#WebAssembly

This project integrates WebAssembly with llama.cpp, facilitating AI model inference in browsers without backend or GPU reliance. It offers TypeScript support, high-level APIs for completions and embeddings, and low-level APIs for fine-tuned operations. Recent updates include offline model caching and custom logger support. Models can be loaded in parallel, with inference processes running in a worker to avoid blocking UI interfaces. Demo applications are available on Hugging Face Spaces to showcase its features.

Genact is an application designed to simulate computer activities that appear productive or indicative of waiting time. It runs across platforms like FreeBSD, Linux, macOS, and Windows, utilizing WebAssembly for browser support. Users can adjust their experience with various modules, like ansible and cryptomining, setting speed factors and selecting specific tasks for local or web execution. Genact is straightforward to use without dependencies and can be easily installed via Homebrew, MacPorts, Scoop, or Cargo.

The llama-cpp-wasm project provides WebAssembly builds and bindings, enabling developers to integrate Llama models into web applications. It supports single and multi-threaded deployment, enhancing JavaScript interactions via WebAssembly. The project includes comprehensive setup instructions, from cloning the repository to executing shell commands. Online demos and code examples facilitate testing and implementation of AI functionalities, with support for models like 'stablelm-2-zephyr-1_6b' from Hugging Face. This flexibility allows for improved AI model performance in web environments.

OnnxStream is a memory-efficient inference library for running large models like Stable Diffusion on devices with limited memory. It separates the inference engine and model weights for versatile data handling, achieving significant memory reductions compared to traditional methods. Ideal for resource-constrained environments, recent updates include WebAssembly and Stable Diffusion XL support, enhancing machine learning potential on compact hardware.

Discover DECtalk, an open-source text-to-speech project featuring the latest code integrations and cross-platform compatibility across Linux, Windows, and MacOS. Connect with the community on Discord or GitHub, download updates, or test in-browser through WebAssembly. Experience enhanced speech synthesis for improved accessibility and development potential.

Hash-WASM provides rapid hash calculations using optimized WebAssembly binaries, supporting algorithms like MD5, SHA-256, and BLAKE3. Its modular design allows for efficient resource use and tree shaking, making it suitable for diverse data streams. Zero dependencies and concurrent hash calculations, combined with an open-source development model, ensure reliable performance and an easy-to-use promise-based API.

The experimental SDK assists in converting .NET Core applications to standalone WASI-compliant WebAssembly files, compatible with both console and ASP.NET Core projects. It facilitates integration with WASI environments using tools like wasmtime and wasmer, and includes support for ASP.NET Core static file bundling and networking API usage.

Sherpa-ncnn provides local real-time speech recognition and voice activity detection across various platforms such as Android, iOS, Windows, macOS, and Linux. It supports architectures like x86, x64, ARM, and RISC-V and works with programming languages including C++, Python, and Go. This project minimizes external dependencies by relying solely on the ncnn framework, enhancing portability through static linking. Users can access ample resources like detailed documentation and practical demo videos, which offer insights into operations such as developing Android apps and deploying pre-trained models.

Pyodide facilitates the execution of Python packages in both web browsers and Node.js through WebAssembly. It supports pure Python packages available on PyPi and incorporates numerous widely-used packages with C extensions. Featuring robust interoperability between JavaScript and Python, Pyodide ensures smooth language integration. Users can experiment with Pyodide in-browser using a REPL without any installation. The project provides several ways to begin, such as utilizing a hosted version or compiling from source. Initiated in 2018 under Mozilla's Iodide project, Pyodide now functions as an independent, community-powered open-source initiative.

Experience a server-less Linux virtual environment in browsers using a comprehensive client-side solution. Utilizing HTML5 and WebAssembly, it offers a secure sandbox for x86 binaries execution. With CheerpX powering an x86-to-WebAssembly JIT compiler and Linux syscall emulator, it ensures ABI compatibility with Debian. Network functions are supported through Tailscale VPN. Customize with Dockerfiles, and deploy locally or via GitHub Pages with tailored uses like Python3 REPLs. WebVM assures continued reliability with immutable builds, freely accessible for individuals.

This module describes converting Tesseract OCR from C to JavaScript WebAssembly. It includes comprehensive Docker-based compilation instructions and details specific changes for enhanced functionality. Suitable for developers, it covers aspects like page angle detection, parameter control, and progress logging in web and node settings. The modified repository ensures optimized OCR performance across various platforms.

Discover FILTER.js, a comprehensive JavaScript library for image and video processing, utilizing HTML5 technologies such as WebGL, WebAssembly, and Web Workers. Offering an array of filters and plugins for synchronous and parallel processing with CPU and GPU support, it is versatile for both browser and Node.js applications. This framework supports custom builds with a wide selection of filters for efficient media processing, making it ideal for integrating advanced image manipulation and real-time computer vision into projects.

WasmEdge delivers a top-notch WebAssembly runtime optimized for web applications, edge computing, and serverless environments, facilitating secure execution of bytecode from C++, Rust, and JavaScript. Its lightweight, extensible design enables seamless integration with Kubernetes and Dapr, boosting performance and security in embedded applications, smart contracts, and SaaS platforms. WasmEdge's fast execution and robust isolation make it ideal for running user-defined code securely, providing flexibility across various technological landscapes.

Explore the integration of seamless virtual backgrounds in live video streams, leveraging cutting-edge ML models such as BodyPix, MediaPipe Meet, and ML Kit Selfie Segmentation directly in your browser. This demo is optimized for performance across various devices and rendering pipelines, utilizing technologies like WebGL, WebAssembly, and TensorFlow Lite. Designed for developers interested in real-time video processing, the project provides insights into implementing and enhancing background segmentation techniques. Discover how these technologies can seamlessly transform video communications.