool for developers and AI enthusiasts alike. Its all-in-one approach simplifies the process of building and integrating various AI components, making it an essential tool for anyone looking to enhance their AI workflows.