A cross-platform real-time captioning software that generates live subtitles and translations from system audio or microphone input using local or cloud-based AI models.

useAgent is an open‑source, self‑hosted AI coworker platform that lets LLM agents (Claude Code, Codex, OpenCode, Pi) operate inside isolated Linux sandboxes, browse the web, run commands, edit code, and interact with team tools (Slack, GitHub, Gmail, Notion, etc.). It provides a UI with a live timeline, terminal, and workspace, stores every run in PostgreSQL for replay, and supports reusable “skills” and a shared knowledge base. The project is written for the Bun runtime, uses pgvector for embeddings, and is released under AGPL‑3.0 with an optional commercial license.

Hibiki is a model for streaming speech-to-speech translation that generates natural target speech and text in real-time as the user speaks.

An Android input method editor (IME) that uses the Whisper engine to provide private, offline voice recognition and translation.

An audio processing monorepo for React Native and Expo that provides on-device speech-to-text, text-to-speech, and audio analysis across iOS, Android, and web.

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.

Guitarix is a Linux‑only, modular virtual guitar amp that combines classic DSP effects with AI‑based amp models (Neural Amp Modeler and RTNeural). It runs as a standalone app, VST3, or LV2 plugins, can be controlled via MIDI, and even operates headless on devices like a Raspberry Pi. The README provides build steps, dependencies, and links to documentation and a Docker‑based container for optimized deployment.

A Rust‑based CLI (also packaged for Homebrew, npm, and binaries) that rewrites Codex’s SQLite state DB and rollout JSONL logs so that chat threads remain visible after you change the `model_provider`. It offers one‑off sync, continuous watch mode, and optional background‑service installation for macOS, Linux, and Windows.

vitest‑evals is a monorepo that adds AI‑aware evaluation to Vitest. It provides harness adapters for various model SDKs, reusable judges (e.g., factuality), a custom reporter that outputs detailed JSON, a local React UI for inspecting runs, and a GitHub Action that posts summaries and can gate PR merges based on pass‑rate or scores.

PUAX is a production‑ready runtime that adds a mind‑set layer to LLM agents. It provides arena/gate/dream primitives, thin‑prompt compression, a zero‑dependency Python SDK, hybrid trigger detection, 59 built‑in (and custom) roles, outcome‑driven verification, a Carbon Shield for human‑targeted manipulation detection, and a reproducible multi‑model benchmark suite. The system is accessed via an MCP server (`npx puax-mcp-server`) that injects native hooks into popular IDEs, and can be embedded directly into LangChain/AutoGen via a one‑line AMP middleware. MIT‑licensed and marked production‑ready.

A local-first audio and video transcription tool that converts recordings into structured Markdown notes and enables AI-powered Q&A over the content.

A high-fidelity neural audio codec that compresses 44.1 kHz audio into discrete codes at 8 kbps, providing approximately 90x compression for speech, music, and environmental sounds.

A speech recognition framework for Espressif chips that enables offline wake word detection, voice command recognition, and audio preprocessing.

A Go SDK that provides a thin, idiomatic wrapper around Anthropic’s Claude API, enabling Go developers to call Claude models (e.g., Claude Opus 4.6) directly from their applications. Install with `go get`, create a client with an API key, and use the `Messages.New` method to send prompts and receive responses.

incur is a TypeScript library that lets you build command‑line tools that are easy for LLM agents to discover and use. It provides typed command definitions (via Zod), token‑efficient TOON output, built‑in MCP and skill‑file generation, and can wrap or expose HTTP APIs as CLI commands.

A Python package for text normalization and inverse text normalization, used to prepare text for speech processing systems.

Better Agent Terminal (BAT) is a cross‑platform desktop app (Windows/macOS/Linux) built with Tauri 2 and React. It lets you manage multiple project workspaces, each with split‑panel terminals, a file browser, Git viewer, snippet manager, and an embedded AI coding assistant (Claude Code or OpenAI Codex). The AI runs inside the app, offering streaming output, permission modes, session resume/fork, and detailed usage stats. Optional integration with the `cx` CLI provides lightweight semantic code navigation. Install via Homebrew, AppImage, or a one‑line script; you can also build from source with Rust and Node. Ideal for developers who want a single window to run commands and interact with LLMs safely and efficiently.

LLM4AD Next is a Python toolkit that turns a natural‑language problem description into a full, runnable evolutionary‑algorithm search project. An interactive LLM consultant generates the code skeleton, configuration, and evaluator, then built‑in evolutionary methods (IslandGA, Diverse Island GA, EoH, etc.) automatically evolve the generated algorithm. It also offers an advisor to suggest evolution‑ready code blocks, long‑term memory support, Docker images, and an online demo.

Specula is an open‑source tool that uses LLM coding agents (Claude Code, Codex, etc.) to automatically generate TLA⁺ specifications for concurrent or distributed code, model‑check them, and reproduce discovered bugs as source‑level patches. It supports auto, CI‑integrated, and interactive modes, requires Python 3.10+, Java 21+, and a capable LLM, and is released under Apache 2.0.

codexU is a native macOS menu‑bar (and Windows Tauri) app that reads local OpenAI Codex (and optional Claude Code) data to display quota percentages, token‑usage trends, project/skill rankings, a locally‑computed AI‑leadership score, and an estimated API‑cost (“羊毛进度”). All processing stays on the user’s machine; no usage data is uploaded.

OpenClaw.NET is a .NET‑native AI‑agent runtime and gateway. It provides a NativeAOT‑compatible executable, a web‑based chat/admin UI, OpenAI‑compatible APIs, 80+ built‑in tool adapters, multi‑channel support (Slack, Telegram, etc.), and first‑class integration with cloud and local LLM providers (OpenAI, Claude, Gemini, DeepSeek, Ollama, embedded Gemma‑4). Optional governance features let you audit and approve high‑risk tool actions. The project is aimed at developers who want a self‑hosted, observable AI‑agent platform that runs on Windows, macOS, or Linux.

A privacy-first, local voice-to-text tool for macOS that transcribes speech and pastes it directly into any active application.

DoWhy is a Python library for end‑to‑end causal inference. It lets you define a causal graph, automatically identifies the causal effect, estimates it with a variety of statistical methods, and runs refutation tests to check robustness. The library also supports graphical causal models for root‑cause analysis, counterfactuals, and interventional sampling. Install via `pip install dowhy` and follow the four‑step workflow (model → identify → estimate → refute).

InternAgent‑1.5 is an open‑source autonomous‑agent framework for long‑horizon scientific discovery. It can generate hypotheses, design and run (simulated) experiments, optimise algorithms (via the MLEvolve core), and answer research questions using a Deep Research pipeline. The system includes a persistent memory module, task suites for algorithm discovery and paper‑reproduction, and unified launch scripts for discovery or QA modes.

A marketplace of LSP server definitions that lets Claude Code (Anthropic’s AI coding assistant) provide IDE‑like features—go‑to‑definition, hover, find‑references, etc.—for dozens of languages. Install Claude Code ≥ 2.1.50, run `npx tweakcc --apply` to patch it, add the marketplace via `/plugin marketplace add Piebald‑AI/claude‑code‑lsps`, then install the underlying language‑server binaries (rust‑analyzer, pyright, clangd, etc.) as instructed. The repo includes validation scripts to keep the marketplace JSON in sync with the `.lsp.json` configs.

OGhidra is a genuine open‑source project that adds large‑language‑model assistance to the Ghidra reverse‑engineering platform. It lets users ask natural‑language questions, runs AI‑driven analysis (function renaming, malware pattern detection, report generation), and supports both local Ollama models and cloud APIs while keeping binaries private.

Xybrid is an open‑source, cross‑platform SDK that runs LLMs, speech‑to‑text and text‑to‑speech models locally on devices (mobile, desktop, Unity). It provides native bindings for Flutter, Swift, Kotlin, Unity, Rust and Python, supports many popular open‑source models, and lets you chain them together in a multi‑model pipeline—all without requiring cloud services.

LakeSoul is an open‑source lakehouse framework with a Rust‑native metadata and I/O core, offering ACID transactions, incremental upserts, automatic compaction, fine‑grained RBAC, and support for Spark, Flink, Presto, Ray, Daft, and Python‑based tools. It lets you store regular rows and multimodal/vector data (via the Vortex format) and provides real‑time CDC ingestion, snapshot queries, and seamless integration with AI/ML pipelines.

ByteChef is an open‑source platform that merges AI‑agent orchestration with visual workflow automation. It offers built‑in agent loops, 250+ pre‑made connectors, multi‑provider LLM support, memory/RAG back‑ends, guardrails, and a Docker‑ready self‑hosted runtime. The free core (Apache 2.0) covers the editor, agents, connectors, and polyglot code; enterprise features add API‑exposed workflows, Git‑native promotion, SSO, and AI‑gateway controls. Quick‑start via Docker Compose gets a local instance running in minutes.

A lightweight macOS voice input tool that uses ASR and LLM correction to paste cleaned-up spoken text directly into any app.

A Home Assistant integration that removes audio lag by locally combining notification chimes and TTS audio into a single seamless file for playback.

A lightweight Vietnamese text-to-speech system featuring multi-speaker synthesis and zero-shot voice cloning that runs efficiently on CPU.

Claude Sidebar is an Obsidian plugin that embeds a terminal running Claude Code (or other AI‑agent CLIs) directly in the sidebar. It auto‑launches agents, supports multiple tabs, context‑menu shortcuts, YOLO mode, and can work on any folder on your machine. Install via Obsidian’s community plugins (or manually), and it runs on macOS, Linux, and Windows using xterm.js and a small Python PTY backend.

Polymarket MCP Server is a Python‑based MCP server that lets Claude (or any LLM) autonomously discover, analyse, trade, and monitor Polymarket prediction‑market positions. It offers 45 tools across market discovery, analysis, trading, portfolio management, and real‑time monitoring, all guarded by configurable safety limits and full‑node authentication. A web dashboard provides visual control, and the project includes real‑API tests, MIT licensing, and detailed docs for both demo (read‑only) and live‑trading setups.

ChatGPT CLI is a cross‑platform command‑line client for many LLM providers. It offers streaming queries, thread‑based chat history, prompt‑file injection, multimedia I/O, and an experimental agent mode (ReAct / Plan‑Execute) that can run shell/file/web‑search tools under budget and policy constraints. It also supports the Model Context Protocol (MCP) for custom tool calls. Install via Homebrew or a pre‑built binary, configure API keys in `~/.chatgpt-cli/config.yaml`, and start chatting with `chatgpt "your question"` or `chatgpt --interactive`.

A cross-platform React Native and Expo library that provides a unified API for speech-to-text functionality across iOS, Android, and the Web.

ClaraVerse is a private AI workspace focused on AI agent teams, layered memory, and a permissive license, offering chat, multi-agent collaboration, terminal agent, workflow automation, and 150+ integrations.

A lightweight 120M-parameter text-to-speech model family that enables high-quality, zero-shot voice cloning and streaming audio synthesis on laptop CPUs or in the browser.

MidiTok is a Python package that converts MIDI and abc music files into token sequences for machine‑learning models such as Transformers, enabling music generation, transcription, and other MIR tasks.

A fast, local speech-to-text system for Home Assistant that recognizes specific known phrases and device names rather than general speech.

Matcha-TTS is a fast, non-autoregressive text-to-speech architecture that uses conditional flow matching to enable high-quality, efficient speech synthesis.

DBreeze is a pure‑C#, embeddable database for .NET that supports ACID transactions, schema‑free tables, full‑text search, and—crucially for AI—an integrated vector store with HNSW similarity search, enabling fast nearest‑neighbor queries on LLM embeddings.

An AI-powered tool for transcribing audio and video from local files or YouTube links, providing translation and subtitling capabilities.

Sudo Code is an open‑source, Rust‑based CLI coding assistant that runs LLMs (Anthropic, OpenAI, xAI, Gemini, etc.) directly in the terminal. It is model‑agnostic, file‑driven, and designed for heavy‑use developers who want full visibility and control over the agent’s actions. The single `scode` binary works interactively or as a pipe‑compatible one‑shot tool, can serve an ACP API for editors/web UIs, and supports multiple deployment modes (local, remote Nexus VFS, or managed agent). Installation is via a simple install script; the project is MIT‑licensed and aims for privacy‑first, no‑telemetry, and no vendor lock‑in.

LuCI plugin that lets OpenWrt routers host the OpenClaw AI gateway, providing a web UI for installing Node.js, managing LLM models, and enabling channels such as WeChat.

Reddit AI Trend Reports is a Docker‑based, MongoDB‑backed system that fetches posts from AI‑focused subreddits, optionally enriches them with image captioning, YouTube transcript summarisation, and web‑page scraping, then uses Groq or OpenRouter LLMs to generate daily bilingual (English/Chinese) trend reports.

A toolset consisting of a CLI and an in-game addon that uses ElevenLabs TTS to generate and play AI voiceovers for World of Warcraft NPC dialogue.

A Python package for efficient forced alignment of audio and text using Hugging Face CTC models, supporting over 1,100 languages.