Top Open-Source GitHub Projects : colibri, ArtCraft, openGym, WeKnora, SCSKiller & Whirl #299 Transcript, AI Summary & Key Points

ArtCraft is a desktop IDE for interactive AI image and video creation, developed by Storytold. It lets artists compose in 2D, stage scenes in 3D, and select models for image, video, music, sound, 3D-mesh, and world generation. Its visual tools include 2D and 3D compositing, image-to-location scene planning, image-to-3D mesh conversion, character posing and identity transfer, kitbashing, background removal, canvas editing, masking, inpainting, prompted image editing, text-to-image generation, and image-to-video animation. The project supports integrations including ArtCraft's catalog, Grok, Midjourney, Sora, and World Labs, and is distributed through stable Windows and macOS releases; Linux users can build it from source.

C64 Keyboard is a font that reconstructs the printed lettering on classic Commodore 64 keycaps from photographs of a Central European machine. It includes uppercase letters, numbers, punctuation, arrows, £, π, selected accented letters, function-key labels, and all 63 PETSCII graphic front legends; lowercase input displays as uppercase because the font represents keycap legends rather than the Commodore 64 screen font. The PETSCII symbols are reconstructed outlines with key identities and assignments based on the Ultimate Commodore 64 Reference, while frame thickness, curves, and print proportions are inferred from the photographs rather than traced exactly. The project provides TTF and OTF desktop fonts, a webfont, editable outlines, build scripts, documentation, and a browser-based specimen and type tester. Its repository contributions, including the font files, outlines, scripts, preview, mappings, and documentation, are dedicated to the public domain under CC0 1.0; the reference photographs and rights in historical keycap designs or trademarks are not included or granted.

Colibrì is a pure-C inference engine and open research platform for running large mixture-of-experts models on consumer and heterogeneous hardware without engine dependencies. It treats VRAM, system RAM, and storage as a unified multitier inference hierarchy, streaming expert weights from disk and coordinating placement, storage I/O, scheduling, kernels, speculation, and CPU/GPU overlap. The project provides chat, server, and web front ends, with support for CPU, CUDA, Metal, and dual-SSD streaming; its design rules prohibit silently changing model precision or router semantics when fast memory is insufficient.

decider is an open-source AI model family and Python package for one-pass typed decisions with calibrated probabilities. Instead of generating text, it reads a text or JSON state and a set of typed questions—Choice, Score, or Noul yes/no questions—and returns a probability distribution over each question's allowed answers from one forward pass. It renders each question as an answer slot, reads the model's hidden state at that slot, projects it onto the valid option-label tokens, and applies a fitted temperature; there is no decoding or output parsing. The package provides `decide()` and `system_one()` interfaces, an HTTP server with `/decide` and `/v1/systemone` endpoints, and GGUF loading through llama.cpp. Models are available in sizes intended for CPUs and laptops as well as CUDA and Apple Silicon GPUs, and the repository includes training, calibration, serving, evaluation, and schema-caching components. The project is an independent reproduction of the System One model class, is not affiliated with TypeSafe AI, and is released under the Apache 2.0 license.

Diagram Design is a diagram-generation skill for Claude Code, Codex, Factory Droid, and Pi. It produces self-contained HTML and SVG diagrams through 39 editorial visual types, including architecture, flowchart, sequence, state-machine, data-flow, timeline, Sankey, Wardley map, kanban, user journey, deployment, dependency graph, UML class, story map, and database-schema diagrams. It can read a website to match its brand, and can redraw draw.io or Mermaid sources at a selected format, size, and detail level. Its semantic system patterns describe behavior separately from layout, so queues, policy traces, and trust boundaries can use an existing diagram type without adding new layout types. The repository supplies each type in minimal-light, minimal-dark, and full-editorial variants. Static HTML is the default; optional accessible motion is available for ordered explanations. The diagrams require no build step, JavaScript, or external image dependency.

Fieldwatch is a receive-only Android observer for nearby Wi-Fi access points and Bluetooth Low Energy advertisements. It passively listens through the phone’s built-in radios, working offline without a dongle, account, backend server, or ads. The app provides configurable views, filtering, an extensible device-signature library, local storage, and reports of detected radio sources. It is distributed as a sideloadable APK for Android 10 and later, with source code and installation materials in the repository, and is licensed under the MIT License. It does not support Wi-Fi monitor mode, Bluetooth Classic inquiry, cellular detection, or direction finding.

i-have-adhd is an open-source skill/plugin for coding assistants that formats responses as ADHD-friendly output. It leads with the next action, numbers multi-step tasks, suppresses tangents, restates state, gives specific time estimates, uses matter-of-fact errors, limits lists to five items, and ends with one concrete next step. The skill is installed through a coding assistant's CLI and can be customized by editing its SKILL.md file. It is licensed under the MIT License and does not require an ADHD diagnosis.

Mortiflix is a local motion-design studio for producing AI-assisted videos with Claude. It guides a project through briefs, scripts, style frames, transition boards, animatics, builds, and final renders, pausing at approval gates between stages. Reviewers can pin notes to video frames, moments, or script paragraphs; subsequent sessions must address each note before continuing, while approved checks and preferences can be carried across projects. It runs through Claude Code with the user's login or through an Anthropic API key, and provides both a terminal CLI and a local web studio. Built-in pipelines cover explainers, vertical social shorts, and logo animations; renders use Remotion and are checked with ffmpeg. The repository also documents narration through ElevenLabs, local Qwen3-TTS via ComfyUI, user-recorded voice, Strudel-based music generation, asset acquisition through a browser harness, and Blender-based 3D tooling. Users can define additional pipelines as folders containing configuration, craft instructions, and skills. Mortiflix is distributed under AGPL-3.0.

An open-source SDK and firmware project for building custom Muse gadgets from off-the-shelf ESP32 boards, Raspberry Pi devices, or other Linux boxes. The ESP32 and Linux device SDKs let developers connect displays, buttons, audio, sensors, actuators, and custom commands, then pair the gadgets with the Muse app on iOS or Android. Gadgets require an SDK token and developer mode in the Muse app before pairing. The project is licensed under Apache License 2.0, with stated exceptions for specified third-party files and dependencies that retain their upstream licenses.

Open Code Review is an open-source, AI-powered command-line code-review tool developed from Alibaba Group’s internal review assistant. It reads Git diffs and sends changed files to a configurable LLM through an agent with tool-use capabilities; the agent can read complete files, search the repository, and inspect other changed files for context before producing structured comments tied to source lines. Its hybrid architecture combines deterministic engineering with dynamic agent decisions: deterministic stages select and bundle files, match rules to file characteristics, and apply independent comment-positioning and reflection modules, while scenario-specific prompts and tools guide context retrieval and review. The `ocr scan` command audits complete files or directories without requiring a meaningful diff, and the project provides multilingual rules for issues including null-pointer exceptions, thread safety, cross-site scripting, and SQL injection. It supports configurable model endpoints, including OpenAI- and Anthropic-compatible endpoints, and requires Git 2.41 or newer.

openGym is a free, open-source gym and body-weight tracker with an Android app and Docker-based self-hosting. It provides guided workouts with previous weights, timed rest periods, personal-record notifications, weekly routines, supersets, cardio, timed sets, warm-up sets, and progression rules including linear progression, Greyskull LP, double progression, and added time. The tracker records workout history, body weight, RIR or RPE effort ratings, estimated one-rep maxes, and progress through charts, muscle-balance, fatigue, strength, and training-heatmap views. It can import histories from FitNotes, Strong, Hevy, and Apple Health exports, and export routines as files or PDFs. The project states that it requires no account, cloud service, or telemetry; its code, releases, container images, and APK builds are distributed under the AGPL-3.0 license.

Paperclip is an open-source, self-hosted control plane for organizing and running teams of AI agents. Its Node.js server and React UI let users define company goals, assign agents roles and tasks, and monitor work, budgets, approvals, costs, and audit activity from a shared dashboard. Agents run through scheduled or event-based heartbeats. The system maintains organization charts, goal ancestry, task dependencies, persistent session state, isolated workspaces, scoped secrets, runtime skill injection, and atomic task checkout with budget enforcement. It supports agent adapters and runtimes including Claude Code, Codex, Cursor, Bash, HTTP/webhook agents, and plugins; recurring routines can create tracked issues and wake assigned agents. Paperclip supports multiple isolated companies in one deployment, governance and approval gates, cost hard stops, agent pause/resume/termination, company export and import with secret scrubbing, and optional OpenTelemetry and Sentry integrations. It can be installed through the Paperclip CLI or run from source; local setups use embedded PostgreSQL and file storage, while production deployments can use an external PostgreSQL database. The project is distributed under the MIT license.

RE:Dox is an open-source, Apache-2.0-licensed structured-data engine for .NET, developed by CAPCOM as a core component of its REX technology. It parses JSON and other supported formats into a compact token DOM/intermediate representation in which fixed-size tokens can retain source offsets, lengths, links, or inline values; this representation supports reading, editing, serialization, deserialization, and format conversion. The mutable token model provides node-like insert, remove, and replace operations without rebuilding a managed object tree, while lazy decoding can defer materializing strings, numbers, timestamps, and binary values. Its asymmetric I/O design writes known values directly through a DataWriter, while deserialization first builds an addressable token DOM for look-ahead, out-of-order constructor binding, reference resolution, and optional automatic parallel processing of large arrays. The core packages support JSON, JSON5, CBOR, MessagePack, INI, and the binary DOX format, with TOML, XML, HTML, and CSV integrations listed as preview components. A common DataReader/DataWriter and DataConverter model enables conversion between formats such as JSON, CBOR, MessagePack, TOML, XML, and DOX, subject to structural projection where formats have different data models. JSON5 editing can preserve comments and other trivia, and JsonSequence supports asynchronous processing of NDJSON or large top-level JSON arrays without loading the entire input. Compatibility packages are provided for System.Text.Json, Newtonsoft.Json, and DataContractJsonSerializer. The project requires .NET 10 or later, is distributed through CAPCOM.REDox.* NuGet packages, and is under active development; its README notes that APIs, package boundaries, and preview-format support may change before the first stable public release.

SCSKiller is a free, open-source Windows tool that pre-compiles a game's shaders into the NVIDIA or AMD GPU driver's cache before play, reducing stutter caused by in-game shader compilation. It reads the shaders shipped with a game, determines the pipelines to build, and compiles them in a separate process while the game is closed; an optional recorder can capture pipelines encountered during play when the files do not provide them. It supports DirectX 12 games and DirectX 11 games on NVIDIA, detects games from several PC launchers or from user-added locations, and can recompile caches after driver updates. The tool does not modify game files during normal operation and leaves anti-cheat games alone apart from reading their files, with documented opt-in exceptions for offline sessions of Elden Ring and Armored Core VI. It is distributed for 64-bit Windows 10 and 11 under GPL-3.0-or-later, with an additional permission covering certain bundled components.

Strands Decider is a small decision model for agentic AI workflows, developed for use with the Strands Agents SDK. It selects among options, answers yes/no questions, or scores inputs against an ordered rubric, returning a calibrated confidence value for each decision. These capabilities support tasks such as model routing, tool selection, argument checking, triage, guardrail checks, and evaluation. The model uses a pretrained decoder LLM torso with its language-modeling head replaced by a pointer head. The head compares hidden states at the answer position and at each option's final token, scoring all options in one forward pass without text generation or a decoding loop. The same masked-softmax design supports its choice, noul, and score question types. The repository includes a command-line interface, a local HTTP server, an example that gates a Strands agent's tool call, and optional image inputs through a vision mode. It can run on CUDA, Apple silicon through MLX or MPS, and CPU. The project is distributed under the Apache License 2.0 and can be installed with pip.

VibeWise is a Claude Code plugin that supports learning-oriented software development while Claude writes the code. It asks the developer to propose an approach, works through architecture, tradeoffs, failure cases, and unfamiliar concepts, and uses Build, Design, and Implementation checkpoints before writing code. The developer confirms or discusses each decision; Claude then implements the agreed step, runs tests, and reports what changed, how it works, and which checks passed. Teaching support can be adjusted for beginner, intermediate, or advanced users, with configurable checkpoint frequency and project-specific learning notes and maps stored in `.vibe-wise/`. The plugin requires Claude Code and Python 3, uses no additional Python packages, has no extra account, backend, or telemetry, and is released under the MIT license.

WeKnora is an open-source, LLM-powered knowledge platform from Tencent for turning documents into queryable knowledge bases, retrieval-augmented Q&A, autonomous reasoning workflows, and self-maintaining Markdown wikis. It supports RAG-based quick Q&A and a ReAct agent that orchestrates retrieval, MCP tools, sandbox skills, and web search for multi-step tasks. Its Wiki Mode distills source documents into structured, interlinked Markdown pages with an interactive knowledge graph, manual editing, revision history, line-level diffs, and rollback. The platform ingests formats including PDF, Word, Markdown, HTML, images, spreadsheets, presentations, and XMind, and can synchronize sources such as Feishu, GitLab, Tencent IMA, Notion, Yuque, and RSS. Its modular pipeline supports interchangeable parsers, LLMs, embedding providers, vector databases, and storage backends, with dense, sparse, hybrid, parent-child, and graph-based retrieval strategies. WeKnora provides a web interface, REST API, command-line client, MCP server, website embed widget, and integrations with messaging channels. It can be deployed locally, with Docker, or on Kubernetes, including private and offline deployments. The repository states that it is licensed under the MIT License and supports workspace RBAC, scoped API keys, audit logs, and Langfuse-based observability.

Whirl is a full-stack, self-hostable AI chat application developed by Anterra and published under the MIT license. Built with Next.js and Convex, it routes conversations through multiple models using OpenRouter, allowing users to switch models and adjust thinking levels within one conversation. It provides long-term memory for preferences and projects, live web search and page reading, MCP integrations with OAuth, installable skills, voice input, image generation, file attachments, and editable living artifacts such as documents, charts, and interactive pages that can be shared by link. It also includes encrypted locked chats, incognito mode, message queueing, folders, conversation sharing, and an installable mobile web app.