Evaluations

Open-source projects, reviewed the honest way — including who should skip them.

AI Tools★ 4.7k · PythonKiroCrewKiro Crew is an open source development workspace that runs on hardware you control. Its README describes persistent sessions, memory and scheduled work, accessible through a desktop app, web dashboard, CLI and connected messaging services. Our run checked setup steps, not persistence or unattended task completion.SetupDocsCommunityMaturityVerdict inside AI Tools★ 4.9k · JavaScriptOrkasOrkas is a desktop application whose Commander agent coordinates specialists and installed CLI agents such as Claude Code and Codex. Its README describes projects that retain goals, rules, memory and files between conversations. Our sandbox did not verify completed agent work or project continuity.SetupDocsCommunityMaturityVerdict inside AI Tools★ 27k · TypeScripthypitHypit gives coding agents such as Claude Code and Codex a workflow for creating videos with captions, footage and effects tied to spoken words. It keeps the composition editable so you can change a presenter, script or product and reuse suitable existing material. Its published license adds conditions to Apache 2.0, including restrictions on hosted services and commercial redistribution.SetupDocsCommunityMaturityVerdict inside AI Tools★ 16k · TypeScriptParadevParadev is a desktop workspace for CLI coding agents, terminals, code review and browser previews. It gives tasks separate Git worktrees. Its source is available under Elastic License 2.0, which restricts providing substantial functionality as a hosted or managed service to third parties.SetupDocsCommunityMaturityVerdict inside AI Tools★ 5.2k · Pythonsee-throughSee-through is a research pipeline that separates a single anime character illustration into editable layers and exports a layered PSD. It uses layer generation and depth estimation to help recover hidden parts and infer drawing order. The output is a starting point for an artist's work, with rigging explicitly outside the project's scope.SetupDocsCommunityMaturityVerdict inside AI Tools★ 29k · PythonResume-MatcherResume Matcher is a self-hosted resume editor that tailors a master resume to individual job descriptions. You can review suggested wording, check the resulting document and export a PDF, using a remote model, a local model or a coding-agent workflow.SetupDocsCommunityMaturityVerdict inside AI Tools★ 4.2k · TypeScriptopenbotOpenBot is a desktop workspace for persistent AI agents, with separate conversations and workspaces, queues, file transfers and an embedded browser. Its current README calls it a development preview and warns that agents operate with broad host access without per-action confirmations after first-launch consent.SetupDocsCommunityMaturityVerdict inside AI Tools★ 11k · TypeScriptinkosInkOS is an AI writing system for fiction and other story projects, with a Chinese primary README and separate English documentation. It documents chapter generation, persistent story state and revision through a browser workbench or terminal commands. English creation is documented, but our run stopped during dependency installation before any writing could be evaluated.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.8k · PythonskillsHiggsfield AI Skills gives coding agents Markdown instructions for creating media and brand assets through Higgsfield's CLI and service. It works with agents including Claude Code and Codex, and groups tasks such as generation, product photography and brandbooks into separate workflows. You get instructions and supporting tools for an authenticated service workflow, not a local model engine.SetupDocsCommunityMaturityVerdict inside AI Tools★ 5.1k · Pythonobsidian-second-brainObsidian Second Brain gives AI agents workflows for saving, finding and maintaining linked Markdown notes in an Obsidian vault. The supplied README describes agent commands for capture, retrieval, ingestion and note revision; those descriptions are not evidence that the workflows succeeded in our sandbox.SetupDocsCommunityMaturityVerdict inside AI Tools★ 9.4k ▲317/7d · TypeScriptinfinite-canvasInfinite Canvas is an image-creation workspace that combines a visual canvas, AI generation, reference-image editing, prompts and assets. A local Canvas Agent connects the canvas with coding assistants through MCP. The supplied main and Canvas Agent README extracts are primarily Chinese; an English operating guide was not assessed.SetupDocsCommunityMaturityVerdict inside AI Tools★ 41k ▲5/7d · PythonColossalAIColossal-AI supplies components for distributing model training and inference across accelerators. It helps engineers divide computation and memory use when a model or training job outgrows a simpler PyTorch setup.SetupDocsCommunityMaturityVerdict inside AI Tools★ 11k ▲562/7d · TypeScriptgenofficeGenOffice is a desktop suite for editing Word documents, spreadsheets, presentations and PDFs, with AI editing alongside the file. It also handles Markdown and HTML, and supplies a command line and agent integrations for working on Office files without manually opening each editor.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.2k ▲650/7d · RusthexHEX is a local voice dictation app from anomalyco. Its main Mac release targets Apple silicon and macOS 15 or newer, with separate Windows and Linux betas. It transcribes speech locally and pastes text into the active app. Optional OpenCode rewriting sends text and application context to a configured model provider.SetupDocsCommunityMaturityVerdict inside AI Tools★ 6.4k ▲198/7d · Pythonopen-duetOpen Duet is a workspace for named AI agents with their own instructions, conversations, memory and assigned folders. It lets one operator organize delegated work and recurring prompts while bringing a separately configured model engine.SetupDocsCommunityMaturityVerdict inside AI Tools★ 10k ▲115/7d · Pythondrawio-skillDrawio Skill is an agent skill that turns descriptions, code, and infrastructure definitions into editable draw.io diagrams. It helps a coding assistant maintain the diagram after the first draft, including source synchronization and different views of the same architecture model.SetupDocsCommunityMaturityVerdict inside AI Tools★ 16k ▲38/7d · Pythondocker-androidDocker-Android runs Android emulators in Docker containers so you can see their screens in a browser and control them remotely. It gives developers a reusable Android workspace for app debugging and automation, with separate MCP and sponsor-only AI agent options.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.3k ▲217/7d · JavaScriptCozyClayCozyClay is a browser-based studio for arranging a 3D scene, posing characters and setting camera shots before handing references to an AI video model. It exports frames, reference clips and prompts, and lets an AI assistant control the scene through an in-app panel, MCP or a terminal.SetupDocsCommunityMaturityVerdict inside AI Tools★ 8.0k ▲64/7d · Pythonseedance-2.0Seedance 2.0 Skill OS is an agent skill for planning shots, binding references and continuing from an accepted clip. Your provider generates the video and handles its costs. The supplied README identifies v6.7.0; our sandbox measurements concern commit 4668457.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.2k ▲64/7d · RustBeavBeav is an AI-assisted workspace for collecting source material, drafting social content, and preparing media and publishing packages. The project is primarily Chinese-language, with an English README available. Its public code is a 2.5.0 snapshot, while current desktop installers are maintained separately.SetupDocsCommunityMaturityVerdict inside AI Tools★ 4.9k ▲64/7d · TypeScriptPilotDeckPilotDeck is an AI assistant that organizes ongoing work into projects with their own files, memory, and skills. It provides English and Chinese documentation and lets you inspect and edit remembered information. It targets recurring work that needs more continuity than a single conversation.SetupDocsCommunityMaturityVerdict inside AI Tools★ 6.4k ▲591/7d · TypeScriptrakazoRakazo is a self-hosted AI assistant with persistent bots, conversations, memory, and computer workspaces. Web, Electron, and mobile clients connect to the same backend. Bots can use a browser, terminal, and files with your chosen model and computer provider. The README labels Rakazo beta.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.0k ▲132/7d · TypeScriptBeefTVBeefTV is a desktop video-creation workspace that connects prompts, reference material, and generated media on a freeform canvas. Its README and linked quick-start guide are Chinese, with no English edition linked there. It keeps a local workspace while letting you configure the model services that generate content.SetupDocsCommunityMaturityVerdict inside AI Tools★ 6.2k ▲233/7d · TypeScriptOpenDotsOpenDots is a customizable workspace for persistent AI assistants, with editable documents, chat, optional computer access, and connections to external tools. It helps developers build a single-owner assistant application, while conversations depend on CopilotKit Intelligence and a configured model provider.SetupDocsCommunityMaturityVerdict inside AI Tools★ 291 ▲108/7d · JavaScriptreelmimicReelMimic is a local web app that studies a reference video's pacing and visual style, then coordinates coding agents to make a new animated video from your brief. You approve a storyboard before production and can request changes at specific moments in the finished video.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.1k ▲37/7d · PythoneasyreadEasyRead is a local paper reader that translates PDFs, keeps the original page beside the translation, and attaches notes and AI discussion to passages. Its main documentation is Chinese, with an English README and English interface available. You can also read the original without translating it.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.3k ▲77/7d · Pythonmotion-video-kitMotion Video Kit is a Claude Code skill with instructions, reference notes, and helper scripts for making business commercials. It helps organize storyboards, code-built animation, audio, and independent critique, but you still supply the renderer, assets, and judgment.SetupDocsCommunityMaturityVerdict inside AI Tools★ 540 ▲8/7d · PythonlivecanvasLiveCanvas is a Chinese-first agent skill that turns a brief into an animated visual and Apple Live Photo resources; an English README exists, while detailed production guides remain in Chinese. It guides a coding assistant through research, animation, and packaging so you can deliver an editable project alongside the preview.SetupDocsCommunityMaturityVerdict inside AI Tools★ 21k ▲38/7d · TypeScriptvoyagerVoyager is a browser extension that adds folders, reusable prompts, message timelines, and other controls to AI chat websites. It helps you find and reuse work inside Gemini, AI Studio, ChatGPT, Claude, and DeepSeek, with different features available on each site.SetupDocsCommunityMaturityVerdict inside AI Tools★ 7.6k ▲30/7d · TypeScriptsuperdesignSuperdesign is the original IDE extension for generating and previewing interface drafts beside your code. Its README says this version is no longer actively maintained and directs users to a web app and a separate coding-agent skill.SetupDocsCommunityMaturityVerdict inside AI Tools★ 515 ▲31/7d · PythonV3R1T4SV3R1T4S is a research kit that directs coding agents toward open problems and packages proposed answers with evidence for checking. It includes research objects, problem definitions, and a review process that separates promising claims from confirmed submissions.SetupDocsCommunityMaturityVerdict inside AI Tools★ 923 ▲20/7d · JavaScriptsuperdesign-skillSuperdesign Skill gives coding agents a workflow for designing interfaces, presentations, and graphics through the Superdesign canvas. It helps an agent use the existing application as design context, but requires a separate CLI and a hosted-service login.SetupDocsCommunityMaturityVerdict inside AI Tools★ 9.2k ▲294/7d · Pythonuniversal-modderUniversal-modder gives coding agents game-modding instructions, a Python toolbox, and shared notes about specific games. It helps connect engine research, asset preparation, and in-game verification, with fal MCP support for generating art and sound.SetupDocsCommunityMaturityVerdict inside AI Tools★ 283 ▲11/7d · C++DeepEP-AscendDeepEP-Ascend moves data between Huawei Ascend chips during training and inference, especially when a mixture-of-experts model sends work to different experts. Its README opens in Chinese and provides an English introduction and detailed English operating instructions. It is a hardware-specific communication library for infrastructure engineers.SetupDocsCommunityMaturityVerdict inside AI Tools★ 623 ▲12/7d · C++DeepGEMM-AscendDeepGEMM-Ascend is a library of matrix-multiplication kernels for Huawei Ascend processors used in AI workloads. It gives model and kernel developers a familiar DeepGEMM interface while handling hardware-specific data layouts and execution details.SetupDocsCommunityMaturityVerdict inside AI Tools★ 527 · PythonhypoarenaHypoarena is an offline workbench for checking how hypothesis-generation pipelines handle evidence, debate, and ranking on synthetic material. The README is in English, while detailed documentation, including its limits and grounding rules, is largely in Chinese.SetupDocsCommunityMaturityVerdict inside AI Tools★ 432 ▲1/7d · PythonjeevesJeeves is a self-hosted model that answers yes/no, multiple-choice, and rating questions about text, returning probabilities your application can use. It can generate reasoning before choosing an answer and provides an API compatible with Jev-style decision workflows.SetupDocsCommunityMaturityVerdict inside AI Tools★ 26k ▲207/7d · Pythonfaster-whisperFaster-whisper is a Python library that turns audio into text using Whisper models on your own machine. It uses the CTranslate2 inference engine and gives an application control over CPU or NVIDIA GPU execution, batching, and timestamps.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.2k ▲20/7d · TypeScriptHeliosGenHeliosGen is a desktop canvas for connecting prompts, reference media, and image or video generators into reusable workflows. It keeps your project database and saved media on your computer, while its default generation service, kie.ai, receives jobs over the internet.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.2k ▲84/7d · TypeScriptanidoodleAnidoodle is a coding-agent skill and TypeScript engine for making drawn illustrations, animations, and interactive web art as editable code. It helps keep a picture and its moving versions in the same source instead of treating each export as a separate asset.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.6k ▲19/7d · PythonjeffJeff is a local model that takes a situation and a list of choices, then returns a probability for each choice. It helps applications route requests or classify text without generating and parsing a written answer.SetupDocsCommunityMaturityVerdict inside AI Tools★ 760 ▲25/7d · Offensive-Security-AI-ModelsOffensive-Security-AI-Models is an English-language reading list of open-weight models for authorized security research, with links and comparison tables. It helps you shortlist models by task, format, and tool support, but supplies no runnable application.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.3k ▲340/7d · awesome-claude-video-skillsAwesome Claude Video Skills is a directory of video-making skills and toolkits for Claude Code, Codex, and other coding agents, with English and Simplified Chinese READMEs. It groups projects by the videos they make and links to previews, source code, and README-based security grades.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.4k ▲26/7d · RustollayaOllaya runs local models that turn messages or other data into choices, scores, and probabilities without generating a written answer. It downloads models by name and serves them through a local API, including a format compatible with TypeSafe clients.SetupDocsCommunityMaturityVerdict inside AI Tools★ 5.4k ▲836/7d · HTMLlogo-design-skillLogo Design Skill gives Claude Code and other compatible agents a process for developing logos, from the brief to editable SVG artwork and delivery files. It combines design instructions, a searchable reference library, and Python helpers for inspecting and presenting concepts.SetupDocsCommunityMaturityVerdict inside AI Tools★ 13k ▲1.7k/7d · awesome-opus5-5-videosAwesome Opus 5.5 Videos is a curated collection of code-made animations, creator posts, and shared prompts, with links to live remakes. It helps you find a visual direction to try with a coding agent, though some entries contain only partial directions rather than a reproducible recipe.SetupDocsCommunityMaturityVerdict inside AI Tools★ 9.8k ▲2.4k/7d · JavaScriptlemo-opuscarLemo-Opuscar is a library of code-made short films, reusable style directions, and a Claude Code skill for creating your own. It gives a coding agent examples and production guides so a film can be drawn and assembled with local tools rather than a video-generation service.SetupDocsCommunityMaturityVerdict inside AI Tools★ 422 ▲1/7d · TypeScriptproduct-film-skillProduct Film is a Claude Code skill that turns your product's components and design rules into a scripted promotional video. It guides the brief, music edits, scene creation, and export so you can keep a product film alongside the application code.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.2k · PythonFuxiFuXi is a terminal AI coding assistant that reads a project, edits files, and runs commands using a model provider you choose or FuXi-managed models. English and Chinese documentation are available, but the agent's source is proprietary: the public repository contains documentation, installers, and an issue tracker.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.2k ▲25/7d · PythonjarvisJarvis is a desktop voice assistant that can use recent room conversation and saved personal memory to answer requests through local models. It also supports text chat, dictation, and connected tools. Its source is available under a noncommercial license, with a separate license required for commercial use.SetupDocsCommunityMaturityVerdict inside AI Tools★ 34k ▲98/7d · Pythonfish-speechFish Speech turns text and reference recordings into synthetic speech, including scenes with several characters. Its current Drama 3 model supports revising a performance and editing part of an audio recording, giving developers and studios control beyond reading text aloud.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.5k ▲290/7d · Pythononetakeonetake is a Claude Code skill for creating short animated product films in which objects carry the viewer into the next scene instead of disappearing between slides. Its documentation includes full English and Chinese guidance, with Python scripts and a JavaScript motion library supporting the workflow. It is source-available under a noncommercial license.SetupDocsCommunityMaturityVerdict inside AI Tools★ 852 ▲82/7d · Pythonawesome-claude-5-5-videosAwesome Claude 5.5 Videos, formerly awesome-opus-5-5-videos, collects source-linked animation and video examples alongside research data and scripts for maintaining the collection. It has an English overview and guides, though some original research and prompt resources are available only in Chinese.SetupDocsCommunityMaturityVerdict inside AI Tools★ 180 · Pythoninterference-searchInterference Search explores several possible solutions at once, merges repeated states, and uses a judge to decide which branches deserve more work. It is a research codebase for arithmetic puzzles and generated programs, useful for studying how search can reduce repeated reasoning.SetupDocsCommunityMaturityVerdict inside AI Tools★ 376 ▲5/7d · TypeScriptshipvideoShipvideo, called LaunchVideo in its README, turns a URL or text prompt into an animated launch video. A language model writes an HTML scene, then a browser renderer and ffmpeg produce an MP4 inside an OpenComputer session.SetupDocsCommunityMaturityVerdict inside AI Tools★ 236 · TypeScriptstorycastStorycast turns a topic and chosen character into a narrated animated film, coordinating writing, images, speech, music, and editing through fal. It gives you a browser interface for that workflow, using your own fal key rather than local generation.SetupDocsCommunityMaturityVerdict inside AI Tools★ 457 ▲15/7d · Rustomarchy-meeting-recorderOmarchy Meeting Recorder records your microphone and computer audio, then turns the meeting into an editable transcript with audio playback. It is built for the Omarchy Linux desktop and helps you revisit a call without inviting a recording bot.SetupDocsCommunityMaturityVerdict inside AI Tools★ 34k ▲174/7d · TypeScriptAionUiAionUi is a desktop workspace where AI agents read and change files, generate documents, and show their output beside the conversation. It includes a built-in agent and can bring tools such as Claude Code into one interface, with browser access and scheduled tasks for work away from the desktop.SetupDocsCommunityMaturityVerdict inside AI Tools★ 27k ▲65/7d · PythonMiniCPM-VMiniCPM-V is a family of models for asking questions about images, documents, and video on local devices or servers. The repository also contains MiniCPM-o for speech interaction, and provides English and Chinese documentation.SetupDocsCommunityMaturityVerdict inside AI Tools★ 30k ▲1.6k/7d · PythonUnlimited-OCRUnlimited-OCR is Baidu's document-parsing model project, with English instructions for turning images and PDF pages into parsed text. It includes examples for processing several pages together and a separate batch script that handles pages individually.SetupDocsCommunityMaturityVerdict inside AI Tools★ 410 ▲31/7d · TypeScriptbridgeclipBridgeClip is a desktop app that finds moments in long videos and exports captioned clips. It renders locally while sending audio and transcripts to AI services through your credited OpenRouter account, with optional social publishing through Zernio.SetupDocsCommunityMaturityVerdict inside AI Tools★ 602 ▲36/7d · Markdownai-engineering-courseAI Engineering Course is an English reading curriculum that links to lessons about building and operating AI systems. It gives developers an ordered path through model foundations, retrieval, agents, and deployment, with articles on Outcome School and videos for some topics.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.1k ▲16/7d · PythonCLMCLM scores a supplied set of answers or actions against the situation you describe, then ranks them or returns a typed decision. It is meant for choosing among existing candidates, such as support departments or proposed agent actions, rather than generating an unrestricted answer.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.3k ▲36/7d · GoFounderOSFounderOS is a business dashboard for a solo founder, with communication, client work, and AI agents in one operator console. It aims to reduce the effort of coordinating a small business across separate tools, while letting you control which connections and actions are enabled.SetupDocsCommunityMaturityVerdict inside AI Tools★ 566 ▲2/7d · Python3dicon3dicon is a Claude Code plugin and Python pipeline that turns a prompt or still image into a looping animated icon with a transparent background. It combines hosted image and video generation with local background removal and WebP encoding for app artwork.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.3k ▲19/7d · Pythonmini-AGImini-AGI is an experimental language model that trains on a stream of bytes and can keep learning from new files or conversations. It moves model components between disk, memory, and the GPU so researchers can study growing models on limited hardware. The README calls its current capabilities toy-level.SetupDocsCommunityMaturityVerdict inside AI Tools★ 4.1k ▲143/7d · TypeScriptOpenIntelligentUIOpenIntelligentUI is a chat application that turns agent answers into interactive explanations, comparisons, and small tools. It combines a web frontend with a Python agent so a user can change inputs inside an answer instead of asking for another paragraph.SetupDocsCommunityMaturityVerdict inside AI Tools★ 523 ▲9/7d · Pythonjev-chat-jarvis-macJarvis for Mac reads supported chat windows and suggests replies in a floating panel, with sending left to you. Its main documentation is Chinese, and there is no separate English README in the repository. It helps you draft responses without replacing your chat application.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.1k ▲38/7d · PythondeepopenDeepOpen is a Python engine that assigns predefined labels, ratings, or true-or-false probabilities to text and structured records. Its documentation is substantially Chinese, with an English introduction and detailed English API guidance available in the same README. It aims to support local routing decisions without generating free-form answers.SetupDocsCommunityMaturityVerdict inside AI Tools★ 865 ▲11/7d · Pythonjev-chat-windowsJevChat-Windows reads a supported chat window with local text recognition and suggests replies that you can insert and send yourself. Its main documentation is Chinese, with an English guide for KakaoTalk support rather than a full English manual. The interface supports Simplified Chinese and Korean, and model judgment and drafting use configured external services.SetupDocsCommunityMaturityVerdict inside AI Tools★ 8.2k ▲90/7d · Kotlinjev-chat-jarvisJev reads chats on Android and suggests replies that it can fill into the message box for you to send. The repository contains a Chinese edition and a separate English WhatsApp edition, with English documentation for both. It helps you choose a response and check it for unwanted commitments before sending.SetupDocsCommunityMaturityVerdict inside AI Tools★ 28k ▲177/7d · pm-skillsPM Skills is a collection of instructions and guided workflows that help an AI assistant with product discovery, planning, and launch decisions. It gives product managers a repeatable way to question assumptions and draft working documents inside tools such as Claude Code and Cowork.SetupDocsCommunityMaturityVerdict inside AI Tools★ 6.5k ▲52/7d · PythonDecepticonDecepticon is an autonomous red-team system that uses language-model agents to plan authorized security engagements and run tools inside a Kali Linux sandbox. It coordinates work across an engagement, including interactive shells and stored findings, for operators who can review the plan and its results.SetupDocsCommunityMaturityVerdict inside AI Tools★ 8.7k ▲3.9k/7d · TypeScriptcindyCindy is a desktop and mobile assistant that lets AI agents work with your files and applications while sharing context across providers. Its recent issue discussions and release notes are predominantly Chinese, but English README and contributor docs exist. This repository contains the clients and shared packages, while the backend is separate.SetupDocsCommunityMaturityVerdict inside AI Tools★ 48k ▲120/7d · PythonUmi-OCRUmi-OCR is an offline desktop tool for extracting text from screenshots, batches of images, and scanned documents, with a Chinese primary README and an available English guide. It gives you a local interface for selecting text, cleaning up reading order, and producing searchable PDFs without sending documents to a hosted OCR provider.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.5k ▲11/7d · PythonHy-MT2Hy-MT2 is Tencent's family of translation models for text that also has terminology, style, or formatting instructions. English and Chinese documentation are available, while much of the issue discussion is Chinese. The repository includes IFMTBench to check whether translators obey those constraints.SetupDocsCommunityMaturityVerdict inside AI Tools★ 18k ▲284/7d · Pythonlingbot-mapLingBot-Map turns a sequence of camera images into a three-dimensional reconstruction and estimates where the camera moved. It processes incoming frames while retaining geometric context, which makes it relevant to researchers working with long videos and robot perception.SetupDocsCommunityMaturityVerdict inside AI Tools★ 16k ▲46/7d · RustburnBurn is a Rust tensor library and deep-learning framework for training and running models from one codebase. Its main appeal is a common model API that can target CPU, CUDA, ROCm, Metal, Vulkan, WebGPU, remote devices, browsers, and some embedded environments.SetupDocsCommunityMaturityVerdict inside AI Tools★ 763 ▲23/7d · RustMimiMimi's primary README is Chinese, and maintained English setup and usage guides are available. It is a desktop and Android app that turns computer or microphone speech into floating original-language or translated subtitles, using cloud speech providers or supported Apple on-device services.SetupDocsCommunityMaturityVerdict inside AI Tools★ 11k ▲64/7d · TypeScriptcodex-securityCodex Security is a command-line scanner and TypeScript SDK that uses coding agents to find, validate, and help fix vulnerabilities in source repositories. It can turn a repository, path, Git diff, or imported alert backlog into findings with evidence, coverage records, patches, and machine-readable exports.SetupDocsCommunityMaturityVerdict inside AI Tools★ 15k ▲1.5k/7d · TypeScriptvideo-shotcraftVideo Shotcraft is a Claude Code and Codex skill for making product promos with Remotion, real interface captures, reusable shot recipes, and a browser editing workbench. It gives an agent a production process and working motion code, rather than generating a finished video from one prompt by itself.SetupDocsCommunityMaturityVerdict inside AI Tools★ 144 · JavaScriptdashengDasheng (大声读 · ReadAloud) is a Next.js English reading-practice app with Chinese-language documentation. It aligns a speech transcript with a passage and combines word matching, completeness, pace arithmetic, and Jev text judgments into feedback. It does not assess pronunciation, stress, or accent. Our recorded run verified dependency installation, a build, and the selected test command, not a live reading session.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1 · TypeScriptNorthstarNorthstar is a Chinese-language, Windows-first personal strategy workspace, and the repository provides no English guide. It turns long-term goals into weekly capacity, daily actions, reviews, guarded AI proposals, memories, and growth records, with a browser build and an unfinished Tauri desktop path.SetupDocsCommunityMaturityVerdict inside AI Tools★ 333 ▲8/7d · TypeScriptcrush-monitorCrush Monitor is Chinese-first; an English README exists, but the interface and analysis labels remain Chinese. It runs a local web app that sends pasted two-person chats to Jev or a text model, then labels emotions and intentions, scores signs of affection, and grades the user's replies.SetupDocsCommunityMaturityVerdict inside AI Tools★ 8.2k ▲90/7d · Kotlinjev-chat-jarvisJev Chat is an Android reply assistant written mainly in Kotlin, with English documentation plus Chinese and Vietnamese translations. It reads the open chat through Android accessibility APIs, asks models to judge intent and risk before drafting replies, then fills your chosen text into the message box without sending it.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.1k ▲45/7d · PythonQwen-Image-2.1Qwen-Image-2.1 is a downloadable image model for creating pictures from text and editing one or more reference images. Its English documentation covers transparent output, local edits, multi-image composition, and several third-party runtimes, while this repository itself mainly supplies examples and a separate prompt-rewriting utility.SetupDocsCommunityMaturityVerdict inside AI Tools★ 298 ▲22/7d · Markdownai-engineer-roadmapAI Engineer Roadmap is a Markdown reading list that orders AI topics from basic terminology through machine learning, language models, agents, inference, safety, deployment, and interviews. It solves the blank-page problem for beginners, but it is a publisher-led syllabus rather than a hands-on course or neutral survey.SetupDocsCommunityMaturityVerdict inside AI Tools★ 95 · PowerShellDeepseek-Harness-Easy-SetupDeepSeek Harness Easy-Setup is a Chinese-language Windows installer project for DeepSeek Harness, and the repository does not provide an English README. It bundles the upstream command-line package with a portable Node runtime, then adds shortcuts that start its local web interface or stop the process.SetupDocsCommunityMaturityVerdict inside AI Tools★ 8.4k ▲153/7d · TypeScriptZCodeZCode's primary README is written in Simplified Chinese, and the repository includes a detailed English README. ZCode is an AI coding workspace that puts the same agent behind desktop, browser, and terminal interfaces, with local and remote project support.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.8k ▲9/7d · Gopaperless-gptPaperless-GPT connects to Paperless-ngx and uses an LLM to suggest titles, tags, correspondents, dates, custom fields, and links between related documents. It can also run vision-based OCR, create searchable PDFs, and send approved changes back to the archive.SetupDocsCommunityMaturityVerdict inside AI Tools★ 12k ▲102/7d · TypeScriptG0DM0D3G0DM0D3 is an open-source chat interface that sends one prompt to several hosted or local language models, compares their replies, and returns a winner. Its main use is red-team prompt research, with a single-file browser app plus optional Next.js and Express surfaces.SetupDocsCommunityMaturityVerdict inside AI Tools★ 20k ▲10/7d · PythonolmocrolmOCR converts PDFs and page images into readable Markdown using a 7-billion-parameter vision-language model. It is built for documents whose tables, equations, handwriting, columns, headers, and reading order are hard for plain text extraction, with local GPU, remote inference, Docker, and S3 batch paths.SetupDocsCommunityMaturityVerdict inside AI Tools★ 329 ▲8/7d · Pythonembodied-jevEmbodiedJev is a Chinese-language browser workbench for letting rules, Jev, local models, or chat APIs choose actions for a Franka Panda arm inside MuJoCo. The main README and technical guides are in Chinese, and the repository has no English README; the interface covers three simulated manipulation tasks, experiment playback, and side-by-side model runs.SetupDocsCommunityMaturityVerdict inside AI Tools★ 289 ▲8/7d · RustunderclassUnderclass combines several ChatGPT/Codex and GitHub Copilot subscriptions behind one local OpenAI-compatible address. It keeps sessions on the same account, removes exhausted accounts from rotation, and brings them back when their quotas reset.SetupDocsCommunityMaturityVerdict inside AI Tools★ 601 ▲1/7d · JavaScriptminecraft-agentMinecraft Agent is a JavaScript Minecraft Java 1.16.5 experiment that combines Astra or Sol planning, JEV action selection, and Mineflayer execution on a fixed, surveyed route. It is useful for studying bounded model control. The supplied evidence does not establish general play on unknown worlds or isolate the models' contribution to completion.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.6k ▲16/7d · Pythonlaya-coremlLaya-CoreML runs specialized decision models locally on Apple Silicon and returns choices, scores, or yes/no probabilities instead of free-form text. It is for apps that need a small, predictable decision step without sending input to a hosted model API.SetupDocsCommunityMaturityVerdict inside AI Tools★ 649 ▲3/7d · RustCometixCodeCometixCode is an unofficial Rust reimplementation of Anthropic's Claude Code terminal assistant. It follows the published Claude Code interface and file structure closely, aiming to provide familiar coding-agent behavior in a native binary.SetupDocsCommunityMaturityVerdict inside AI Tools★ 306 ▲2/7d · PythonopenJev-verdict-2.0openJev-verdict-2.0 is a Python research project that turns a block of text into bounded choices, scores, or yes/no-style decisions with probabilities instead of generating prose. The repository covers two related models, a 151M-parameter Verdict checkpoint and a separate Verdict 2.0 architecture for typed software workflows.SetupDocsCommunityMaturityVerdict inside AI Tools★ 60 ▲4/7d · Pythoneditaplot2026EditaPlot's main README is written in Simplified Chinese, and the repository also provides a full English README. It is a Codex skill that inspects scientific tables, asks the user to confirm what each column means, then drives Origin 2026 to produce an editable OPJU project plus PNG, PDF, and TIF exports.SetupDocsCommunityMaturityVerdict inside AI Tools★ 257 ▲3/7d · post-production-skillPost Production Skill's README is in Simplified Chinese, with no English README, although the bundle includes an English master prompt. It is a narrow Codex skill that rewrites an idea or reference into a Seedance 2.5 prompt for one warm 1990s studio look, with a stable presenter, moving cameras, CRT hardware, and three-dimensional editing interfaces.SetupDocsCommunityMaturityVerdict inside AI Tools★ 7.0k ▲20/7d · Pythonlaya-mlxLaya-MLX is a native Apple Silicon runtime for open-weight Laya models that answer typed choice, score, and probability questions without generating text. It ports the encoder and decision heads to MLX for local macOS inference and includes English plus linked Chinese documentation.SetupDocsCommunityMaturityVerdict inside AI Tools★ 313 ▲2/7d · Pythonjev-alignjev-align is an experimental Python CLI for turning labeled examples into reusable Jev decision functions. It finds uncertain rows, asks a person to label them, and uses GEPA plus a separate reflection model to propose a better typed definition without accepting that proposal automatically.SetupDocsCommunityMaturityVerdict inside AI Tools★ 753 ▲4/7d · awesome-jev-toolsawesome-jev-tools is a curated index of projects and practices built around TypeSafe AI's Jev typed-decision model. It helps developers find Jev examples by use case, while explicitly declining to verify that listed projects build, pass tests, reproduce their claims, or permit reuse.SetupDocsCommunityMaturityVerdict inside AI Tools★ 4.6k ▲460/7d · Pythongeo-sleuthGeo-sleuth's working skill instructions and script output are written in Chinese, while the repository provides a full English README and says the agent can reply in your language. It gives coding agents a photo-geolocation workflow built from metadata, OCR, maps, terrain, satellite imagery, street views, and geometric checks, then asks for coordinates with an error radius and evidence image.SetupDocsCommunityMaturityVerdict inside AI Tools★ 326 ▲1/7d · PythonCovalent-MASCovalent-MAS is a small Python interface layer for wiring molecule generators, optimizers, docking tools, and trajectory records into a covalent-drug design workflow. The repository currently supplies data contracts, JSONL storage, a sequential AutoDock Vina adapter, and a CLI for already prepared ligands, not the generation and optimization models described in its results section.SetupDocsCommunityMaturityVerdict inside AI Tools★ 763 ▲16/7d · PythonopenjevOpenJev is a self-hosted decision server that answers yes-or-no, multiple-choice, and scored questions with probabilities instead of free-form text. It speaks TypeSafe Jev's wire protocol and can use DiffusionGemma or four smaller open models, giving developers a typed alternative to parsing an LLM response.SetupDocsCommunityMaturityVerdict inside AI Tools★ 873 ▲50/7d · TypeScriptagency-agents-appAgency Agents App is a native desktop installer and tracker for personas from the separate agency-agents catalog. Its README describes browsing source personas, rendering files for coding tools, recording managed installs and detecting outside edits. The coding tools execute the personas; this app does not run agents.SetupDocsCommunityMaturityVerdict inside AI Tools★ 9.1k ▲35/7d · CudaDeepGEMMDeepGEMM is a CUDA kernel library for the matrix multiplication, mixture-of-experts, and attention work inside modern language models. It gives engineers optimized FP8, FP4, BF16, and TF32 operations for recent NVIDIA data-center GPUs, with kernels compiled at runtime through DeepJIT.SetupDocsCommunityMaturityVerdict inside AI Tools★ 888 ▲20/7d · Pythonguizang-product-video-skillGuizang Product Video Skill is a Chinese-first Claude Code and Codex skill. The repository documentation is in Chinese, and the README does not link an English version. It turns a software repository or changelog into a code-rendered product promo, including the story, product UI, animation, music, sound effects, MP4, and editable project.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.3k ▲15/7d · PythonnimbleNimble is a Python recipe and local model for making typed decisions about text. You provide a flat schema of boolean or multiple-choice questions, and it returns one allowed answer per field with candidate probabilities instead of generating prose.SetupDocsCommunityMaturityVerdict inside AI Tools★ 837 ▲6/7d · TypeScriptlocaljevLocalJev is a Bun and TypeScript service that accepts typed Jev decision questions and sends them to a local model through an OpenAI-compatible chat endpoint. It solves a specific compatibility problem for developers who want Jev-shaped choices, scores, and confidence values without running OpenJev's patched vLLM backend.SetupDocsCommunityMaturityVerdict inside AI Tools★ 326 ▲4/7d · Pythonjev-visualJev Visual is an English-language educational project, with a Simplified Chinese README too, that uses a small vision-language model on Apple Silicon to answer typed questions about an image. It reuses shared visual context and scores candidate answers directly, giving you a local UI, CLI, and HTTP API for learning the technique.SetupDocsCommunityMaturityVerdict inside AI Tools★ 249 ▲4/7d · TypeScriptjev-reviewJev Review is a local MCP server that gives coding agents structured software-quality scores while they work. It sends only the task, diff, files, and repository context supplied by the agent to TypeSafe's Jev API, then turns the response into per-dimension scores and comparisons.SetupDocsCommunityMaturityVerdict inside AI Tools★ 35k ▲408/7d · PythonlayaLaya is a local decision model that turns text into typed choices, scores, or yes/no probabilities in one model pass. Its router selects among English, multilingual, and specialized checkpoints, giving applications a smaller decision layer for triage, guardrails, and routing.SetupDocsCommunityMaturityVerdict inside AI Tools★ 38k ▲1.2k/7d · RustartcraftArtCraft is a desktop workspace for composing AI images, video, sound, 3D meshes, and virtual scenes. It lets an artist arrange layers, props, poses, and cameras before sending work to hosted generation models. The code is source available under a custom work-in-progress license, not an OSI open-source license.SetupDocsCommunityMaturityVerdict inside AI Tools★ 386 ▲17/7d · JavaScriptZeroScript-FreeZeroScript-Free connects a supported AI chat website to Roblox Studio through a browser extension and a local Python bridge. It lets the chat read or edit scripts, run Luau, inspect a place, and control play tests through Studio's MCP server. You avoid an API key, but the workflow depends on provider webpages that can change without notice.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.8k ▲88/7d · Goawesome-ai-agentsAwesome AI Agents is a curated directory of public agent projects, frameworks, memory systems, evaluation tools, guardrails, and MCP software. It helps builders find candidates by category while showing repository ownership, submission provenance, maintainership evidence, stars, and recent star growth.SetupDocsCommunityMaturityVerdict inside AI Tools★ 412 ▲3/7d · TypeScriptjev-experimentsJev Experiments is a collection of 21 small applications showing how TypeSafe's Jev model can make fast, probability-backed decisions without writing prose. Each demo applies that pattern to a concrete job such as support triage, command safety, search, moderation, or routing, but the repository is a gallery rather than one reusable product.SetupDocsCommunityMaturityVerdict inside AI Tools★ 587 · PythonOrcaBonsai-27B-UncensoredOrcaBonsai is a runtime add-on that reduces refusal behavior in the separate Ternary Bonsai 2 27B model without rewriting its compressed weights. It is research tooling for Apple Silicon and a particular model pack, not a new model checkpoint or a general chat application.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.6k ▲9/7d · PythonNanoJevNanoJev is a 0.6B-parameter model that scores supplied choices, Boolean claims, and ordered levels without generating an answer token by token. It packages one checkpoint, training data, an HTTP inference service, and browser replays for four game tasks so researchers can inspect how direct probability-based decisions behave.SetupDocsCommunityMaturityVerdict inside AI Tools★ 371 ▲4/7d · Pythonuplifting-biomolecular-modelingUplifting Biomolecular Modeling is a reference collection of 36 optimization kits for protein and genomics model inference. Each kit pins one upstream tool, adds faster or lower-memory execution modes, and documents the exact software, GPU, weights, and patches needed to compare its optimized path with stock behavior.SetupDocsCommunityMaturityVerdict inside AI Tools★ 529 ▲8/7d · JavaScriptprocedural-filmprocedural-film is an agent skill for making short animated films or a browser platform game from JavaScript-generated pictures and sound. Its detailed pipeline coordinates research, storyboards, scene-writing subagents, automated visual checks, browser rendering, audio synthesis, and MP4 export.SetupDocsCommunityMaturityVerdict inside AI Tools★ 695 ▲3/7d · TypeScriptjev-reviewJev Review is a local TypeScript workflow that asks TypeSafe Jev to review a Git branch diff or scan a JavaScript and TypeScript codebase. It screens files first, spends extra model calls on the strongest signals, and saves findings to a local dashboard for human review.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.5k ▲24/7d · Dream-RSIDream-RSI is a research paper and project page that evaluates a new coding-agent exploration method across 8 tasks. It treats the tree of earlier attempts as a replayable world, then tests new branching and stopping policies against recorded outcomes before using a revised policy online.SetupDocsCommunityMaturityVerdict inside AI Tools★ 458 ▲3/7d · Pythontypesafe-marioTypeSafe Mario lets the Jev model choose controller moves in the original Super Mario Bros. from structured emulator data instead of screenshots. It turns RAM and telemetry into JSON about Mario, nearby threats, terrain, timing, and recent actions, then records each choice for inspection.SetupDocsCommunityMaturityVerdict inside AI Tools★ 4.9k ▲26/7d · PythonSemIf-OpenJevSemIf, formerly OpenJev, uses an open language model to score a set of choices directly instead of generating an answer and parsing it back into a decision. It is aimed at small runtime judgments such as routing a request or checking whether evidence supports a claim, with local GPU, Apple Silicon, CPU, and browser paths.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.4k ▲4/7d · PythonjevlikeJevlike is a small Python model for choosing among a list of text options that can change with each request. It returns one probability per option in a single pass, which suits routing and menu selection when generating a prose answer would be wasteful.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.2k ▲22/7d · PythonNeuralScreenNeuralScreen is a Windows desktop overlay that applies NVIDIA neural rendering and optional frame generation to a whole display or one window. It can also capture screenshots, record processed video with sound, convert local media, and publish frames to OBS through Spout2.SetupDocsCommunityMaturityVerdict inside AI Tools★ 558 ▲4/7d · Pythonvideo-generator-clientVideo Generator Client is an async Python wrapper that puts Seedance, Kling, MiniMax/Hailuo, and Wan behind one interface. It includes a CLI, a local FastAPI web screen, and a small MCP server, so one prompt workflow can submit jobs, poll status, and download results across providers.SetupDocsCommunityMaturityVerdict inside AI Tools★ 227 ▲1/7d · Swiftclaude-siri-aiClaude for Siri is an experimental macOS 27 app that puts a signed-in Claude Code account in Spotlight's native Ask menu. A local Swift bridge sends text prompts to the Claude CLI and streams the reply back into Apple's response interface.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.0k · GoARTEXARTEX is a Chinese-language, self-hosted system that lets multiple AI agents plan and carry out penetration-testing work inside an asset and evidence dashboard. Its README and sole file in `docs/` are in Chinese, and the repository has no separate English README or English documentation file. It tracks targets, tool calls, traffic, findings, approvals, and the chain of evidence behind each task.SetupDocsCommunityMaturityVerdict inside AI Tools★ 13k ▲62/7d · PythonchandraChandra is an OCR model and Python toolkit that turns images and PDFs into Markdown, HTML, or JSON while retaining page layout. It is aimed at documents that ordinary text extraction handles badly, including tables, handwriting, forms, diagrams, and mathematical notation.SetupDocsCommunityMaturityVerdict inside AI Tools★ 10.0k ▲22/7d · Pythonproduction-agentic-rag-courseProduction Agentic RAG Course is a seven-week Python course built around an arXiv research assistant. It teaches the whole path from PostgreSQL and keyword search to hybrid retrieval, a local Ollama model, tracing, caching, LangGraph decisions, and a Telegram bot.SetupDocsCommunityMaturityVerdict inside AI Tools★ 653 ▲6/7d · HTMLGPT-as-PolicyGPT-as-Policy is an English and Chinese research release that compares GPT 6 Astra acting directly as a robot policy with a hybrid policy that can correct actions proposed by a vision-language-action model. It packages evaluation code, selected results, videos, and a rebuildable report, while leaving simulator assets, model checkpoints, and original operations data to their upstream or private environments.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.2k ▲22/7d · PythonNeuralScreenNeuralScreen is a Windows desktop overlay for NVIDIA neural rendering and optional frame generation over a whole display or one window. The README documents screenshots, processed recording and Spout2 output. Our recorded Debian run checked repository commands at commit e7dd41d, not Windows rendering quality or GPU compatibility.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.7k ▲21/7d · KotlinmuralMural is a native iPhone and Android language-learning app built around spoken conversations with an AI tutor. It saves transcripts, vocabulary, and learning observations on the phone, then revisits words in later conversations instead of presenting a fixed lesson path.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.1k ▲181/7d · HTMLrecurrent-looped-tranformerRecurrent Looped Transformer is a research report and project website for an experimental language-model architecture that carries a hidden state from one token to the next. The repository explains the design and publishes papers, figures, and reported results, but it does not contain a runnable training or inference implementation.SetupDocsCommunityMaturityVerdict inside AI Tools★ 211 · Pythongpu-timegpu-time is a local neural parser that turns short English or Spanish time phrases into dates, ranges, and recurrence rules. It runs in JavaScript on CPU or WebGPU, so reminder fields and command bars can interpret a schedule without sending the text to a server.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.1k ▲28/7d · PythonOpenWAMOpenWAM is a research stack for training robot policies that learn from video, state, and action data. It gives researchers interchangeable model parts, dataset adapters, training code, deployment tools, and benchmark clients so they can compare world-action model designs without rebuilding the whole pipeline.SetupDocsCommunityMaturityVerdict inside AI Tools★ 192 · xialingguo-ipxialingguo-ip and all of its documentation are written in Chinese, with no English guide in the repository. It is an instruction pack for asking an image-capable AI assistant to make article covers for a Chinese AI creator, using nine named visual styles and a short intake form.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.2k ▲21/7d · JavaScriptreelbench-skillsReelbench Skills is a Chinese-first set of three Claude Code and Codex skills for analyzing video shots, compositing shot notes beside footage, and removing video metadata. English README and per-skill documentation exist, while the main skill instructions and much of the source commentary are Chinese.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.6k ▲40/7d · TypeScriptSoL-PiSoL-Pi is an extension for the Pi coding agent that tries to cut repeated tool calls, replayed output, long diagnostic logs, and stale task context. It adds four opt-in mechanisms without patching Pi itself.SetupDocsCommunityMaturityVerdict inside AI Tools★ 151 · flybookStonkfly feeds Coinbase price charts into a simulated fruit-fly nervous system, then turns selected neural activity into buy, sell, or hold proposals. It can paper trade or place guarded spot orders, but the project states plainly that profitable learning has not been demonstrated.SetupDocsCommunityMaturityVerdict inside AI Tools★ 465 · Pythoncrypto-ragCrypto RAG is an Indonesian-language command-line assistant for crypto concepts, live market data, alerts, and portfolio tracking; it does not provide English documentation. It keeps explanatory material in a local retrieval index and fetches changing prices, funding, TVL, sentiment, and network data from public APIs when you ask.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.9k ▲111/7d · TypeScriptanything2explaineranything2explainer is a Claude Code and Codex skill for producing narrated, horizontal motion-graphics explainers in English or Chinese. It supplies a Remotion template plus a nine-stage method for research, narration, speech, storyboarding, shot construction, rendering, subtitles, and visual review.SetupDocsCommunityMaturityVerdict inside AI Tools★ 183 · Pythoncek-probe-modelCek Probe Model is an Indonesian-language set of Python scripts for checking OpenAI-compatible model endpoints; it does not provide English documentation. It lists available models, sends identity or logic prompts, and prints a summary intended to flag unavailable endpoints or suspicious model labels.SetupDocsCommunityMaturityVerdict inside AI Tools★ 524 ▲9/7d · CudaDeepSelectDeepSelect is a bilingual English-and-Chinese PyTorch extension for the TopK operation used in DeepSeek Sparse Attention and sampling. It supplies specialized NVIDIA CUDA and Huawei Ascend kernels for a narrow set of bfloat16 and float32 workloads, with full English setup and algorithm notes alongside Chinese material.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.6k ▲126/7d · PythonMicroduck-build-tutorialMicroduck is a compact biped-robot build whose primary README opens in Chinese and later repeats the guide in English. It combines Raspberry Pi deployment code, printable parts, wiring instructions, a prebuilt SD-card image, and a MuJoCo reinforcement-learning environment for training the included ONNX walking policy.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.1k ▲53/7d · RustChoruzChoruz is a local-first, Slack-like workspace where people and command-line AI agents share messages, threads, tasks, and files. It runs real Claude Code, Codex, Grok, or plugin-supplied agent sessions in separate workspaces, giving a team one place to assign work and inspect what those agents send back.SetupDocsCommunityMaturityVerdict inside AI Tools★ 264 ▲1/7d · haiming-app-monetizationHaiming App Monetization is a Chinese-language agent skill for reviewing mobile onboarding, paywalls, purchase options, and competitor pricing; it has no English documentation. It reads an app project or supplied screenshots and produces a product plan, with code changes allowed only when the user explicitly asks for implementation.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.7k ▲45/7d · Pythonlieflat-gongwenLieflat Gongwen is a Chinese-language agent skill for drafting seven kinds of Chinese government and Party-affairs material; it has no English documentation. It gives Codex, Claude Code, Moxt, and other agents.md-compatible tools genre-specific writing rules plus a Python checker, while explicitly excluding fixed-format statutory documents.SetupDocsCommunityMaturityVerdict inside AI Tools★ 34k ▲27/7d · JavaScriptnetronNetron opens machine-learning model files as interactive graphs, so you can inspect layers, connections, shapes, and stored values without writing a loader for each framework. It runs in a browser, as a desktop app, or from Python, which makes it useful for checking an unfamiliar model before you wire it into an application.SetupDocsCommunityMaturityVerdict inside AI Tools★ 43k ▲25/7d · PythonBettaFishBettaFish is a Chinese-first multi-agent system for collecting online discussion, analyzing sentiment, and producing public-opinion research reports. An English README and contribution guide exist, though the main documentation, issue activity, and bundled report templates are primarily Chinese. It combines web and social collection, private database analysis, several model roles, agent discussion, and HTML, PDF, or Markdown report generation.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.4k ▲162/7d · PythonUniMateUniMate is an English-language research system that turns a text description into motion for humanoids, animals, and other rigged 3D objects. One model handles different skeleton layouts, then exports motion data and rendered previews that can be applied back to a mesh.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.5k ▲178/7d · JavaScriptunigit-ecosystemUNIGIT Ecosystem is the public brand, vision, and partner-intake repository for a proprietary AI workbench. It explains what UNIGIT wants to build and screens public contributions, but it does not contain the product runtime or an extension SDK.SetupDocsCommunityMaturityVerdict inside AI Tools★ 6.6k ▲372/7d · HTMLhandraw-styleHandraw Style is a Chinese-language skill and visual catalog for turning numbered drawing styles, layouts, and colors into image-generation prompts. Full English documentation exists. It helps creators choose a repeatable visual direction without knowing the usual art and composition vocabulary.SetupDocsCommunityMaturityVerdict inside AI Tools★ 731 ▲4/7d · KotlinsouchastnikSouchastnik is a Russian-language Android keyboard that uses a local model to flag possible Russian legal consequences while you type. Its README, interface, and developer notes are in Russian, and we found no English documentation. It also inserts labels for named organizations and people from local dictionaries, without requesting internet access.SetupDocsCommunityMaturityVerdict inside AI Tools★ 4.9k ▲135/7d · dlssg_for_sm86DLSSG for SM86 is a Chinese-language Windows mod that enables NVIDIA DLSS Frame Generation on RTX 20-series and RTX 30-series cards. A full English README is available. It places a proxy DLL beside a game's executable so supported D3D12 games can expose frame-generation modes normally reserved for newer GPUs.SetupDocsCommunityMaturityVerdict inside AI Tools★ 87 · HTMLglm-flash-offline-clientGLM Flash Offline Client presents itself as a Windows desktop app for running GLM-5.3-Flash locally. The repository is a documentation and landing-page shell, however: its two named source files are empty, and GitHub has no release containing the client described in the README.SetupDocsCommunityMaturityVerdict inside AI Tools★ 188 ▲2/7d · Pythoncyber-resume-reviewer-skillCyber Resume Reviewer is an Agent Skill that tells an AI assistant how to critique, tailor, score, or rewrite IT and cybersecurity resumes without inventing evidence. It can produce a Markdown report and a matching PDF, but it is a set of instructions, references, templates, and local helpers rather than a standalone hiring app.SetupDocsCommunityMaturityVerdict inside AI Tools★ 618 ▲41/7d · JavaScriptRuiC-card-skillRuiC Card Skill's primary README is Chinese, and the repository also provides a full English README. It gives a multimodal coding agent a recipe and scripts for turning layered artwork into an interactive holographic card website, an editable Blender project, renders, and a browser verification report.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.2k ▲10/7d · PythonitsgivingIt's Giving is a local webcam effect that recognizes 14 facial expressions or poses and places a matching meme over your head. It can publish the altered frame as a virtual camera for Zoom, Google Meet, Teams, Discord, or OBS. A calibrated second script learns your neutral face for seven seconds so fixed thresholds are less likely to misread you.SetupDocsCommunityMaturityVerdict inside AI Tools★ 274 ▲4/7d · JavaScriptholo-cardHolo Card is a Codex skill that turns supplied artwork into an interactive layered card with tilt, depth, flip, moving background, and contour glow. It guides image generation and repair, then assembles the approved layers into self-contained HTML and a ZIP; an optional local API covers agents without native image tools.SetupDocsCommunityMaturityVerdict inside AI Tools★ 10 · PythonAwesome-OKFAwesome-OKF is a 29-entry catalog of tools, plugins, Claude Code skills, proposals, and documentation for Google's Open Knowledge Format. It also supplies a Python CLI, a content converter, and an MCP server so people and agents can search the same YAML catalog. The practical goal is to make agent-readable knowledge projects easier to discover and turn ordinary documents into OKF bundles.SetupDocsCommunityMaturityVerdict inside AI Tools★ 30k ▲69/7d · PythonfacefusionFaceFusion is a face-manipulation application with interactive, headless, batch and queued-job commands. Its documented automation options make it worth evaluating for repeatable media work, but our September 30, 2026 sandbox run at commit 7247081 did not establish working media processing or production readiness.SetupDocsCommunityMaturityVerdict inside AI Tools★ 39k ▲4.8k/7d · PythoniFixAiiFixAi is a Python diagnostic that probes whether an AI agent follows the business rules, permissions, evidence standards, and escalation paths defined for it. It produces a scorecard and grade from fixed inspections, while marking checks inconclusive when the adapter cannot expose enough evidence.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.8k ▲27/7d · JavaScriptdream-loopDream Loop is an agent skill that turns a generated target image into a visual specification, then asks a coding agent to keep revising a game, app, or 3D scene toward that image. It solves the vague feedback problem in visual builds by giving the builder a fixed screenshot and, in Pro mode, a separate critic with a scoring rubric.SetupDocsCommunityMaturityVerdict inside AI Tools★ 183 · PythonCodex-Minecraft-GameplayCodex Minecraft Gameplay gives a computer-use agent bounded keyboard, mouse, and screenshot controls for Minecraft on a visible Windows desktop. It supplies the control layer and visual checks, while you supply Minecraft, an agent, and the goal.SetupDocsCommunityMaturityVerdict inside AI Tools★ 445 ▲9/7d · kunKun is an English-language agent skill that answers questions and directs engineering work using a living collection of Kun Chen's public opinions, tools, workflows, and writing style. The installed skill stays small, then fetches about 131 KB of Markdown from the repository's current `main` branch during a session.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.9k ▲96/7d · Pythonscreenwriting-skillsscreenwriting-skills is a Chinese-language collection of 26 Agent Skills for film, television, and theatre writing; English, Japanese, Korean, and French READMEs explain the package, but the instruction bodies remain Chinese. Claude Code, Codex, and other skill-aware agents use the material to guide structure, character, dialogue, series development, adaptation, opera, workflow, and revision in the language of the user's question.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.0k ▲33/7d · Pythonholo-card-studioHolo Card Studio is a Codex skill whose primary documentation is in Chinese, with a full English README available. It turns a prompt or reference image into an editable Blender card and an interactive Three.js viewer, covering both holographic and two-image lenticular designs.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.5k ▲139/7d · HTMLmicroduck-replicaMicroduck Replica is primarily a Chinese-language reconstruction project for Pollen Robotics' small biped robot duck, with a substantial English README available. It combines mechanical findings, electronics, firmware, a servo console, board-image work, and reinforcement-learning training material so builders can reproduce either the original-servo design or a cheaper Feetech version.SetupDocsCommunityMaturityVerdict inside AI Tools★ 7.8k ▲44/7d · PythonreefReef is infrastructure for collecting agent interactions, matching feedback, producing candidate updates, evaluating them, and publishing accepted versions. It can change model weights through training recipes or revise an agent harness, including prompts, rules, skills, and commands.SetupDocsCommunityMaturityVerdict inside AI Tools★ 243 ▲1/7d · TypeScriptmobilecodeMobileCode is a fork of the OpenCode coding agent that adds iOS Simulator and Android Emulator controls beside the chat session. It can detect a mobile project, build and install it, stream the virtual device into the app, and return compiler or Gradle errors to the agent.SetupDocsCommunityMaturityVerdict inside AI Tools★ 8.2k ▲933/7d · Shellopen-seo-mcp-skillsOpen SEO MCP Skills is a set of eight English-language Claude skills for audits, keyword work, rank tracking, backlinks, content briefs, and ad overlap. The repository contains the instructions, while the data comes through Ryze's hosted MCP connector and your connected analytics or advertising accounts.SetupDocsCommunityMaturityVerdict inside AI Tools★ 11k ▲35/7d · TypeScriptvoltagentVoltAgent is a TypeScript framework for building agents with tools, memory, workflows, supervisors, guardrails, retrieval, voice, evaluations, and MCP connections. Its companion VoltOps product adds tracing, automation, deployment, and a console, either in the cloud or self-hosted.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.0k ▲30/7d · PythonInferenceXInferenceX is an open research platform for repeatedly benchmarking large-language-model serving software on current accelerators. It supplies configurations, launchers, evaluation tooling, result processing, and a public dashboard so performance comparisons can change as vLLM, SGLang, TensorRT-LLM, CUDA, ROCm, and hardware change.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.5k ▲22/7d · ShellBonsai-demoBonsai Demo packages the scripts, web interface, and custom runtimes needed to run PrismML's ternary Bonsai 2 27B model on a local computer. It gives Mac, Linux, and Windows users one place to download the model, start an OpenAI-style server, send images, call tools, and connect MCP servers.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.1k ▲68/7d · JavaScriptqwen-audio-agentQwen Audio Agent is a runtime for keeping a voice conversation active while a separate coding or task agent works in the background. It connects replaceable voice services to agent backends, then presents the result through a terminal, browser, or desktop app.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.7k ▲31/7d · Pythonwechat-intelligence-hubWeChat Intelligence Hub is a Chinese-first Python project for turning your local WeChat history into search, daily briefings, reply queues, promises, follow-ups, and opportunity notes. The root README and full usage guide are Chinese; the repository has some English security text but no complete English setup guide. It ships a read-only database reader plus Codex skills that interpret the local records without sending messages.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.7k ▲75/7d · Pythondlss5-visual-enhancerVisual Enhancer is a portable Windows app for arranging RTX image and video effects into a previewable export pipeline. Its code is public, but its own license calls it source-available proprietary software, not open source.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.4k ▲10/7d · PythonABot-ReconABot-Recon turns a long, ordered video into camera poses and a 3D point reconstruction while keeping only a 12-frame local context. It is aimed at researchers who need streaming geometry without model memory that grows with every frame.SetupDocsCommunityMaturityVerdict inside AI Tools★ 163 · TypeScriptunreelUNREEL is a Next.js streaming site that writes an episode in batches and generates each video clip while the previous clip plays. It demonstrates an endless AI channel using fal-hosted video, image, and language models, with one mode for continuous stories and another for hard-cut comedy channels.SetupDocsCommunityMaturityVerdict inside AI Tools★ 855 ▲4/7d · JavaScriptcamera-to-blenderCamera to Blender turns one phone or webcam photo into a 3D model and sends that model into an open Blender scene. A small FastAPI relay connects the camera page, Tripo's hosted model generator, optional Gemini background removal, and a Blender add-on.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.3k ▲136/7d · DLSS-NR-on-AMDDLSS-NR on AMD is a Windows game mod that reimplements the runtime needed to apply NVIDIA's DLSS 5 Neural Rendering pass on recent AMD Radeon cards. GitHub detects no source language because the repository contains documentation and release binaries rather than the implementation; the English README explains how to inject the mod into a DirectX 12 game that already supports FSR.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.4k ▲44/7d · Pythoncommerce-agentsCommerce Agents is Anthropic's reference code for two Claude-based assistants: one helps customers shop, and one helps staff manage catalog, inventory, pricing, and campaigns. It gives teams reusable tool contracts, safety gates, and demo storefronts, but leaves each company to connect its own systems and enforce its own business rules.SetupDocsCommunityMaturityVerdict inside AI Tools★ 65 · PythonnovoweaveNovoWeave is a Python blueprint for arranging the parts of a generative protein-design system: constraints, backbone proposals, sequence design, evaluation, ranking, and provenance. It does not generate proteins; its own example ends in NotImplementedError, so the useful product is the software contract rather than a working model.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.2k ▲14/7d · Pythonrizzo-piiRizzo PII is an Italian-first desktop app and local HTTP service for finding personal data, replacing it with typed placeholders, and restoring it after a cloud model responds. Its main README and setup guidance are available in English, while the detector concentrates on Italian legal identifiers such as codice fiscale, partita IVA, and cadastral references.SetupDocsCommunityMaturityVerdict inside AI Tools★ 3.3k ▲39/7d · PythonredamonRedAmon is a self-hosted offensive-security platform that joins reconnaissance, scanners, an AI agent, a Neo4j attack graph, finding triage, and proposed code fixes. It is built for authorized assessments where a human can set scope, approve dangerous actions, and review any pull request it creates.SetupDocsCommunityMaturityVerdict inside AI Tools★ 19k ▲108/7d · PythonSkillOptSkillOpt improves the Markdown instruction files that guide an AI agent while leaving the model itself unchanged. It runs tasks, studies the results, proposes bounded edits, and keeps a candidate only when it passes validation; a separate preview called SkillOpt-Sleep can learn from supported coding-agent session histories.SetupDocsCommunityMaturityVerdict inside AI Tools★ 672 ▲3/7d · Astroawesome-ai-agent-platformsAwesome AI Agent Platforms is a curated directory of 49 agent products, frameworks, automation systems, browser agents, and coding agents. The README and companion Astro site record one category, a short description, the stated license, and the hosting model for each entry.SetupDocsCommunityMaturityVerdict inside AI Tools★ 522 ▲7/7d · Pythonguizang-yingzao-skillYingzao is a Chinese-language Claude Code and Codex skill that turns photos of Chinese buildings, streets, shops, crafts, and local food into editorial posters. The repository has no English README, so English-only users must translate a detailed workflow before they can judge or operate it safely.SetupDocsCommunityMaturityVerdict inside AI Tools★ 660 ▲14/7d · Pythonvdn-minimax-h3VDN-Minimax-H3 is a video-generation model and inference stack that adds hybrid attention to MiniMax H3. It generates text-guided video with audio, accepts first or last frames and reference images, and includes the code used to train its added attention branch and adapters.SetupDocsCommunityMaturityVerdict inside AI Tools★ 266 ▲1/7d · JavaScriptimage-to-3d-pipelineImage to 3D Pipeline is a published reconstruction experiment built around one fictional submersible. It compares several image-to-mesh models under a fixed Blender inspection routine, then displays the selected GLB in a Three.js explorer.SetupDocsCommunityMaturityVerdict inside AI Tools★ 51k ▲924/7d · Pythonnature-skillsNature Skills is a Chinese-first collection of agent instructions and helper code for scientific writing, literature work, figures, presentations, and research administration. A full English README and English pages for individual skills are available, so English-speaking researchers can evaluate and install it without translating the main documentation themselves.SetupDocsCommunityMaturityVerdict inside AI Tools★ 368 ▲5/7d · Pythonimage-story-video-wizardImage Story Video Wizard is a Codex and WorkBuddy skill that guides an audio-first illustrated video through 16 gated stages, from the brief to final feedback. The main README is Chinese with no English README, while the core skill and reference documents are in English; it organizes decisions and handoffs rather than generating a finished video by itself.SetupDocsCommunityMaturityVerdict inside AI Tools★ 2.4k · JavaScriptBrowserKittenBrowserKitten, identified in the README as PawWork_ZhuaZhua, is an unpacked Chrome MV3 extension for controlling the current browser tab through a side-panel agent. The README describes browser actions, guest JavaScript execution, and session artifacts such as sheets, Univer documents, and site HTML. It requires your own OpenAI-compatible model key.SetupDocsCommunityMaturityVerdict inside AI Tools★ 1.5k ▲20/7d · HTMLconsulting-pptx-skillConsulting PPTX Skill is a Japanese-first Claude Code skill for planning, laying out, checking, and reviewing consulting-style slide decks. The README has a short English summary, but its working instructions, design rules, and review prompt are mainly Japanese. It produces 16:9 HTML and PDF from a 62-layout library rather than generating an editable PowerPoint file directly.SetupDocsCommunityMaturityVerdict inside AI Tools★ 155 ▲1/7d · JavaScriptaipass-bridgeaipass-bridge turns a logged-in AiPASS browser tab into a local OpenAI-compatible endpoint for chat, files, media generation, and a small coding agent. It is aimed mainly at Thai citizens who already receive model access through AiPASS and want to use that account from a terminal or program.SetupDocsCommunityMaturityVerdict inside AI Tools★ 670 ▲17/7d · image-prompt-reverseImage Prompt Reverse is a Chinese-first Codex skill that turns a reference image into bilingual prompts for image generators. English documentation is included, but the default deliverable starts with a 450 to 700 Chinese-character prompt and then gives an English equivalent.SetupDocsCommunityMaturityVerdict inside