Coding & Development

Marauder is a dedicated privacy application designed to keep your location history confidential and secure. It operates on the principle that your location data should belong solely to you, ensuring that all history is saved directly on your device and never transmitted elsewhere. This commitment to on-device storage provides users with full control over their personal information, eliminating concerns about data breaches or unauthorized access. Marauder is ideal for individuals who prioritize their digital privacy and seek a reliable solution to manage their location footprint without compromising their personal data. The application emphasizes user trust and protection, making it a suitable choice for those concerned about their digital footprint.

Tonbo is an embedded database specifically designed for serverless and edge runtimes, bridging the gap between stateless compute and persistent data. It stores data as Parquet files on S3, with coordination handled through a manifest, ensuring compute remains fully stateless. The entire storage and query engine is async-first, making it ideal for modern cloud environments. Tonbo supports rich data types and declarative schemas, allowing users to query with zero-copy RecordBatch. It runs anywhere, including Tokio, WASM, and various edge runtimes, and can serve as a storage engine for custom data infrastructure. With open formats, its standard Parquet files are readable by any tool, offering flexibility and interoperability.

rl-tools is an open-source deep reinforcement learning library designed for speed and portability, making it ideal for continuous control tasks. It supports a range of popular reinforcement learning algorithms including TD3, PPO, and SAC, with examples provided for various environments like Pendulum and MuJoCo Ant-v4. The library offers C++ notebooks for documentation and local tinkering via Docker, alongside Python bindings available through PyPI for seamless integration into Python projects. Benchmarks demonstrate its efficiency across different devices and architectures, including macOS and Ubuntu, with specific optimizations for fast training. rl-tools also supports embedded platforms like iOS, Teensy, Crazyflie, and ESP32 for inference and training.

Merge Cooking® offers an engaging mobile gaming experience that blends classic merge puzzle mechanics with restaurant management simulation. Players are tasked with merging ingredients to create a wide array of delicious dishes, managing their own restaurants, and expanding their culinary empire across various global themes. The game allows users to renovate different settings like mansions, gardens, or cafes, master cooking skills, and dive into exciting stories. It provides a fun and interactive way to enjoy cooking and restaurant management, appealing to those who love casual puzzle-solving combined with a culinary theme.

Azen is a free platform designed to integrate various AI tools, including popular models like ChatGPT, DALL·E 2, and ElevenLabs. It offers users a centralized hub to access and leverage multiple AI capabilities without needing to switch between different applications. The platform is particularly well-suited for individuals interested in experimenting with artificial intelligence and for developers looking to prototype AI-powered applications. Azen aims to provide a convenient and accessible way to explore the diverse functionalities offered by different AI models.

CookBook Recipe & Meal Planner, also known as Cook AI, is an iOS mobile application designed to simplify meal preparation and recipe management. Users can effortlessly organize recipes from diverse online sources, including social media platforms and websites, converting them into clear, easy-to-follow formats. Beyond recipe organization, the app offers robust meal planning functionalities, allowing users to schedule their meals efficiently. It also generates smart grocery lists, complete with nutritional information, making it a comprehensive culinary assistant for everyday cooking needs. The app aims to streamline the entire cooking process from discovery to plate.

mcphub.nvim is a comprehensive MCP client designed for Neovim, enabling seamless integration of Model Context Protocol (MCP) servers directly into your development environment. It offers an intuitive interface for managing, testing, and utilizing MCP servers alongside popular chat plugins like Avante.nvim and CodeCompanion.nvim. Key features include real-time updates for tools, resources, and prompts, support for various MCP server transports (Streamable-HTTP, SSE, STDIO), and robust authentication options for remote servers. The tool also boasts universal configuration syntax, VS Code compatibility, and advanced workspace management with project-local configs and smart file-watching, making it a powerful solution for developers working with AI models.

EnCharge AI delivers breakthrough advances in performance, cost, and sustainability for AI computation, offering fully validated hardware and flexible software. Their technology provides 20x higher efficiency (TOPS/W), 9x higher compute density (TOPS/mm²), 10x lower Total Cost of Ownership (TCO), and 100x lower CO2 emissions compared to cloud or GPU alternatives. EnCharge AI's core technology integrates into existing semiconductor supply chains, enabling versatile products from chiplets to ASICs and standard PCIe cards for seamless orchestration between on-device and cloud deployments. This allows for broadening access to AI, enabling new capabilities on-device, and promoting sustainability and affordability in AI deployment.

FocusOnDepth is an AI tool designed for depth estimation in images, hosted as a Hugging Face Space. While the tool aims to provide capabilities for analyzing and processing images to determine depth, it is currently experiencing runtime errors due to insufficient hardware capacity. This makes it unavailable for immediate use. When operational, it would be suitable for researchers and developers interested in image processing and AI model testing, particularly those working with depth perception in computer vision applications. The tool is free to use, making it accessible for experimentation and academic purposes.

The Hallucination Evaluation Leaderboard is a dedicated platform for assessing and comparing the performance of various AI models in detecting and mitigating hallucinations. Hosted on Hugging Face Spaces by Vectara, this tool offers a live ranking system, allowing users to instantly view how different models or queries perform against a set of established metrics. It serves as a valuable resource for researchers and developers who need to benchmark their AI models, understand current industry standards, and identify areas for improvement in hallucination detection. The platform emphasizes transparency and provides a clear, real-time overview of model efficacy in this critical aspect of AI reliability.

Gemma 2 llama.cpp 2B/9B/27B is a Hugging Face Space that provides an interactive interface to the Gemma-2 language model. Users can input questions or prompts into a chat box and receive replies generated by the AI. A key feature is the flexibility to select different model sizes, specifically 2B, 9B, or 27B, catering to varying computational needs and desired output complexity. Additionally, users have control over settings such as the response length, allowing for tailored interactions. This tool is licensed under Apache-2.0, making it an open-source option for those interested in experimenting with or integrating the Gemma-2 model.

Gradio_YOLOv5_Det is an AI tool designed for object detection, leveraging the powerful YOLOv5 model. It provides a user-friendly interface built with Gradio, enabling individuals to easily upload images and perform object detection tasks. This tool is particularly useful for automating image analysis and various computer vision applications. While the live website currently shows a runtime error, the underlying purpose is to offer a straightforward way to apply advanced object detection capabilities. It is licensed under GPL-3.0, indicating its open-source nature and potential for community contributions and modifications.

iBUG Emotion Recognition is an AI tool hosted on Hugging Face that specializes in detecting emotions from facial images. Users can upload an image to the platform, and the application will automatically identify faces and determine their emotional states. The tool provides flexibility by allowing users to select different models for analysis and specify the maximum number of faces to process within a single image. This makes it suitable for various applications requiring facial analysis and emotion detection, particularly in research and development contexts. The results are displayed directly on the uploaded image, offering a clear visual representation of the detected emotions.

Thehiddenwiki is an online directory specifically designed to index and list websites and resources found on the dark web, primarily accessible via the Tor browser. It provides a curated collection of .onion links, acting as a gateway for users interested in exploring content beyond the surface web. The platform categorizes various types of hidden services, including financial services, drug marketplaces, and other commercial links, as well as informational sites and forums. The Hidden Wiki aims to offer a reliable and updated list of active dark web sites, helping users navigate this often-ephemeral part of the internet. It emphasizes its role as one of the oldest and most comprehensive link directories for the deep web.

OpenMed NER Model Discovery is a specialized tool designed to assist researchers and developers in the biomedical field in finding suitable Named Entity Recognition (NER) models. It offers a user-friendly interface with search and filter capabilities, allowing users to efficiently explore a wide range of NER models tailored for medical and clinical text analysis. The platform provides detailed model cards, including descriptions and code snippets, to help users understand each model's capabilities and integrate them into their projects. This tool streamlines the process of selecting appropriate models for tasks such as identifying diseases, drugs, genes, and other entities within medical literature or clinical notes, enhancing the accuracy and efficiency of biomedical text processing.

Archestra is an Enterprise AI Platform designed to simplify AI usage within organizations by offering a comprehensive suite of tools. It features guardrails to prevent data exfiltration and prompt injections, an MCP (Model, Code, Prompt) registry for private sharing and governance, and a Kubernetes-native MCP orchestrator for managing AI servers, API keys, and OAuth. The platform also includes robust cost monitoring, limits, and dynamic optimization to reduce AI expenses by automatically switching to cheaper models for simpler tasks. With observability features providing metrics, traces, and logs, Archestra helps platform teams, developers, and management gain full visibility into AI adoption, usage, and data access, all built on a strong security foundation.

MultiNet is an open-source AI tool designed for real-time joint semantic reasoning in autonomous driving applications. It excels at simultaneously performing road segmentation, car detection, and street classification, offering state-of-the-art performance in segmentation while maintaining real-time processing speeds. The model is built as an encoder-decoder architecture, utilizing a VGG encoder and independent decoders for each task. This repository combines several TensorFlow models, specifically KittiSeg for road segmentation, KittiBox for car detection, and KittiClass for street classification, which are included as submodules. MultiNet is compatible with the TensorVision backend for organized experiment management and requires Python 2.7 and TensorFlow 1.0.

tstorage is a lightweight, open-source, embedded time-series database designed for efficient handling of large volumes of time-series data. It features a straightforward API with massively optimized ingestion capabilities, ensuring goroutine-safe writes and reads. The database partitions data points by time, using a linear data model structure rather than B-trees or LSM trees, which is ideal for time-series workloads that are mostly append-only. It supports both in-memory and persistent disk storage, allowing users to specify a data path for on-disk persistence. tstorage also handles out-of-order data points by buffering them in memory partitions, making it robust against network latency or clock synchronization issues. This design ensures fast read operations, especially for recent data, and efficient storage by sequentially writing larger files when partitions are full.

YOLOv5-Lite is an optimized object detection model evolved from YOLOv5, designed for enhanced lightness, speed, and deployment ease. It significantly reduces model size, with versions as compact as 900kb (int8) and 1.7M (fp16), making it highly efficient for resource-constrained environments. The tool boasts impressive performance, reaching 15 FPS on a Raspberry Pi 4B, making it suitable for edge computing and embedded systems. Key optimizations include the removal of the Focus layer and four slice operations, along with the addition of shuffle channels and a YOLOv5 head for channel reduction. YOLOv5-Lite supports various frameworks and backends like PyTorch, ncnn, mnn, OpenVINO, TensorRT, and TFLite, and provides models tailored for different platforms and precision levels. It includes comprehensive documentation for installation, inference, training, and deployment on diverse hardware.

2d-gaussian-splatting provides an official implementation for creating geometrically accurate radiance fields using 2D Gaussian Splatting. This open-source project represents scenes with 2D oriented disks and utilizes perspective-correct differentiable rasterization. It includes regularizations to enhance reconstruction quality and offers various meshing approaches for Gaussian splatting, including both bounded and unbounded mesh extraction. The tool supports COLMAP and NeRF Synthetic datasets, and provides scripts for training, rendering, and evaluation of novel view synthesis and geometric reconstruction. It also features integrations with community resources like WebGL/Three.js viewers and offers performance improvements through CUDA operator fusing.

advanced_lane_detection is an open-source project designed for advanced lane detection using computer vision techniques. Developed as part of the Udacity Self-Driving Car Nanodegree, it provides a comprehensive pipeline for identifying lane boundaries in images and video streams. Key steps include camera calibration and distortion correction, creating thresholded binary images using color transforms and gradients, applying perspective transforms for a bird's-eye view, and fitting polynomial curves to detect lane lines. The tool also calculates lane curvature and vehicle position relative to the lane center, and annotates the original image with this information. It's built with Python and relies on libraries like NumPy, OpenCV, Matplotlib, and Pickle.

airframe-react is a free and open-source dashboard template designed for building high-quality admin and analytics interfaces. It leverages Bootstrap 4 and React 16, ensuring responsiveness across smartphones, tablets, and desktops. The template is available under an MIT license, making it highly accessible for developers. It features a minimalist design with an innovative Light UI, perfect for large-scale applications. The project includes React Router and customized reactstrap, with dependencies regularly updated. It offers over 10 layout variations, ready-to-use applications, a large collection of UI components, and more than 120 unique pages, making it ideal for CRMs, CMSs, Admin Panels, and Analytics dashboards.

Awesome-state-space-models is a comprehensive collection of research papers and repositories focused on state-space models and hybrid models. This GitHub repository serves as a centralized resource for academics, researchers, and engineers interested in the latest advancements and implementations in this field. It includes a wide array of topics, from foundational theories to specific applications in areas like language models, vision, reinforcement learning, and biomedical imaging. The collection is regularly updated with new arXiv preprints and conference papers, offering insights into various model architectures, optimization techniques, and practical use cases, including Mamba, RWKV, and other hybrid approaches.

Awesome-VLA-Robotics is a curated, open-source repository offering an extensive collection of resources focused on Vision-Language-Action (VLA) models in robotics. This includes a detailed list of excellent research papers, various VLA models, relevant datasets, and other valuable materials for researchers and practitioners in the field. The repository defines VLA models, outlines their core concepts, and details key components like Vision Encoders, Language Understanding modules, and Action Decoders. It also explores the relationship between VLAs, VLMs, and Embodied AI, tracing the evolution from VLM adaptation to integrated VLA systems. The resource is structured to provide quick glances at key models and datasets, categorized by application area and technical approach, making it an invaluable reference for understanding and advancing VLA robotics.