#1
World Monitor is a real-time global intelligence dashboard built by Elie Habib that aggregates AI-synthesized news, geopolitical signals, financial data, and infrastructure tracking into a unified situational awareness interface. The project ingests 500+ curated feeds across 15 categories from 65+ external data providers, covering geopolitics, finance, energy, climate, aviation, cyber, military, and news intelligence. A standout feature is its dual map engine combining a 3D globe (globe.gl/Three.js) and a WebGL flat map (deck.gl/MapLibre GL) with 56 layer types, alongside a proprietary Country Instability Index (CII v8) scoring 31 Tier-1 countries and a 7-signal finance market composite tracking 29 stock exchanges, commodities, and crypto. The platform ships six domain-specific variants (world, tech, finance, commodity, happy, energy) from a single TypeScript/Vite codebase, with native desktop apps via Tauri 2 for macOS, Windows, and Linux, and supports 25 languages including RTL. Notably agent-friendly, it exposes an MCP server, REST API with OpenAPI spec, an official CLI (`worldmonitor` npm package), and zero-dependency SDKs in Python, Ruby, and Go, making it well-suited for AI agent pipelines and programmatic intelligence workflows. Local AI inference is supported via Ollama with no API keys required, and cloud inference via Groq and OpenRouter is also available. The project is actively maintained under AGPL-3.0, has a stable public deployment, and is notable for its breadth of data integration, multi-platform reach, and first-class agent/MCP tooling — positioning it as a trending open-source geopolitical intelligence platform.
AppsMCPAgentslocal-firstmultimodalworkflow-automation
★ 74,912 total+12,862 this week
#16
Voicebox is a free, open-source, local-first AI voice studio created by Jamie Pine that positions itself as a unified alternative to both ElevenLabs (voice output) and WisprFlow (voice input), combining the full voice I/O stack into a single desktop application that runs entirely on the user's machine. The project's core philosophy is complete privacy — no voice data, model weights, or audio captures ever leave the local device, making it a compelling option for privacy-conscious developers, creators, and AI power users. At its heart, Voicebox supports seven distinct TTS engines — Qwen3-TTS, Qwen CustomVoice, LuxTTS, Chatterbox Multilingual, Chatterbox Turbo, HumeAI TADA, and Kokoro — each with different strengths in terms of language coverage, speed, expressiveness, and VRAM requirements, switchable per generation. Voice cloning is zero-shot, requiring only a short reference audio sample, and the app also ships with over 50 curated preset voices across Kokoro and Qwen CustomVoice for users who don't have reference audio. The platform supports 23 languages spanning English, Arabic, Japanese, Hindi, Swahili, and many more, with Chatterbox Multilingual offering the broadest coverage. Expressive speech is supported through paralinguistic tags like `[laugh]`, `[sigh]`, and `[gasp]` via Chatterbox Turbo, while Qwen CustomVoice enables natural-language delivery control such as "speak slowly" or "whisper." Post-processing is handled by eight audio effects powered by Spotify's Pedalboard library, including pitch shift, reverb, delay, chorus, compression, and filters, with built-in and custom presets assignable per voice profile. A Stories Editor provides a multi-track timeline for producing conversations, podcasts, and narratives with drag-and-drop composition, inline trimming, and synchronized playback. On the input side, Voicebox offers global dictation via a configurable hotkey with push-to-talk and toggle modes, accessibility-verified auto-paste on macOS, and optional LLM-based cleanup of transcripts using a bundled local language model. Transcription is powered by OpenAI Whisper, running on MLX for Apple Silicon or PyTorch for CUDA, ROCm, and DirectML backends. A standout feature is agent voice integration: any MCP-aware AI agent such as Claude Code, Cursor, or Cline can invoke a single tool call (`voicebox.speak`) to speak back to the user in a cloned voice, with support for voice personas and LLM-driven compose, rewrite, and respond modes over MCP. The app is built with Tauri (Rust) rather than Electron for native performance, and runs on macOS (MLX/Metal), Windows (CUDA), Linux, AMD ROCm, Intel Arc, and Docker. It exposes a REST API and
AppsMCPModelsaudiolocal-firstmultimodal
★ 46,944 total+3,417 this week
#18
AI Job Search is an open-source, locally-run job application automation framework built on top of Anthropic's Claude Code CLI, created by Mads Lorentzen, a geophysicist-turned-AI-engineer who used the exact same system to conduct his own job search after being laid off in late 2025. The project is notably grounded in real-world validation: 69 tailored applications, 20 first interviews, and one signed contract resulted in Lorentzen starting as an AI engineer in June 2026, giving the framework a credibility that purely theoretical tools lack. The core idea is to fork the repository, populate a personal profile, and then let Claude Code handle the heavy lifting of evaluating job postings, tailoring CVs, writing cover letters, and preparing for interviews — all running locally on the user's machine without cloud dependencies beyond the Claude Code CLI itself. The workflow is organized around three primary commands: `/scrape` searches multiple job portals and deduplicates results sorted by fit, `/apply` takes a job URL or pasted description and runs a full drafter-reviewer pipeline producing LaTeX-compiled CVs and cover letters, and `/setup` builds the user's profile from uploaded documents, a pasted CV, or a guided interview. A second-agent reviewer critiques drafts before final output, encoding career guidance best practices such as structured fit evaluation across five dimensions, forward-looking cover letter framing, and optional salary benchmarking. The framework is language- and country-agnostic in its core logic, though its job portal search skills are currently built for the Danish market, covering platforms like Jobindex, Jobnet, Akademikernes Jobbank, JobDanmark, LinkedIn, and FreeHire, with a clear extension pattern for adding portals in other markets via the `/add-portal` command. Beyond the core trio, ten additional commands extend the system: `/interview` builds stage-specific prep packs with STAR-mapped questions and mock interview roleplay; `/outcome` tracks application results, archives submitted materials, and drafts follow-up messages; `/rank` batch-scores scraped postings in parallel before committing to full applications; `/expand` enriches profiles by scanning linked public sources like GitHub, Kaggle, and Google Scholar; `/upskill` performs gap analysis between the user's profile and target postings with web-searched learning resources; `/html-report` generates a fully offline, self-contained HTML dashboard with funnel charts and application tables; `/notion-sync` publishes a live read-only pipeline view to Notion via MCP; `/gmail-sync` reads Gmail for application status signals and proposes tracker updates for user approval; and `/add-template` registers custom CV or cover letter templates in LaTeX or Typst. Security is addressed explicitly: job postings are treated as untrusted input, the workflow follows no embedded instructions from postings, and a dedicated SECURITY.md file outlines the agentic defense model while acknowledging it is instruction-level rather than a full sandbox. Prerequisites include Python 3.10+, Bun for the CLI search tools, and a LaTeX distribution supporting lualatex and xelatex, with optional poppler for ATS parseability checks. The project is actively trending on Trendshift, has CI badges indicating maintained quality, and explicitly disclaims any affiliation with Anthropic or any cryptocurrency scheme, signaling a community-first, transparent
AppsAgentsToolslocal-firstworkflow-automationsynthetic-data
★ 27,348 total+3,237 this week
#36NEWPartial week
X4G is a fast, modern gateway project for VLESS tunneling over WebSocket and XHTTP protocols, combined with a built-in HTTP Proxy, designed primarily for censorship circumvention and traffic management. Written in Persian, it targets users in Iran and similar regions who need flexible, configurable proxy infrastructure. Key features include a full web management dashboard with hourly traffic charts, live connections, and activity logs; unlimited link management with per-config traffic quotas (KB/MB/GB), bandwidth throttling (Mbps), simultaneous IP limits, and auto-expiry by day. It supports QR code output, configurable uTLS fingerprints (Chrome, Firefox, Safari, etc.), custom ALPN, and arbitrary connection ports. A standout feature is "Sub Groups" — professional public subscription pages grouping multiple configs with optional password protection. An optional Telegram management bot enables full config and sub-group administration via a step-by-step wizard without opening the web panel. Deployment targets Railway.app via GitHub fork, with persistent state stored in JSON on a mounted Volume. The project appears early-stage but feature-rich, likely trending due to growing demand for advanced anti-censorship tooling in Persian-speaking communities.
AppsMCPToolssecurityworkflow-automation
★ 6,821 total+2,082 this week
#37
Open Design is an open-source, agent-native design platform positioned explicitly as an alternative to Anthropic's Claude Design and to Figma for the AI agent era. It is a local-first native desktop application for macOS and Windows that transforms any coding agent CLI already installed on a developer's laptop into a full-featured design engine. Rather than requiring users to push pixels on a canvas, Open Design delivers real, runnable single-page artifacts written in actual CSS, real fonts, and real components, all shaped by a team's `DESIGN.md` brand contract and exported directly to HTML, PDF, PPTX, or MP4. The core philosophical premise is that the agent-native loop Anthropic shipped with Claude Design — discover the brief, lock the direction, stream the artifact, critique, deliver — should be open, composable, and filesystem-readable rather than locked inside a proprietary product.
The architecture is built around three primary distribution mechanisms: a filesystem of composable skills, a CLI (`od`), and an MCP (Model Context Protocol) server. Once Open Design is installed, a single command such as `od mcp install claude` wires the MCP server into a supported agent's configuration, enabling the same design tools to be called from inside any of the 25+ supported coding agent CLIs. Supported agents include Claude Code, OpenAI Codex CLI, Cursor, VS Code with GitHub Copilot, GitHub Copilot CLI, OpenCode, OpenClaw, Antigravity, Cline, Trae, Kimi CLI, Kiro, Pi Agent, Mistral Vibe CLI, Hermes Agent, DeepSeek Reasonix, Raven, and more. This breadth of agent compatibility is a defining feature and a clear signal of the project's ambition to become the universal design layer across the fragmented AI coding agent ecosystem.
The platform is organized around several core pages and concepts. The Home page serves as the entry point where users pick a skill, select a design system, type a brief, and initiate generation. The Automation page allows repetitive design workflows to be orchestrated into reusable, schedulable automations. The Design System page is where teams distill their `DESIGN.md` into a brand contract that shapes every output the agents produce. The Plugin page enables browsing, installing, and distributing workflow plugins to extend generation capabilities on demand. The Integrations page connects external systems and MCP tools, allowing Open Design to be invoked from any IDE, script, or automation pipeline.
Inside any project's Studio, the same design system streams out multiple artifact types. Prototypes are single-page HTML artifacts rendered in a sandboxed iframe, previewable instantly and downloadable as source. HyperFrames are programmatic motion graphics and animated outputs rendered to real MP4 files at resolutions such as 1920×1080 at 30fps. Decks are pitch presentations navigable by keyboard and exportable to PPTX or PDF. Images are brand-grade visual assets with high-resolution generation and download. This multi-artifact approach means a single design system definition can consistently power wildly different output formats without manual reformatting.
For model access, Open Design Cloud is the official managed service, offering 20+ flagship models including GPT, Claude, Gemini, and DeepSeek under a single recharge, billed by real token usage with zero configuration. For users who prefer to bring their own keys, a BYOK proxy endpoint supports OpenAI, Anthropic, Azure OpenAI, Google Gemini, Ollama, LM Studio, vLLM, Atlas Cloud, and any OpenAI-compatible endpoint. Atlas Cloud specifically uses `https://api.atlascloud.ai/v1` with OpenAI-compatible model IDs. Per-target SSRF protection blocks internal IPs, link-local addresses, and CGNAT ranges at the daemon edge, indicating meaningful attention to security in the proxy layer. Runtime definitions live in the daemon source under `apps/daemon/src/runtimes/defs/`, with an adapter contract documented in `docs/agent-adapters.md`.
The project is internationalized across 14 languages including English, Spanish, Portuguese, German, French, Simplified Chinese, Traditional Chinese, Korean, Japanese, Arabic, Russian, Ukrainian, Turkish, and Thai, suggesting a global community target. The README is structured with a product tour, platform compatibility matrix, demo screenshots, and quickstart instructions (described as "3 commands"), indicating a project that is actively maturing and investing in developer experience. The Open Design Fellow program is open for contributors who want to shape the product alongside the core team, with details in `MAINTAINERS.md` and a Discord community. The project is licensed under Apache 2.0, making it genuinely open for commercial and community use. Its combination of broad agent compatibility, open licensing, composable skill architecture, and direct challenge to both Claude Design and Figma makes it a notably ambitious and potentially trending project at the intersection of AI agents, developer tooling, and design automation.
AppsAgentsMCPlocal-firstmultimodalworkflow-automation
★ 81,788 total+2,003 this week
#40
Codex Dream Skin is an external theming/skinning tool for OpenAI's Codex desktop client that injects custom backgrounds and CSS via local CDP (Chrome DevTools Protocol) without modifying the official app binary or code signature. It works on both macOS and Windows, offering a native menu bar/system tray interface for switching, importing, and saving themes. The project features a companion platform, DreamSkin.cc, with a community theme gallery, an online Studio for designing themes, and a one-click skin-apply flow secured by SHA-256 verification, source validation, and native confirmation dialogs. Themes are packaged as ZIP files containing a background image, theme.json, and Safe CSS scoped to 12 registered UI components. Preset themes include the gothic sci-fi "Gothic Void Crusade" (contributed by community member seansong-ideogram) and an "Arina Hashimoto" preset. The project is actively maintained with installer releases for both platforms, detailed security boundaries, and contribution templates. It is trending likely due to growing interest in AI coding tools and desktop customization, combined with its polished community ecosystem and zero-modification safety approach.
AppsMCPToolscodinglocal-firstsecurity
★ 12,382 total+1,826 this week
#42NEWPartial week
Bento is an open-source, self-contained PowerPoint alternative where the entire application — viewer, editor, and presenter — lives inside a single ~560 KB HTML file. Users can open, edit, and present decks in any modern browser with no installation, account, or cloud dependency required. Key features include morph-based slide transitions, end-to-end encrypted (AES-GCM) live collaboration via a custom CRDT engine, built-in dependency-free charting, and a local-first offline mode. The document format is plain JSON embedded in the file, making it natively editable by AI agents (Claude Code, Cursor, Aider, Ollama, etc.) without plugins. The file rewrites itself on save using the File System Access API, and signed in-app updates never touch user documents. A roadmap extends the "bento" format to notes (bento/spaces), sheets (bento/dash), and a vault app. Released under the MIT License, the project is notable for its radical portability, privacy-first architecture, and AI-agent-friendly design, positioning it as a trending alternative to cloud-locked office suites.
AppsToolsMCPlocal-firstmultimodalworkflow-automation
★ 2,337 total+1,721 this week
#45
Palmier Pro is an open-source, Swift-native video editor for macOS built specifically for AI-assisted workflows, positioning Adobe Premiere Pro as its north star. Developed by Y Combinator S24 company Palmier, it enables users and AI agents to collaboratively generate and edit videos directly within a timeline interface. It integrates state-of-the-art generative AI models (Seedance, Kling, Nano Banana Pro) for video and image generation, and exposes a built-in MCP (Model Context Protocol) server at localhost, allowing seamless connection with AI coding agents like Claude Code, Claude Desktop, Codex, and Cursor. The core editor and MCP server are fully open source under GPLv3, while generative AI processing remains proprietary and requires a subscription. Currently limited to macOS 26 (Tahoe) on Apple Silicon, the project is notably trending on Trendshift, has multilingual README support across 15 languages, and signals strong momentum as an agentic AI-video editing platform at the intersection of open-source tooling and frontier AI model integration.
AppsMCPAgentsvideomultimodalon-device
★ 12,464 total+1,632 this week
#46
Astryx is Meta's open-source design system, now publicly released after eight years of internal development where it powered 13,000+ apps across the company. Currently in beta, it ships 150+ accessible React components, brand-level theming, dark mode support, ready-to-ship templates, and a CLI as one cohesive package. Built on React and StyleX, it notably avoids styling lock-in — consumers can override with Tailwind, CSS modules, or plain CSS without adopting StyleX themselves. A standout feature is its "swizzle" mechanism, which ejects a component's full source into your project for deep customization without forking. Astryx is explicitly designed for human-AI collaborative development, with its API, docs, and CLI co-designed so developers and AI assistants work from the same reference. It offers seven pre-built themes and a component lab, plus experimental chart packages. Its trending appeal likely stems from Meta open-sourcing a mature, large-scale internal design system with first-class AI-agent ergonomics — a timely differentiator in the current AI-assisted development landscape.
AppsToolscoding
★ 10,804 total+1,603 this week
#62
**daily_stock_analysis** (股票智能分析系统) is an AI-powered daily stock analysis platform supporting A-shares, Hong Kong, US, Japanese, Korean, and Taiwan markets. It automatically generates structured "Decision Dashboard" reports covering core conclusions, scoring, trend direction, buy/sell levels, risk alerts, catalysts, and operational checklists, then pushes them to WeChat Work, Feishu, Telegram, Discord, Slack, or email. The system integrates multiple LLM backends (Gemini, OpenAI, DeepSeek, Claude, Qwen, Ollama) and aggregates market data from AkShare, Tushare, YFinance, Baostock, TickFlow, and Longbridge, with news sourced from SerpAPI, Tavily, Bocha, Brave, and MiniMax. It features a FastAPI-based web workbench with manual analysis, backtesting, portfolio management, Agent strategy Q&A (supporting 15 built-in strategies including moving averages, Elliott Wave, and Chan Theory), and smart import via image/CSV/clipboard. Deployment is zero-cost via GitHub Actions (5-minute setup), Docker, or local scheduling. Ranked **#1 Python Repository of the Day** on Trendshift and featured on HelloGitHub, it is actively trending in the Chinese quant/retail investing community. Two companion projects, AlphaSift and AlphaEvo, extend it with multi-factor stock screening and strategy backtesting/evolution.
AppsAgentsToolsworkflow-automationmultimodal
★ 59,120 total+1,186 this week
#65
OmniGet is a free, open-source, cross-platform desktop application for Windows, macOS, and Linux that consolidates downloading, media playback, reading, and music management into a single no-terminal interface. Built with Tauri, Svelte, and Rust, it targets users who are tired of juggling multiple tools — yt-dlp in a terminal, separate course downloaders, and standalone music apps — by unifying all of that into one window. Its core downloading engine wraps yt-dlp and FFmpeg, which self-install and auto-update with SHA256 verification, giving it access to over 1,800 supported sites including YouTube, TikTok, Instagram, Twitter/X, Reddit, Twitch, Vimeo, Bluesky, and Bilibili. Bilibili support is notably deep, covering 4K, HDR, Dolby Vision, Hi-Res lossless audio, Dolby Atmos, Danmaku in XML/ASS/JSON, NFO metadata for Kodi and Jellyfin, and eleven distinct URL types. Online course downloading is a headline feature, supporting Udemy, Hotmart, Kiwify, Skool, Teachable, Kajabi, Wondrium, Thinkific, and Rocketseat, with an integrated player that resumes playback, supports timestamped notes, and displays attached PDFs side by side. A global hotkey (Ctrl+Shift+D / Cmd+Shift+D) lets users copy any link and trigger a background download without ever opening the app window. The built-in music player handles local libraries (MP3, FLAC, M4A, OGG, Opus), synced lyrics, an equalizer, and integrations with Spotify, SoundCloud, YouTube Music, Qobuz, and Last.fm. A built-in reader supports EPUB, PDF, CBZ, TXT, and HTML with highlights, bookmarks, and focus mode. Additional features include batch downloading, partial video clips, subtitle embedding, Whisper-generated subtitles, SponsorBlock, channel following with tray notifications, and direct P2P file transfer via short code. The project is licensed under GPL-3.0, requires no admin rights, and earned thousands of GitHub stars within its first months — notable because no comparable open-source all-in-one tool
AppsToolsDatalocal-firstvideoaudio
★ 7,889 total+1,101 this week
#71
Toonflow is an AI-powered short-drama production workbench developed by HBAI-Ltd, designed to automate the full creative pipeline from raw text to finished video. The platform targets the rapidly growing short-video and micro-drama content market, positioning itself as a zero-barrier, end-to-end AI factory for turning novels or scripts into polished video episodes — claiming a 10x improvement in production efficiency. Its core workflow follows a structured loop: planning → scriptwriting → storyboarding → video output, all organized on an infinite-canvas interface where script nodes, character assets, storyboard frames, and video clips can be freely arranged, branched, and iterated in parallel rather than linearly. A standout architectural feature is its three-layer Agent collaboration system comprising decision, execution, and supervision layers, which together handle task decomposition, content generation, quality review, and revision feedback to improve consistency across long productions. Toonflow also implements persistent Agent memory using local ONNX-based vector retrieval, supporting short-term message history, long-term summarization, and semantic recall across sessions — a technically sophisticated approach to maintaining narrative continuity. The platform features a programmable vendor system allowing users to write TypeScript logic for AI provider integrations directly in the settings panel, taking effect instantly without restarting or modifying source code, making it highly adaptable for private deployments and multi-model setups. Chapter event graph-driven adaptation automatically extracts and structures plot events from source novels, enabling precise context retrieval during script adaptation to reduce information loss in long texts. Skill files externalize core prompts for ScriptAgent and ProductionAgent as editable Markdown files, supporting rapid prompt tuning without code changes. A demo production was completed in approximately two hours using Seedance 2.0 for video, GPT Image 2 for images, and Claude Opus 4.6 for language tasks, at a total cost of roughly ¥130 (~$18 USD). The project supports seven languages including Simplified Chinese, Traditional Chinese, English, Thai, Vietnamese, Japanese, and Russian, and is deployable via desktop installers (Windows, macOS, Linux), Docker, cloud servers with PM2, or through an officially authorized cloud platform called 智星云 (AI-Galaxy). Built on TypeScript, Node.js, Electron, and Docker, the project has received AtomGit G-Star No. 540 certification and Gitee GVP recognition, signaling meaningful community traction. It is licensed under Apache 2.0 and maintains an active Discord community, suggesting it is an emerging, actively developed open-source project at the intersection of generative AI, agentic workflows, and creative content production — a notably trending combination in 2024–2025.
AppsAgentsDatavideomultimodalworkflow-automation
★ 12,704 total+1,032 this week
#81
T3 Code is a minimal web-based GUI designed to serve as a unified front-end for AI coding agents, currently supporting Codex, Claude Code, Cursor, and OpenCode, with more providers planned. The project allows developers to interact with these agents through a single interface rather than switching between separate tools. It can be run instantly via `npx t3@latest` without installation, or installed as a desktop application on Windows (via winget), macOS (via Homebrew), or Arch Linux (via AUR). The build system relies on Vite+ (vp), a custom toolchain wrapper around Vite. The project is explicitly in a very early stage — bugs are expected, public documentation is minimal, and contributions are not yet formally accepted, though a Discord community exists for support. Its notable appeal lies in aggregating multiple leading AI coding assistants into one lightweight GUI, which aligns well with the current surge of interest in agentic coding workflows. The combination of multi-provider support, cross-platform distribution, and zero-install usage makes it a compelling early-stage tool in the AI developer tooling space.
AppsAgentsToolscodingworkflow-automation
★ 15,112 total+922 this week
#83
Awesome Zhuiju Free is a curated, community-maintained awesome-list repository focused on providing free, ad-free resources for Chinese-language drama and film streaming (追剧, meaning "binge-watching dramas"). The project is authored by GitHub user laoma2053 and is hosted at the companion website zhuiju.me. Its core mission is to aggregate and verify high-quality, freely accessible video streaming resources for Chinese-speaking audiences, covering a wide spectrum of content discovery methods and tools. As of the latest badge update, the repository has catalogued 82 resources across eight major categories: online streaming sites (30 entries), mobile video apps (3), cloud drive search tools (4), magnet/BitTorrent resources (11), subtitle resources (3), TVBox/影视仓 shell apps (2), TVBox/影视仓 configuration endpoints (15), IPTV live stream sources (1), membership group-buying services (0 currently listed), and open-source projects (13). A defining feature of the project is its automated daily availability checking system, powered by GitHub Actions workflows. Every day around 09:00 Beijing time, the system probes each listed resource and updates its status to one of four states: accessible (🟢), restricted (🟡), inaccessible (🔴), or unchecked (⚪). Results are stored in a machine-readable reports/availability.json file, and resource metadata is maintained in resources/resources.json, making the entire dataset open and programmatically accessible. This automation distinguishes it sharply from typical navigation sites, which often go unmaintained and accumulate dead links. The project also runs a data validation workflow (validate-data.yml) to ensure structural integrity of the resource data. Community participation is a central pillar: users can submit new resources or report broken links directly via GitHub Issues using structured templates, lowering the barrier for contribution. All resources are hand-curated with a stated emphasis on being free and ad-free, and each entry carries a star-based recommendation rating. The resource list spans both domestic Chinese streaming sites and overseas-accessible platforms, with many entries noting whether a VPN or proxy (梯子) is required for access. Notable included platforms span a wide range of quality tiers, from true 4K sources like IFN and 好好看, to popular aggregators like 爱看机器人, 泥视频, PPnix, and 多瑙影视, to anime-focused sites like 青空次元. The project is licensed under Creative Commons Attribution 4.0 (CC BY 4.0), ensuring the data remains openly reusable. It has gained enough traction to appear on Trendshift's daily trending repositories chart for JavaScript, signaling active community interest and growth momentum. The repository's transparency — open data, open workflows, open contribution — positions it as a notably trustworthy alternative to opaque, ad-laden video navigation sites that dominate this niche. Its combination of automation, community curation, and broad categorical coverage makes it a practical and living reference for Chinese-speaking users seeking free, reliable drama and film streaming options across multiple access methods and device types.
AppsDataMCPvideoworkflow-automation
★ 4,661 total+874 this week
#85
Codex++ is an external launcher and management tool for the OpenAI Codex/ChatGPT desktop application, built with Rust 1.85+ and Tauri 2.x. Rather than patching the official app's `app.asar` or writing files to the installation directory, it operates non-invasively via the Chromium DevTools Protocol and a local auxiliary service. Its core capability is provider management, supporting four distinct modes: official login only, official login mixed with a custom API, pure API mode, and an aggregated provider mode that enables failover, per-session rotation, per-request rotation, and weighted routing across multiple API backends. Each provider can be configured with Responses or Chat Completions protocol, and the tool transparently converts Chat Completions to the Responses protocol Codex expects via a local proxy. Beyond provider switching, Codex++ delivers rich UI enhancements including session deletion and batch management, Markdown export, plugin marketplace unlocking, model whitelisting, rich-text-to-plain-text paste fixing, forced Chinese localization, session width/scroll restoration, Stepwise next-step suggestions, Upstream worktree creation, Zed Remote project integration, and user script support. All enhancements can be individually toggled. The project ships signed installers for Windows (NSIS x64), macOS Intel (x64 DMG), and macOS Apple Silicon (arm64 DMG), with in-app update checking. It has attracted multiple commercial sponsors offering API relay services for Chinese developers, suggesting active community traction. The project maintains a QQ group and Telegram channel for community support, indicating a growing Chinese developer user base and notable momentum on GitHub trending.
AppsToolsMCPcodinglocal-firstworkflow-automation
★ 26,716 total+862 this week
#95
OpenLogi is a native, local-first, open-source alternative to Logitech Options+, written in Rust, that communicates with Logitech HID++ peripherals over Logi Bolt, Unifying receivers, Bluetooth, and USB — without requiring a Logitech account or telemetry. It supports full button remapping, DPI control and presets, SmartShift wheel configuration, per-device scroll inversion, RGB keyboard lighting, and per-application profile switching, all stored in a plain TOML config file. The project is structured into three components: a GPUI-based desktop GUI with an interactive mouse diagram and clickable hotspots, a background agent handling all device I/O via HID++, and a CLI for headless inventory and diagnostics including DPI/SmartShift round-trip testing. It runs on macOS, Linux, and Windows, with Linux treated as a first-class platform via evdev/uinput, udev rules, systemd units, and native packages (.deb/.rpm/.pkg.tar.zst). The project is under active development but already feature-rich, trending on Trendshift for Rust repositories, and notable for being the only cross-platform, privacy-respecting, scriptable replacement for Options+ that supports Linux and gesture-button reassignment. It is dual-licensed under MIT and Apache 2.0.
AppsToolslocal-firston-devicesecurity
★ 7,626 total+663 this week
#114
GeoLibre is a free, open-source, cloud-native GIS platform designed for visualizing, exploring, and analyzing geospatial data across web browsers, desktop (Windows, macOS, Linux), Android, and Jupyter notebooks — all while keeping data local and private. Built with Tauri v2, React, TypeScript, MapLibre GL JS, DuckDB-WASM Spatial, and deck.gl, it offers a unified workspace that adapts responsively across environments. Key features include 3D Tiles rendering, temporal animation via a Time Slider, 700+ geoprocessing tools, a SQL Workspace powered by DuckDB, a plugin API, and remarkable planetary basemap support covering the Moon, Mars, Mercury, Venus, Galilean moons, Titan, Pluto, and Charon with per-body ellipsoid measurements. The project is available via PyPI, conda-forge, Microsoft Store, AUR, and Flatpak, indicating strong packaging maturity. It is notable for its breadth — spanning data science workflows in Jupyter, cloud-native geospatial formats, spatial SQL, and even extraterrestrial mapping — making it a compelling open-source alternative to proprietary GIS tools. Archived on Zenodo with a citable DOI, it appears to be a rapidly maturing, actively developed platform with growing community traction.
AppsDataToolslocal-firstmultimodalvector-database
★ 2,308 total+529 this week
#120NEWPartial week
Vorssaint is a free, open-source macOS menu bar utility that consolidates the functionality of a dozen typically paid Mac apps into a single icon, requiring no account, no telemetry, and no subscription. The project is built in Swift and targets macOS 14 and later on Apple Silicon, positioning itself as a privacy-first, locally-executed alternative to fragmented paid utilities. Its core philosophy is modularity: users install only the features they want through a Features hub, and uninstalled features stop loading entirely, consuming zero CPU, memory, or energy until reinstalled — with settings preserved throughout. The app is organized into broad capability domains covering sound (per-app volume mixing, output switching, microphone management), system monitoring (CPU, GPU, memory, temperatures, battery health, network), window and Dock management (a richer ⌘Tab app switcher with live thumbnails, window snapping, Dock previews), keyboard and mouse enhancements (text snippets, smooth scrolling, Super Key via Caps Lock, key debounce), clipboard and file tools (clipboard history, a file shelf, Finder cut-and-paste, URL cleaning), and everyday utilities (screenshot editor with annotation and QR recognition, offline OCR, color picker, Homebrew manager, app uninstaller, media compression, scratchpad, radial menu, and more). Installation is available via Homebrew or a signed and notarized disk image from the GitHub releases page, ensuring smooth macOS gatekeeper compatibility. The app supports thirteen languages, offers settings export/import for multi-Mac setups, and includes three one-click onboarding bundles (Essentials, Windows, Battery and Quiet). Notably, Vorssaint was ranked the #1 Swift repository of the day on Trendshift, signaling strong community momentum. Its breadth of features, zero-cost model, strict privacy stance, and thoughtful energy-aware architecture make it a compelling and trending alternative to the fragmented Mac utility app market.
AppsToolslocal-firston-device
★ 3,294 total+517 this week