hermes‑claude‑auth is a Python runtime patch that lets the hermes‑agent use a Claude Code subscription after Anthropic added OAuth validation. It installs a .pth shim that monkey‑patches the agent’s request builder, adds the required billing header, fixes prompt layout, and implements auto‑wait for subscription‑window rate limits. The installer works on Linux/macOS and Windows, provides uninstall scripts, and includes git‑hook + cron recovery to survive hermes updates.

A Python wrapper for the WORLD Vocoder that allows for high-quality speech analysis and resynthesis by decomposing audio into pitch, spectral envelope, and aperiodicity.

Dioptra is NIST’s open‑source Docker‑based platform for testing AI models against trustworthiness criteria (fairness, safety, explainability, etc.). It offers a REST API, web UI, and Python client, lets you define experiments in YAML, supports plugins for PyTorch/TensorFlow, and tracks reproducible, traceable runs for developers, auditors, researchers, and red‑teamers.

ZONOS2 is a high-fidelity, multilingual text-to-speech model using a Mixture-of-Experts backbone to deliver naturalistic voice cloning and expressive emotional control.

Quantus is an open‑source Python library (PyTorch/TensorFlow) that provides 35+ quantitative metrics for evaluating neural‑network explanation methods across six categories (faithfulness, robustness, localisation, complexity, randomisation, axiomatic). It enables reproducible, batch‑efficient assessment of XAI techniques for image, time‑series, and tabular data, with tutorials, docs, and community support.

A Python quantitative‑trading platform that structures the workflow into a “Three‑Ministries‑Six‑Departments” hierarchy, offers back‑testing, multi‑strategy management, risk‑first execution, and an AI‑driven natural‑language stock‑screening feature. It runs via a FastAPI + HTML dashboard, supports multiple data sources, and is free for personal/academic use (commercial use needs a license).

Yao Open Skills is an open‑source catalogue of reusable AI “Skills” – structured prompt‑plus‑code workflows for tasks like decision analysis, security auditing, tutorial creation and reading‑report visualisation. Each Skill lives in its own folder with scripts, documentation and multi‑format export (Markdown, HTML, PDF, etc.). The repo provides a registry, publishing rules, and helper scripts to keep the collection clean, versioned and publicly discoverable.

A C# framework for audio and video fingerprinting and recognition, enabling the fast identification of media content through acoustic and video hashes.

Raix is a Ruby library that adds LLM chat, tool (function) dispatch, and prompt‑chain capabilities to any Ruby class. It wraps multiple providers via RubyLLM/OpenRouter, manages a transcript, offers global/class/instance hooks, supports JSON mode, caching, and prompt caching, and provides a DSL for declaring functions and multi‑step prompts.

Flytekit Python is the official SDK for Flyte, letting you write, test, and deploy AI/ML workflows in pure Python using `@task` and `@workflow` decorators. Install with `pip install flytekit`, follow the quick‑start guide, and extend via plugins or contribute through the provided docs.

A lightweight Python library for audio activity detection that splits audio streams into events using energy thresholding or WebRTC VAD.

An intelligent hearable system that allows users to isolate and hear a target speaker in noisy environments by looking at them for a few seconds.

Open ChatGPT Atlas is an open‑source Chrome/Edge extension (and optional Electron app) that lets an AI control the browser (via Gemini 2.5 Computer‑Use) and call over 500 third‑party services (via Composio). Install with Node, load the built `dist` folder, add your Google and optional Composio API keys, then use the sidebar chat to give natural‑language commands like “click this button” or “create a GitHub issue.” It offers visual feedback, safety confirmations, and runs entirely client‑side.

InternAgentHarness (InternBootcamp) is an open‑source framework for creating, running, and evaluating multi‑turn, tool‑augmented LLM agents. It lets researchers define synthetic tasks, expose verified external tools, manage multi‑round interactions, compute rewards, and log full trajectories for data generation or RL training.

A CLI tool that transcribes YouTube videos or local media files using Whisper, with optional AI-powered summarization and translation.

A Python interface for openSMILE that enables the extraction of standardized audio features for machine learning research.

QuietHarness is a tiny, open‑source safety layer for AI‑coding agents (Claude Code, Codex, Cursor). It provides shared configuration templates, dry‑run installation scripts, automatic backups, and risk‑gated actions so the agent won’t overwrite files, skip tests, or perform irreversible operations without confirmation. Installable per‑project or globally, it works entirely offline and is licensed under MIT.

A Streamlit web app that uses LLMs (e.g., Qwen2.5) and BGE‑M3 embeddings to extract claims from news, retrieve web evidence via DuckDuckGo/SearXNG, and verify the claims through a three‑node pipeline. Supports Chinese, English, Japanese, Korean, multiple model providers, user accounts, history, and PDF reports.

Jittor is a high‑performance deep‑learning framework that JIT‑compiles the entire computation graph and uses meta‑operators to generate specialised CUDA/C++ kernels. It offers a PyTorch‑like Python API, supports CPU and GPU back‑ends (CUDA, ROCm, Hygon), and includes a model zoo for vision, rendering, geometric learning, and reinforcement learning.

An efficient singing voice conversion project that enables high-quality AI voice changing with significantly lower hardware requirements and faster training times than traditional SVC models.

A robust forced-aligner built on Kaldi that aligns speech audio with text transcripts to provide precise word-level timing.

CodeGPT is a Go‑based CLI that uses LLM APIs (OpenAI, Azure, Gemini, Anthropic, Ollama, Groq, OpenRouter) to automatically generate Conventional‑Commit messages and brief code‑review summaries. It can install a Git `prepare‑commit‑msg` hook, supports custom prompts, language translation, streaming output, and flexible configuration (API keys, providers, proxy, diff context, etc.). Installable via Homebrew, Chocolatey, script, binary, or `go install`.

The core speech synthesis engine for VOICEVOX, provided as a C and Python library for easy integration into various applications.

DiffMem is a production‑ready FastAPI service that gives LLM‑powered agents a Git‑backed, Markdown‑based memory. Each user lives on an isolated Git branch; the writer agent commits conversation updates, and a retrieval agent uses plain `git`/`grep` commands to fetch the current view or dive into history. The system needs only an LLM API key, runs on a tiny Linux box (Docker or Coolify), and optionally mirrors branches to a private GitHub repo for backup. It includes consolidation tools (dedupe, redistribute, link) to keep the “now” files compact and well‑linked.

AgentForge is a low‑code Python framework for building, testing, and iterating AI‑driven autonomous agents. It uses declarative YAML files (Cogs) to compose agents, attach optional memory, and orchestrate multi‑agent workflows, while supporting OpenAI, Gemini, Claude, and local models via Ollama/LMStudio.

A diffusion-based training framework for Text-to-Speech, Singing Voice Synthesis, and Singing Voice Conversion that supports multi-speaker training and efficient hardware utilization.

A comprehensive toolkit providing a unified interface to over 90 metrics for evaluating the quality, intelligibility, and technical properties of speech and audio.

Crust is an open‑source, locally‑run proxy that sits between AI coding assistants and LLM providers. It inspects every tool call (file reads, shell commands, network requests) and blocks actions that could expose secrets or harm the host. The gateway works as an HTTP proxy, MCP/ACP stdio wrapper, or Docker container, and includes 42 built‑in security rules, 51 DLP patterns, and a plugin system for custom policies. It runs entirely on the user’s machine, stores logs encrypted in the OS keyring, and offers iOS integration via a Swift package.

A high-quality, low-bitrate speech codec that compresses audio into discrete tokens, optimized for use in downstream tasks like text-to-speech.

A collection of practical notebooks for fine-tuning and optimizing audio AI models, including ASR and audio captioning, using the Hugging Face ecosystem.

Kocoro is an open‑source macOS AI‑agent daemon (`shan`) that lets language‑model agents read/write files, control apps, run shell commands, automate browsers, and integrate with Slack/Telegram. It provides a CLI/TUI, HTTP API, MCP streaming, and a real‑time voice front‑brain. Install via npm, a script, or Go build; configure a Shannon Cloud or self‑hosted gateway for LLM completions. The daemon handles permissions, per‑agent memory, scheduling, and extensible tools, while a separate closed‑source GUI (Kocoro Desktop) offers a polished interface. Licensed under MIT.

Milady is a privacy‑first personal AI assistant that runs locally (or via a self‑hosted or cloud backend). It offers a desktop app with a 3‑D avatar, multi‑platform chat connectors, progressive action streaming, and built‑in BNB‑Smart‑Chain trading. Install via signed installers for macOS/Windows/Linux, or use the CLI (`milody setup`). The system is energy‑aware, plugin‑extensible, and can be securely exposed as a remote backend with an API token.

Python bindings for whisper.cpp that provide a simple API for high-performance speech-to-text transcription with support for multiple hardware acceleration backends.

A JavaScript SDK for integrating Microsoft Cognitive Services speech capabilities, enabling speech-to-text and text-to-speech functionality in browsers and Node.js.

MeanVC is a lightweight, zero-shot voice conversion system that uses mean flows and a diffusion transformer to enable real-time, single-step timbre transfer.

Claude Code Multi‑Agent is an open‑source framework that gives Anthropic’s Claude Code project awareness and expert‑agent capabilities. It uses a Python Hook system and a locally‑run Ollama model to automatically detect the codebase type, analyse user intent, load 300+ domain‑specific “Skills” (backend, frontend, testing, security, etc.), and enforce automatic documentation (DEVELOPMENT.md, KNOWLEDGE.md, CHANGELOG.md). After cloning, installing Ollama, a model, and uv, you drop your project into the repo and Claude Code instantly knows its context, suggests appropriate tools, and can be driven with slash commands like `/backend‑specialist` or workflow commands such as `/kiro/spec`. Key points: - Zero‑config start; all configuration lives in `.claude/settings.json` and `prompts.json`. - Hooks run on session start, user prompt, tool use, etc., handling project detection, intent analysis, and doc updates. - Skills are defined by `SKILL.md` files and are auto‑discovered. - Works entirely locally (no external API), integrates with Git, and is MIT‑licensed. Ideal for developers or small teams who want a reproducible, on‑premise AI coding assistant.

Next.js Evals is a Vercel‑maintained framework for automatically testing AI coding agents on real‑world Next.js tasks. It syncs eval fixtures from the `vercel/next.js` repo, runs agents in a Vercel sandbox, checks hidden assertions, records pass/fail, token usage, and cost, and exports a JSON leaderboard that powers the public Next.js evaluation site.

A local implementation of the Kokoro-82M text-to-speech model featuring a web interface, 54 multilingual voices, and full offline support.

A high-quality text-to-dialogue RVQ Transformer model for conversational English speech generation and voice cloning.

A unified open-source framework for evaluating large audio foundation models, supporting both speech understanding and generation across 34 benchmarks.

Kaldi‑Active‑Grammar is a Python wrapper for the Kaldi speech‑recognition engine that lets you compile many small grammars and activate only the needed ones per utterance. It ships binary wheels for all major OSes, includes a pre‑trained English model, and provides a backend for the Dragonfly and Caster voice‑control frameworks. Install via pip, download the model, define rules, and dynamically enable them for context‑aware command‑and‑control.

A deep learning-based Voice Activity Detection system using a 1D-ResNet and MFCC features to classify audio signals as speech or noise.

A Python speech recognition framework that allows users to create custom voice commands to automate computer activities and program by voice.

An iOS library that provides a polished voice-to-text overlay UI, handling permissions and speech recognition using Apple's native SFSpeechRecognizer.

A command line interface for macOS that enables transcription of live microphone input and audio files using the system's built-in speech recognition.

Rhino is Picovoice’s on‑device speech‑to‑intent engine that converts spoken commands into structured intents and slots in real time. It runs locally on everything from micro‑controllers to phones, browsers, and desktops, supports multiple languages, and offers SDKs for Python, .NET, Java, Flutter, React Native, Android, iOS, Web, Node.js, and C.

A framework for building end-to-end automatic speech recognition systems, providing reference implementations of 20+ ASR models and recipes for multiple languages.

A PyTorch-based automatic speech recognition framework that supports streaming and non-streaming inference across multiple model architectures and languages.