kyr0

Re-inventing the webdev ecosystem for determinism and simplicity - full featured; batteries included. React-like, Next.js-like, Shadcn-like, jQuery-like

Browser extension for editors and professionals engaged in text-related research, writing, and evaluation tasks. This tool serves as a comprehensive "AI toolbox" enabling direct interaction with content on any website for extraction, translation, summarization, re-phrasing and more.

Encodes text and binary data to a subset of Unicode plane 1 (BMP). Runs in browser and Node.js. Symbol compression algo. Turns binary data into short, copy-pastable UTF-8 strings. URI-safe in modern browsers (for pasting, locally).

Cross-platform shared memory stream/buffer, header-only library for low-level IPC in C/C++ with FFI bindings for Rust, Zig and Go.

WASM based, asynchronous Brotli compression and decompression package that works in browsers and Node.js; CommonJS and ESM

A novel algorithm for encoding/decoding and character level compression/decompression with character-level compression rates of >500%. The compressed data is copy-pasteable without data loss.

An Astro project template for decent projects: auth, i18next, Bootstrap, sitemap, webworker, robots.txt, preact, react, endpoints, endpoint clients, OAuth, various Astro features and data loading preconfigured

Astro 2.0 / turbo monorepo project template featuring hybrid rendering (SSG + SSR). Configured to deploy on Vercel.

Simple vLLM container deployment for Qwen3-Omni-30B-A3B-Instruct with up-to-date CUDA

Extanium is a OOP javascript framework for rapid & cross-plattform mobile application development in Appcelerator Titanium - ExtJS rewrite

A design system that paints the web in 8 bits.

FastAPI to serve Qwen-ASR with streaming support. Tested. Benchmarked. Flash Attention 2. Fast & Stable.

217 kB, tiny offline reverse geocoding library that works anywhere, browser, Node.js, web worker. High performance (S2 cell based). Looks up country and nearest city, given GPS coordinates

dotProduct(vectorsA, vectorsB) / matmul in JIT-optimized JavaScript, WebAssembly/WASM using SIMD vector instructions. pthread, WebGPU, WebGL, WebNN experiments.

Java-App for creating and validating Factur-X / ZuGFeRD / X-Rechnung invoices conforming with EU-Norm EN 16931.

A highly efficient, isomorphic, full-featured, multilingual text search engine library, providing full-text search, fuzzy matching, phonetic scoring, document indexing and more, with micro JSON state hydration/dehydration in-browser and server-side.

Quickly demonstrates how to pass data between JS and WASM context in the most efficient way (shared memory).

Microtune is an AU/VST MIDI FX plugin. It allows to detune all 12 chromatic tones of any virtual instrument individually. +/- 100 cents per semitone. All major DAWs supported.

In-browser, multi-lingual vector embedding and search

A simple HTTP vhost reverse proxy / port sharing implementation for node.js. Run multiple node HTTP servers (Domains, IP's) through one port.

A node native module that allows to read shared memory cross-platform (raw and using StreamReader / StreamWriter abstractions)

Vanilla JSX + HTML + CSS compiler and static-site generator (SSG) [experimental]

Universal library to integrate with LLMs. Supports Node.js, Deno, Bun, Service Workers, Web Extensions.

Milky music visualizer for Mac, inspired by Ryan Geiss' work on the "Geiss" music visualizer

Easy, fast and WASM/WebGPU accelerated vector embedding for the web platform. Locally and via API. Compatible with Browsers, Workers, Web Extensions, Node.js & co.

LZ-based compression algorithm for TypeScript projects (Browser, Node.js)

A AssemblyScript-like (TypeScript-like, JavaScript-like) language that compiles to native code using ts.asm

Milky music visualizer for Mac, inspired by Ryan Geiss' work on the "Geiss" music visualizer

Waldwicht model family - Gemma 4 quantized down to < 3 GB for the Apple Silicon platform + Technical report on the extensive ablation study that resulted in component-based quantization.

A simple cli app to trim single mp3 files (based on ffmpeg)

Write Quality Prompts - TypeScript/JavaScript edition. Use and evaluate prompting techniques quickly.

Grafting script and vLLM container inference runtime Makefile for kyr0/Ornith-35B-FP8-E4M3-MTP

Multi-threading, SIMD-optimized Audio Engine for the web (experimental)

MilkyMilky music visualizer public releases

A Node.js library that extracts and structures text from HTML files for full-text search indexing.

Simple and fast inference server implementation for Prism ML 8B MLX 1bit

An open source guide to learning all facets of fundamental music theory, composing and electronic music production. It's a lifetime project.

A smart source code globber that collects relevant code and puts it into your agent.md: npx update-agents-md

CrisprWhisper for Mac - MLX - inference code and 8b quants

No description.

Inspects and analyzes HuggingFace models by name/URL provided and estimates/optimizes on-premise inference viability with CLI run/serve command recommendations for inference engines

A fun project to predict the next german lottery numbers using an Attention LSTM neural network trained on the last 1000+ draws aka: A fancy random number generator ;)

Extremely fast and smart HTML + JS + CSS minifier, available for Rust, Node.js, and WASM

Google langdetect for the web platform. Detects 55 languages of of the box. Minimum 1 sentence input for >90% accuracy.

Linux support tool/driver for PCE-PB-N-series scales

Website of redaktool.ai

An experimental web-based synthesizer

Everything I know about running LLMs locally

No description.

Deploy Qwen3.6-35B-A3B (Q4_K_XL) + MTP speculative decoding on a single NVIDIA L4 24GB — GCP g2-standard-8 — via the official llama.cpp Docker image. Decode-optimized to ~92–100 tok/s, lossless (full GPU residency + ECC-off).

Calculate term specificity for a single doc with BM25 formula by using Wikipedia IDF

Local LiteLLM proxy deployment for macOS

SAM 3D Objects with Multi-view Images

Give a query, get a dataroom. Pi + self-hosted Qwen3.6 research harness on a single L4.

⚠️ Experimentelle Skill-Sammlung für deutsches Recht (Arbeits-, Gesellschafts-, Insolvenz-, Datenschutz-, Prozessrecht u.a.) – inzwischen verbessert und im Alltag getestet, aber weiterhin Experiment. Bitte selber ausprobieren, Issues/PRs willkommen! Keine Rechtsberatung. Mandatsgeheimnis (§§ 203/204 StGB, § 43e BRAO), DSGVO, US-Transfer, KI-VO & Co

Build product integrations with AI.

No description.

Aidana.ai local voice agent for macOS

SkillOpt with local AI is a text-space optimizer that trains reusable natural-language skills for frozen LLM agents through trajectory-driven edits, validation-gated updates, and deployable best_skill.md artifacts.

Learn it. Build it. Ship it for others.

Webpage for a wealth of learning for many things Windows NT visit: https://empyreal96.github.io/nt-info-depot/index.html

Lightning-Fast, On-Device, Multilingual TTS — running natively via ONNX.

Run LLMs with MLX

The cpp inference of BiRefNet based on Tensorrt.

No description.

The best-benchmarked open-source AI memory system. And it's free.

Just a simple multimodal avatar interaction platform

HeadAudio: An audio node/processor for real-time audio-driven viseme detection and lip-sync in browsers.

Talking Head (3D): A JavaScript class for real-time lip-sync using full-body 3D avatars.

MLX-Embeddings is the best package for running Vision and Language Embedding models locally on your Mac using MLX.

LLM inference server with continuous batching & SSD caching for Apple Silicon — managed from the macOS menu bar

MLX: An array framework for Apple silicon

Autoresearch for GPU kernels. Give it any PyTorch model, go to sleep, wake up to optimized Triton kernels.

No description.

vLLM + Qwen3.5-122B-A10B-NVFP4 on NVIDIA DGX Spark (GB10/SM121) — single-GPU NVFP4 W4A4 with MTP speculative decoding, self-contained Docker build

Outrageous Voice Assistant - Fully local end-to-end ASR + LLM + TTS pipeline using open weight models and a simple web based UI

SGLang Omni: High-Performance Multi-Stage Pipeline Framework for Omni Models

Just talk to your agent — it learns and EVOLVES.

No description.

No description.

For testing ASR models with remote/public file support

Qwen3-ASR is an open-source series of ASR models developed by the Qwen team at Alibaba Cloud, supporting stable multilingual speech/music/song recognition, language detection and timestamp prediction.

Hello, world! :)

Fast local inference of Qwen3-TTS on Mac, backed by mlx-audio

A text-to-speech (TTS), speech-to-text (STT) and speech-to-speech (STS) library built on Apple's MLX framework, providing efficient speech analysis on Apple Silicon.

APL is a Turing-complete, domain-specific language for writing multi‑step, branching LLM workflows (*Agentic Workflows*) in a **Markdown‑flavoured Jinja-syntax**

VS Code extension for APL - Syntax highlighting & more.

Ludicrously fast multicore math for web browsers using WebAssembly, SIMD-128 and Web Workers

Emulator that can run 32-bit Windows programs/games on multiple platforms via Wine

Frontier CoreML audio models in your apps — text-to-speech, speech-to-text, voice activity detection, and speaker diarization. In Swift, powered by SOTA open source.

A simple inference server (OpenAI compatible) and a trivial WebUI for running/testing MLX models locally on Mac.

Training code for kokoro tts model

https://hf.co/hexgrad/Kokoro-82M

Verbatim Automatic Speech Recognition with improved word-level timestamps and filler detection

No description.

Build your personal knowledge base with Trilium Notes

Hetzner Online Community Project

Chrome extension to extract LinkedIn profile data and send it to Zapier, Make, or any webhook-based automation tool.

Optimizing inference proxy for LLMs

No description.

Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek-R1, Qwen3, Gemma 3, TTS 2x faster with 70% less VRAM.

The best ChatGPT that $100 can buy.

'RoBERTa' (Liu et al., 2019) implementation from scratch in PyTorch

Qwen3-omni is a natively end-to-end, omni-modal LLM developed by the Qwen team at Alibaba Cloud, capable of understanding text, audio, images, and video, as well as generating speech in real time.

No description.

The Grain (Educational) Language, Standard Library, Assembler, Disassembler, Bootloader, Operating System Kernel and Software Distribution

Far Cry 1 Full Source (Developed by CryTek). For NON COMMERCIAL Purposes only. Leaked.

No description.

An audio compressor plugin created with JUCE

A lightweight screen recorder based on ScreenCapture Kit for macOS / 基于 ScreenCapture Kit 的轻量化多功能 macOS 录屏工具

Camera and Microphone streaming library via RTMP and SRT for iOS, macOS, tvOS and visionOS.

Simple load tests for ollama LLM and embedding inferencing

A TTS model capable of generating ultra-realistic dialogue in one pass.

Website of the JSCraftCamp in Munich

The Milky music visualizer for Linux.

Tensor computation with WebGPU acceleration

The best API for running dos programs in browser

Emscripten port of 3D Pinball for Windows – Space Cadet decompilation

Source code of Windows XP (NT5). Leaks are not from me. I just extracted the archive and cabinet files.

A transparent and safe image converter for Node.js, written in Rust. Encodes PNG, JPG, TIFF to WebP

IOTA Wallet

Simple but powerful prompt and RAG workflow templating language.

AIME machine learning container (MLC) management system

Witness the aha moment of VLM with less than $3.

A free and open-source block-based email template builder.

Library for audio processing and synthesis

Wait-free thread-safe single-consumer single-producer ring buffer using SharedArrayBuffer

A Web Audio API synth for demoscene written in literate javascript

IADT Y4 - Final Year Project: A web-based subtractive synthesizer 🥭🎹

Oberheim OB-X inspired synth in a browser

Fast, JavaScript DSP library that creates JIT optimized audio callbacks using code generation techniques

a js library to compile optimized audio callbacks inspired by gen~

ZUgferd validator using Verapdf, discontinued because integrated into Mustangproject 2.0

An adapter for enabling Rayon-based concurrency on the Web with WebAssembly.

Collection of patches for puppeteer and playwright to avoid automation detection and leaks. Helps to avoid Cloudflare and DataDome CAPTCHA pages. Easy to patch/unpatch, can be enabled/disabled on demand.

A transparent, fast and safe image converter for Node.js, written in Rust. Encodes PNG, JPG, TIFF to WebP

Extremely fast JavaScript minifier, available for Rust and Node.js

qpwgraph - A PipeWire Graph Qt GUI Interface

A JS-independent, tree-shakeable CSS framework for building faithful recreations of the Windows 7 UI.

A CSS framework for building faithful recreations of operating system GUIs.

a Fast Fourier Transform (FFT) library that tries to Keep it Simple, Stupid

Just some random guides and docs.

Build faster websites with Astro's next-gen island architecture 🏝✨

A template to demonstrate how to write custom renderers and integrate third-party renderers/frontend frameworks in Astro.

BlackHole is a modern macOS audio loopback driver that allows applications to pass audio to other applications with zero additional latency.

The Geiss screensaver and Winamp music visualization plug-in

State-of-the-art Machine Learning for the web. Run 🤗 Transformers directly in your browser, with no need for a server!

Free full-featured web-based audio & waveform editing tool

StreamFX is a plugin for OBS® Studio which adds many new effects, filters, sources, transitions and encoders! Be it 3D Transform, Blur, complex Masking, or even custom shaders, you'll find it all here.

Generalist and Lightweight Model for Named Entity Recognition (Extract any entity types from texts) @ NAACL 2024

Learn how to use logit bias with OpenAI models to create highly-powerful classifiers in minutes.

Baguetter is a flexible, efficient, and hackable search engine library implemented in Python. It's designed for quickly benchmarking, implementing, and testing new search methods. Baguetter supports sparse (traditional), dense (semantic), and hybrid retrieval methods.

Transformer Explained: Learn How LLM Transformer Models Work with Interactive Visualization

Put together a multilingual corpus from a variety of sources. Used for wordfreq and word embeddings.

Simple word to frequency mappings for the german language based on text corpora and using CISTEM stemmer.

Port of Google's language-detection library to Python.

Fast Open-Source Search & Clustering engine × for Vectors & 🔜 Strings × in C++, C, Python, JavaScript, Rust, Java, Objective-C, Swift, C#, GoLang, and Wolfram 🔍

ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator

No description.

No description.

High-performance In-browser LLM Inference Engine

An ONNX converter script focused on embedding models

The application worker driven frontend framework

Template language for prompting text-to-* AI models. Supports variables, defaults, conditions, loops, meta data and safely compiles. Runs in all JS/TS environments.

No description.

Minimal CSS Framework for semantic HTML

Original Soundtrack of RaptorGPT - Call of the Hallucination

Automatized research at the touch of a button 👆

Lars's datasets

System headers for interfacing WebGPU from C programs compiled via Emscripten to WebAssembly

Header-only C++/python library for fast approximate nearest neighbors

hnswlib-wasm attempts to create a browser friendly version of hnswlib

Specific, up-to-date knowledge for LLM coding experts.

Astro documentation

The Wysiwyg Editor for Markdown: 100% CommonMark + GFM with Uploads and Image Resizing

No description.

FFmpeg for browser, powered by WebAssembly

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding

On-device speaker diarization powered by deep learning

The Milkdrop2 Music Visualization Standalone, only for Windows

Incredibly fast JavaScript runtime, bundler, test runner, and package manager – all in one

No description.

Open source, client-side cookie consent management for your Astro website

No description.

Astroplate is a free starter template built with Astro, TailwindCSS & TypeScript providing everything you need to jumpstart your Astro project. Get started with Astroplate and save yourself hours of work!

Universal and Transferable Attacks on Aligned Language Models

Finds Instagram location IDs near a specified latitude and longitude.

🌟 Build beautiful, high-performance documentation websites with Astro

A Vercel project with an edge function to validate and send HTML form data for free. With captcha-free, top notch spam protection. (Using Sendgrid and OpenAI gpt-3.5-turbo)

Node.js bindings for faiss

No description.

Enjoy our curated collection of examples and solutions. Use these patterns to build your own robust and scalable applications.

Astro version of the Popular Stablo Blog Template

Attention-based CNN-LSTM and XGBoost hybrid model for stock prediction

Comparing Long Term Short Memory (LSTM) & Gated Re-current Unit (GRU) during forecasting of oil price .Exploring multivariate relationships between West Texas Intermediate and S&P 500, Dow Jones Utility Avg, US Dollar Index Futures , US 10 Yr Treasury Bonds , Gold Futures.

Time Series forecasting using Seasonal ARIMA & Prophet. Applied statistical tests like Augmented Dickey–Fuller test to check stationary of series. Checked ACF ,PACF plots to identify Moving average and Auto-regressive order of series. Transformed series to make it stationary.

Fork with a patch for int3 single-step debugging

Created with StackBlitz ⚡️

Next generation frontend tooling. It's fast!

The official distribution of olcPixelGameEngine, a tool used in javidx9's YouTube videos and projects