PocketSphinx is an open‑source, offline speech‑to‑text engine (C library with Python bindings) that runs efficiently on low‑resource devices. It provides a command‑line tool and APIs for recognizing or aligning audio, outputting results as JSON.

A Manim plugin that integrates AI voiceovers and recording tools directly into Python, enabling per-word animation synchronization using OpenAI Whisper.

A simple Azure Speech Service module that uses the Microsoft Edge Read Aloud API to provide text-to-speech synthesis for server-side runtimes.

A local-first voice studio for macOS and iOS that uses MLX and Swift to generate high-quality speech, featuring voice cloning and custom voice design without cloud dependency.

A Python library for offline text-to-speech conversion that utilizes native system engines to generate voice output without an internet connection.

A Python/C library and toolset for forced alignment, automatically synchronizing audio narrations with their corresponding text fragments.

A Payload CMS plugin that adds AI‑powered text, image, and voice generation to content fields, supports multiple model providers, and offers fine‑grained configuration and access control.

Sponsio is an open‑source library that adds deterministic, microsecond‑scale guardrails to LLM‑powered agents. It lets you write formal‑method‑backed contracts (YAML rulebooks) that are enforced on every tool call—blocking, warning, or allowing the action without invoking any model. The library integrates with LangChain, Claude, OpenAI, CrewAI, and other agent frameworks in Python or TypeScript, ships with 22 ready‑made contract bundles, and offers a hosted console for audit and review. Benchmarks show >95 % reduction of misaligned behavior and latency under 0.2 ms per call, making it a fast, reliable safety layer for production AI agents.

Harbor is a CLI‑driven Docker‑Compose orchestrator that lets you spin up a complete local LLM stack—including model back‑ends (Ollama, llama.cpp, vLLM, MLX, etc.), chat front‑ends (Open WebUI, LibreChat, AnythingLLM, …), web‑search, voice, image generation, and Boost‑based agentic workflows—with a single `harbor up` command.

aidevops is an MIT‑licensed AI‑DevOps framework that provides a CLI and OpenCode plugin for orchestrating AI agents across code, infrastructure, product, marketing and more. It breaks work into 14 main agents and thousands of sub‑agents, uses a pulse supervisor to route tasks, manage token budgets, keep secrets safe, and maintain Git hygiene. Install via npm or Homebrew, then run commands like `aidevops init`, `aidevops pulse`, or domain‑specific slash commands to automate builds, deployments, reports, and business processes while staying auditable.

Norse is a PyTorch‑based library that adds spiking neural‑network primitives (LIF, LSNN, etc.) for building, training, and evaluating event‑driven models. It offers pip/conda/Docker installation, ready‑made example tasks (MNIST, CIFAR‑10, cart‑pole), and integrates with PyTorch‑Lightning for scalable training.

Dingo is an open‑source Python library (and SaaS product) for evaluating AI data, model outputs, and RAG systems. It supports rule‑based, LLM‑based, and agent‑driven checks, streams data from files, databases, S3, or Hugging Face, and provides JSON reports plus optional visual dashboards. Install via `pip install dingo-python` (with optional extras), run via CLI (`dingo eval …`) or Python SDK, and extend with custom rules, LLM prompts, or agents.

An Obsidian plugin that uses Whisper to transcribe audio recordings or uploaded files into text, with optional LLM post-processing for grammar and formatting.

A Python library for audio data augmentation that provides a wide variety of transforms to make audio deep learning models more robust for real-world use.

DataChain is a Python library that turns files in S3/GCS/Azure (or local) into versioned, typed datasets stored in a lightweight SQLite DB. It provides fast, sub‑second filtering, joins, aggregations and vector similarity over millions of records without moving the raw files. Pipelines are written as ordinary Python functions; the engine handles parallel execution, checkpointing and incremental updates. Optional features include an auto‑generated markdown Knowledge Base for humans/LLMs and a skill that makes the data layer visible to Claude, Cursor, Codex, Copilot or Pi. A hosted “Studio” service adds shared dataset registry, distributed compute and UI for large media. Install with `pip install datachain` and (optionally) `datachain skill install --target claude`.

Agently is a Python framework for building reliable AI services. It provides structured request/response contracts, instant streaming of partial results, observable actions (tool calls), versioned skills, and a signal‑driven workflow engine (TriggerFlow). The library abstracts model providers, supports model pools, and includes built‑in actions for Python, shell, web browsing, SQLite, etc. Designed for production‑grade assistants, internal copilots, and AI‑backed APIs that need debuggability, restart‑safe workflows, and strict output validation.

GoalBuddy is an npm‑based CLI that creates a file‑backed “board” inside your repo to orchestrate long‑running coding tasks with LLM agents like Codex and Claude Code. It lets you define a goal, break it into safe, verifiable slices, dispatch those slices to a chosen harness, capture a YAML receipt, and verify the work against an oracle (tests, demos, etc.). Boards persist across tools, enabling seamless hand‑off between agents, audit trails, and parallel sub‑goals, all with a tiny runtime footprint.

OASIS is a Python CLI tool that uses locally hosted LLMs (via Ollama or any OpenAI‑compatible server) to scan source code for security vulnerabilities. It runs a fast scan model, then deep‑analysis models, validates findings deterministically, caches results, and outputs canonical JSON plus HTML/PDF/Markdown reports. A password‑protected web dashboard lets users explore findings and ask the assistant for re‑investigations. It supports multi‑model runs, suppression registries, diffing against baselines, and CI exit‑code gating.

A Rust‑based, step‑by‑step course that guides you to implement an in‑memory vector database, covering Arrow tables, DataFusion integration, and several ANN indexes (IVFFlat, NSW, HNSW, IVF‑PQ), with SQL support and benchmark tools.

A fast neural vocoder that synthesizes high-quality audio waveforms from acoustic features using spectral coefficients and inverse Fourier transform.

An extension of openai-whisper that provides accurate word-level timestamps and confidence scores for multilingual automatic speech recognition.

brms is an R package that provides a high‑level, lme4‑style interface for fitting Bayesian generalized (non‑linear) multilevel models using Stan. It supports many response families, flexible priors, and a suite of post‑fit diagnostics and visualizations, enabling users to perform sophisticated Bayesian regression without writing Stan code directly.

A streamlined installer for Open Voice OS, enabling users to deploy a privacy-focused, open-source voice assistant on Raspberry Pi, Linux, or Mac hardware.

Skills Manager is a native macOS app (and accompanying terminal UI) that centralises discovery, installation, testing and organisation of coding‑agent “skills” (prompt extensions) for dozens of LLM‑based developers such as Claude Code, Cursor, Gemini CLI, OpenHands, etc. It stores original skill packages in a local Library, groups them into reusable Collections, and mounts them into agents via symlinks, providing conflict detection, translation, and offline‑first operation.

A command-line tool and Python API for mining multilingual pronunciation data from Wiktionary to create grapheme-to-phoneme (G2P) datasets.

oneDNN (oneAPI Deep Neural Network Library) is an open‑source, cross‑platform performance library that supplies highly optimized CPU and GPU kernels for deep‑learning operators. It implements the oneAPI specification, supports Intel, AMD, Arm, and experimental Power/IBM z/RISC‑V hardware, and is used by major frameworks such as PyTorch, TensorFlow, ONNX Runtime, and llama.cpp.

TensorFlow Quantum is a Python library that integrates Cirq quantum circuits with TensorFlow/Keras, providing high‑performance hybrid quantum‑classical machine‑learning models, automatic differentiation, and scalable simulation via qsim.

A high-performance Rust implementation of the Kokoro TTS model providing fast text-to-speech synthesis via a CLI, Rust crate, and OpenAI-compatible API server.

A collection of audio processing scripts for tasks like vocal separation, loudness matching, and transcription to prepare datasets for AI training.

MineBench is a web‑based benchmark that evaluates LLMs on 3‑D spatial reasoning by having them output voxel coordinates for prompts like “a medieval castle”. It visualises the builds, lets humans compare them head‑to‑head, and ranks models with a Bradley‑Terry system. The platform includes an Arena for pairwise voting, a Sandbox for custom builds, a Gallery of community prompts, and export options (GLB, STL, .vox, .schem). It supports major LLM providers and can be run locally with Node.js, pnpm, and Docker.

A lightweight browser automation tool for AI agents, implemented as a single Zig binary with no Node.js dependencies.

GolemBot is a Node.js runtime that gives existing coding agents (Cursor, Claude Code, OpenCode, Codex) a chat‑bot body. With a single `golembot gateway` command you can expose the agent on Slack, Telegram, Discord, Feishu, DingTalk, WeCom, WeChat, or a custom HTTP API, and swap LLM providers (OpenRouter, MiniMax, DeepSeek, etc.) via a tiny YAML config. It integrates the ClawHub marketplace (13 k+ community skills) so agents can gain new abilities instantly. Features include a built‑in dashboard, cron scheduler, fleet management, and a minimal JavaScript SDK (`createAssistant`). Install with `npm i -g golembot`, run `golembot onboard` to configure, then `golembot gateway` to go live. MIT‑licensed.

roam‑code is a local static‑analysis CLI (and optional MCP server) that builds a symbol graph of a repository, letting AI coding agents query definitions, callers, test impact, and risk (blast radius) without any remote API calls. It provides pre‑flight checks, health summaries, verification gates, and Claude‑specific hooks, all open‑source and privacy‑preserving.

The repository is a genuine software project about AI desktop and browser automation, providing a Rust‑based MCP server that lets AI agents control the computer via JSON‑RPC over stdio, with local OCR and optional vision model integration.

Naomi is an open‑source, always‑on voice‑assistant platform (Python, Linux/Raspberry Pi) that lets you control home devices, query information, and add custom “skills” via simple Python modules.

Clawdbot is an open‑source, locally‑run AI assistant that connects to Claude, GPT, Gemini and other LLMs via configurable API‑proxy endpoints. It offers web and terminal UIs, Telegram/Discord/WhatsApp bots, and stores all data on the host. The README covers installation (Node 22+, npm or script), an onboarding wizard for model and channel setup, detailed JSON configuration for multi‑model proxies, common pitfalls (Node version, env‑var misuse, missing fields), and a full set of CLI commands for managing the gateway, logs, and diagnostics.

Easy Voice Toolkit bundles open‑source speech models (Whisper, GPT‑SoVITS) into a GUI/Colab workflow for cleaning audio, transcribing it, building a voice‑conversion dataset, training a custom voice model, and generating converted speech. It offers a ready‑to‑run Windows portable package and a Colab notebook, with a developer‑friendly install guide. The project is academic‑use only and plans future LLM chatbot integration and Linux support.

JAX‑CFD is a research‑grade, JAX‑based library for differentiable CFD. It provides finite‑volume and pseudospectral solvers, optional ML‑augmented models, and data utilities, enabling gradient‑based experiments on turbulent flow simulations. The package is installable via pip, comes with Colab demos, and is now unmaintained, with newer alternatives recommended.

A self-hosted web application that uses OpenAI's Whisper to locally transcribe audio into text or subtitles in various formats.

A framework for synthesizing music and audio loops by applying diffusion models to mel spectrograms, supporting latent diffusion and conditional generation.

Paper Pilot is a desktop (Electron + React) research assistant that crawls 8+ scholarly databases, downloads PDFs, indexes them with SQLite + FTS5 + vector search, and lets you ask a local or hosted LLM for evidence‑grounded answers. All data stays on your machine; the app provides audit‑ready citation traces and a reproducible workflow for literature review.

ch.at is a single‑binary Go server that lets you chat with an LLM (default GPT‑4o) via HTTP, SSH, DNS, or API, with no JavaScript, no accounts, and all state kept in RAM. It’s privacy‑focused, easy to self‑host, and includes a public board for posting messages over the same lightweight protocols.

EasyJailbreak is an open‑source Python framework that modularises LLM jailbreak attacks. It provides ready‑made recipes (ReNeLLM, GPTFuzz, AutoDAN, etc.) and lets users plug in custom selectors, mutators, constraints, and evaluators to generate, run, and score adversarial prompts against target language models. Install via `pip install easyjailbreak` or from source, then use the high‑level API (e.g., `PAIR` recipe) or build your own attack pipeline. The project includes documentation, a paper, datasets, and downloadable experimental results.

Torch.rb is a Ruby gem that wraps the PyTorch C++ library (LibTorch), giving Ruby developers a full‑featured deep‑learning API. It supports tensors, autograd, neural‑network modules, optimizers, and GPU (CUDA/MPS) acceleration, while mirroring the PyTorch Python API with Ruby‑style method names. Installation requires a matching LibTorch build, after which you can write and run models (e.g., CNNs, GANs) directly in Ruby.

An end-to-end multimodal model for real-time spoken dialogue that supports direct audio understanding and personalized voice cloning.

A Python utility that reconstructs one audio source using the most spectrally similar frames from another audio or video source.

A machine learning-supported lo-fi music generator that uses a VAE model to represent and generate musical tracks through feature vectors.

JJazzLab is an open application that generates realistic, dynamic backing tracks with drums, bass, guitar, and piano based on chord symbols and selected music styles.