AutoPiP is a Safari Web Extension paired with a small Cocoa host app for macOS 13.5+. In production, it automatically enters Picture-in-Picture mode for HTML5 video when the user switches tabs (visibilitychange), switches windows (blur/focus with a 100ms delay that filters internal focus changes like clicking YouTube live chat), or scrolls a YouTube video out of view (IntersectionObserver with a 10% threshold). The toolbar popup provides a global on/off toggle, per-feature switches, blacklist/whitelist site control with full-hostname or root-domain matching, and a configurable keyboard shortcut (default ⌥P) that prefers the W3C requestPictureInPicture API and falls back to webkitSetPresentationMode. The extension is Manifest V2, which is the only option Safari currently supports, and it makes no outbound network requests beyond Sparkle update checks performed by the host app. Architecturally, the product is split into three layers: the content script (content.js) owns PiP state transitions and video discovery via priority-ordered selectors (YouTube, Disney+, Twitch, generic) with a MutationObserver-invalidated cache; the popup (popup.js) persists settings in browser.storage.local and broadcasts changes to all open tabs via runtime messages; and the Swift host app provides a first-launch onboarding flow in a WKWebView, bridges update preferences through a native message handler, and wraps Sparkle 2.9.3 with an opt-in beta channel. The release pipeline is the standout feature: semver.txt is the single source of truth, feature/* branches produce immutable beta tags, main produces stable releases, and a canonical Sparkle appcast plus Homebrew cask are updated automatically via Python scripts with atomic file replacement. The repository is actively maintained (last commit September 2026, beta release v2.1.1 in progress) with a well-instrumented CI setup: dependency-free Node.js tests execute the content script, popup, and host script inside a vm sandbox with fake DOM objects; Swift unit tests cover onboarding JSON parsing; and release-pipeline tests validate appcast/cask transformations and workflow invariants. Documentation is unusually thorough (AGENTS.md, BUILD.md, PRIVACY.md, SECURITY.md). Minor weaknesses: the UI test targets contain only Xcode template boilerplate, popup.js uses unconditional console.log in a few places, and storage keys/message names are duplicated string contracts between popup.js and content.js rather than shared constants.
Bitcore is a comprehensive, production-grade monorepo for building multi-blockchain wallet infrastructure and blockchain data services. It provides a full stack including core cryptographic libraries (bitcore-lib for BTC, BCH, LTC, DOGE), wallet management (bitcore-wallet-service), client SDKs (bitcore-wallet-client), a blockchain indexer and API node (bitcore-node), a CLI wallet, and a React-based block explorer (insight). The architecture is modular with each cryptocurrency having its own package, but shares significant code duplication across coin-specific libraries. The system supports UTXO-based chains (BTC, LTC, DOGE, BCH) and account-based chains (EVM, Solana, XRP), including threshold signature schemes (TSS) and multi-sig wallets. External API adapters (Alchemy, Moralis) and fiat on-ramp integrations (Wyre, Banxa, Simplex, Transak) provide production features. The codebase is written primarily in TypeScript and JavaScript, with extensive test coverage and operational scripts. While functionality is rich, technical drift is medium due to duplicated code across coin libraries, outdated dependencies (sjcl, request, Gulp), and some deprecated patterns.
This repository is a Serverless Framework plugin that enables deploying Python WSGI applications (Flask, Django, Pyramid, etc.) to AWS Lambda via API Gateway or ALB. It automatically converts API Gateway events (v1, v2) and ALB requests into WSGI environ dictionaries, invokes the application, and transforms the response back into the appropriate Lambda proxy format. The plugin also provides a local development server (`sls wsgi serve`), automated dependency packaging from `requirements.txt`, and remote execution commands for Python code, shell commands, Django management commands, and Flask CLI commands. The codebase is split into a Node.js plugin (`index.js`) that hooks into Serverless lifecycle events (packaging, deployment, invoke local) and a set of Python modules (`serverless_wsgi.py`, `wsgi_handler.py`, `serve.py`, `requirements.py`) that handle the actual WSGI conversion, local serving, and dependency installation. The core WSGI adapter (`serverless_wsgi.py`) is a standalone, reusable library that can be used independently of the Serverless framework for any Lambda function needing to handle API Gateway events. The plugin is well-tested with both JavaScript and Python test suites, and follows a clean, modular architecture.
Lossless Claw is a production-grade OpenClaw plugin for lossless context management in LLM conversations. It persists every message in SQLite, compresses older context into a DAG of hierarchical summaries (leaf and condensed), and assembles the active prompt by combining recent messages with summaries. The system provides recall tools (grep, describe, expand) for agents to retrieve details from compacted history, a doctor tool for repairing summary quality, a CLI for diagnostics, and a Go-based TUI for interactive inspection and maintenance. It integrates deeply with OpenClaw's plugin API, using the host's LLM dispatch for summarization and supporting multi-provider fallback, OAuth, and retry chains. The architecture is event-driven with deferred compaction, pending summary pipelines, and atomic publication. While the core is TypeScript, the TUI is written in Go, creating a bimodal maintenance burden. The codebase demonstrates strong engineering practices with extensive test coverage, but its tight coupling to OpenClaw's evolving API and large configuration surface introduce moderate technical drift risk.
Nitric is a multi-cloud framework that enables developers to define cloud infrastructure and application logic in code, then deploy to AWS, Azure, or GCP. The framework provides infrastructure-as-code capabilities for cloud resources (APIs, storage, queues, topics, secrets, etc.) with runtime implementations for handling cloud service events. It features a modular architecture with core framework logic, cloud provider implementations, and a documentation site. The system uses a plugin-based worker system for handling cloud events and provides consistent abstractions across cloud providers through well-defined interfaces.
Thonburian Whisper is a Thai Automatic Speech Recognition (ASR) model fine-tuned from OpenAI's Whisper, provided as a reusable application with associated tools. The repository offers the pre-trained model hosted on Hugging Face, along with utilities for transcription, long-form audio processing, TensorRT conversion, and evaluation. Key components include a CLI tool for long-form transcription using VAD and syllable segmentation, scripts for converting the model to TensorRT format for efficient inference, and evaluation metrics calculation. The architecture consists of modular Python scripts and notebooks that demonstrate usage rather than a monolithic application, with clear separation between model usage, audio preprocessing, and post-processing.
This repository provides a collection of custom nodes for ComfyUI and SD WebUI (reForge) that implement various guidance techniques for diffusion models. It includes implementations of Perturbed-Attention Guidance (PAG), Smoothed Energy Guidance (SEG), Sliding Window Guidance (SWG), PLADIS, Normalized Attention Guidance (NAG), Token Perturbation Guidance (TPG), Frequency-Decoupled Guidance (FDG), Momentum Guidance (MG), and Sliding Mode Control CFG (SMC-CFG). The architecture consists of modular node files, shared utility functions, and backend compatibility layers. Key entry points are the ComfyUI extension registration in __init__.py and the reForge integration scripts in the scripts directory. The system modifies the denoising process of diffusion models by injecting guidance computations at specific U-Net blocks, enhancing image quality and control.
This repository is a structured, comprehensive AI engineering curriculum spanning 20 phases and over 500 lessons, from math foundations and ML fundamentals to deep learning, computer vision, NLP, speech, transformers, generative AI, reinforcement learning, LLMs from scratch, agent engineering, multi-agent systems, production infrastructure, and ethics/safety. It is an educational resource, not a production application or data research project. The code consists of self-contained pedagogical implementations (primarily Python, with some TypeScript, Rust, and Julia) that demonstrate core algorithms from scratch—such as attention mechanisms, BPE tokenization, diffusion models, RLHF, and agent loops—using synthetic data and deterministic stubs. Capstone projects include small web servers (Hono/TypeScript) and test harnesses. The curriculum is actively maintained with CI tooling, certification prep materials, and translation support. While the code is simplified for learning, it exhibits clean patterns, consistent lesson structure, and testing practices that make individual components (e.g., tool registries, state machines, evaluation harnesses) reusable in other contexts.
This is a NestJS boilerplate for building RESTful APIs with TypeScript, PostgreSQL, TypeORM, and modern DevOps practices. The system provides a modular backend architecture with authentication (JWT), user management, post management, email verification via BullMQ queues, health checks, Swagger API documentation, and comprehensive testing setup. It follows NestJS best practices with clear separation of concerns across modules for auth, users, posts, health, and background jobs. Key features include cursor-based and offset-based pagination, internationalization (i18n), Redis caching, security headers via Helmet, and environment-based configuration management. The codebase demonstrates strong engineering practices with reusable utilities for password handling, configuration validation, and sophisticated cursor pagination that could be extracted to other projects with minimal modification.
This repository implements the Ghost CLI, a command-line tool for installing, configuring, updating, and managing Ghost CMS instances on Linux servers (Ubuntu). The system is a monolithic Node.js application with a plugin architecture. Its core responsibilities include: (1) resolving and downloading Ghost versions from the npm registry, (2) setting up required infrastructure (MySQL, nginx, systemd), (3) managing instance lifecycle (start/stop/restart), (4) handling content migration and import/export via the Ghost Admin API, (5) running health checks via the doctor command, and (6) handling major version upgrades with rollback support. Architecturally, the CLI is composed of a Command base class with yargs integration, a layered task system using listr2 for sequential/parallel operations, and an extension system that allows built-in plugins (nginx, systemd, MySQL) to inject commands, setup steps, doctor checks, and custom process managers. Configuration is managed via JSON files, and dependencies are installed using either yarn or pnpm. The codebase is well-structured with clear separation of concerns: commands in lib/commands/, tasks in lib/tasks/, utilities in lib/utils/, and extensions in extensions/. The tool depends on external services: npm registry for version resolution, Ghost Admin API for import/export, system tools (nginx, systemd, MySQL) for production setups, and acme.sh for SSL certificate management. It supports local development (direct node process) and production (systemd service). The codebase is actively maintained, with support for Ghost versions 1.x through 6.x, Node.js 22/24, and MySQL 5.6 through 9.4. While there is some legacy code for older Ghost versions, the overall architecture is modern, testable, and modular.
This repository is a collection of deployable templates and educational cookbooks for building real-time LLM applications using the Pathway framework. It provides ready-to-deploy RAG (Retrieval-Augmented Generation) solutions with live data synchronization from various sources (files, Google Drive, SharePoint, etc.), built-in vector and hybrid search, and integration with LangGraph for agentic workflows. The templates are production-oriented with Docker support, while the cookbooks demonstrate advanced patterns like Self-RAG with self-reflection and query rewriting. The codebase is well-structured but exhibits some duplication across templates, and the core logic relies heavily on the external Pathway library.
UZDoom is a modern, feature-rich source port for Doom-engine games, forked from GZDoom. It is a monolithic game engine application that provides software and hardware-accelerated rendering (OpenGL, Vulkan, GLES), a full audio subsystem (ZMusic) supporting MIDI, tracker modules, and chip emulation, a custom GUI toolkit (ZWidget) for the launcher and in-game menus, a scripting engine (ZScript/ACS) with a Debug Adapter Protocol (DAP) server, and a resource filesystem. The codebase is large and complex, with extensive third-party library vendoring (FreeType, HarfBuzz, Abseil, asmjit, etc.) and multiple rendering backends. It is actively maintained with recent commits and CI pipelines. While the architecture is well-structured with clear subsystem boundaries, the sheer volume of vendored code and platform-specific implementations introduces medium technical drift. Several components (ZMusic, ZVulkan, ZWidget, DAP debug server) are self-contained and could be extracted for reuse in other projects.
Winapp2 is a community-maintained database of cleaning rules for Windows system cleaners (CCleaner, BleachBit, System Ninja, Tron, FluentCleaner) paired with a VB.NET toolchain, winapp2ool, that generates, lints, trims, diffs, combines, and flavor-transforms the database. The repository is a hybrid: the primary artifact is the declarative winapp2.ini data, but the dominant engineering effort is the software pipeline that keeps that data consistent and distributable. It is not an end-user cleaner; it is the source-of-truth repository and build system for the cleaning rules. Architecturally, winapp2ool is a Windows-only .NET Framework console application with both interactive menus and scriptable CLI entry points. Modules such as EntryBuilder, UWPBuilder, and BrowserBuilder consume concise DSL source files (alphabetically split INI catalogs plus shared scaffold templates) and generate standardized winapp2 entries. Transmute/Flavorizer then applies per-tool corrections in a fixed six-stage pipeline, Combine merges artifacts, WinappDebug normalizes and lints, and Diff produces changelogs. The full build is orchestrated by a PowerShell script and enforced by GitHub Actions workflows that byte-compare generated artifacts against committed outputs. The codebase shows strong engineering discipline: unit tests cover key components, CI guards artifact drift and binary parity, and several modules are cleanly separated behind reusable abstractions. Technical drift is moderate because the entire toolchain is tied to .NET Framework, Visual Basic .NET, Windows registry access, and the winapp2/CCleaner data model; these are stable but aging dependencies. The committed generated artifacts and custom DSLs are the main long-term maintenance risks, but the project is actively curated and the pipeline is deliberately designed to catch silent corruption.
This repository is a collection of Odoo modules extending financial reporting capabilities in the Odoo ERP system. It provides standardized financial reports (aged partner balance, general ledger, trial balance, VAT report, etc.), partner statements (activity, outstanding, detailed), tax balance reports, cash flow forecasting, and export functionality in multiple formats (PDF, HTML, XLSX). The architecture follows Odoo's modular pattern with models, views, reports, wizards, and data files. Key entry points are wizard-driven interfaces that collect user parameters and generate reports via QWeb templates or XLSX generators. The system relies on Odoo's core accounting models and reporting infrastructure, with data flowing from user input through ORM queries to report presentation. The codebase demonstrates good reuse through abstract base classes for reports and wizards, reducing duplication across report types.
Meteor is an ultra-simple full-stack JavaScript framework for building modern web and mobile applications. This monorepo contains the Meteor CLI tool (`meteor`), the build system (isobuild, Rspack integration), core runtime packages (reactive data via DDP, MongoDB integration, accounts, Blaze templating, modules, bundling), and a rich ecosystem of OAuth providers, testing infrastructure, and developer tools (ESLint plugin, Babel preset, installer). The codebase is organized as a monorepo with `tools/` for the CLI and build system, `packages/` for framework packages, and `npm-packages/` for standalone npm packages. It demonstrates a real-time data synchronization pattern using DDP (Distributed Data Protocol), a custom constraint solver for package dependency resolution, and a compiler plugin system for transpilation. Despite being a well-established framework with extensive testing, the codebase shows signs of technical drift: deprecated packages remain in-tree, there is a dual sync/async API surface due to ongoing migration from Fibers to async/await, and the build system is transitioning from a legacy compiler to Rspack.
This repository provides `ffmpeg-next`, a Rust library that offers safe, idiomatic wrappers around FFmpeg's C libraries (libavformat, libavcodec, libavutil, libswscale, libswresample). It is a maintained fork of the abandoned `ffmpeg` crate, designed to decode, encode, mux, demux, filter, scale, and resample video and audio from Rust. The architecture is modular, with dedicated modules for codec (packets, frames, encoders, decoders), format (input/output contexts, streams, chapters), filter graphs, device I/O, and software scaling/resampling. Key entry points include `format::input` and `format::output` for opening media files, and the decoder/encoder APIs (`send_packet`, `receive_frame`, etc.). The crate supports FFmpeg versions 3.4 through 8.0 via extensive conditional compilation, and includes advanced features like custom stream I/O, interrupt callbacks, and type-safe audio plane access. It is primarily a library for integration into Rust multimedia applications, not a standalone tool.
This repository contains the public-facing source code for Draw Things, a production-grade AI image and video generation application for macOS/iOS/Linux. The codebase implements a comprehensive suite of diffusion models (Stable Diffusion, FLUX, HiDream, Qwen, MiniMax, etc.), samplers, LoRA training, model conversion, and a gRPC-based distributed inference server. It is written primarily in Swift, using Bazel for builds, and includes extensive use of FlatBuffers for data serialization, SQLite for persistence, and a custom neural network compiler (NNC) for on-device inference. The architecture is modular, with separate libraries for tokenization (CLIP, Tiktoken, SentencePiece), model zoos (downloadable model specifications), image generation pipelines, and cloud authentication. The repository also includes CLI tools (draw-things-cli, gRPCServerCLI, model converters) and server management scripts for deploying GPU-backed inference services with Tailscale, Envoy, and Docker. While the core application is a consumer-facing app, the codebase also contains research-grade components such as instruction count estimators for performance profiling, formal verification tools (Z3, Lean), and web search integration with browser fallback. The code is well-structured with many reusable patterns, though it carries some maintenance overhead from vendored dependencies and generated code.
MoneyPrinterTurbo is a production short-video generation platform. Given a topic or script, it orchestrates LLM-based script/keyword generation, multi-provider TTS (Edge, Azure, Gemini, ElevenLabs, etc.), video material sourcing (Pexels, Pixabay, Coverr, several AI video generators), subtitle generation (Whisper or TTS timestamps), background music, and final video composition via moviepy/FFmpeg. It exposes three consistent interfaces: a FastAPI REST API, a Streamlit WebUI, and a CLI. All three share the same underlying task pipeline in `app/services/task.py`, which manages state, progress, and cross-posting to YouTube/TikTok/Instagram. The architecture is a service-oriented monolith: the core pipeline is a linear sequence of stages (script → terms → audio → subtitles → materials → video → optional BGM/cross-post). External integrations are abstracted behind provider-specific modules (e.g., `voice.py`, `material.py`, `bgm.py`). A Redis-backed or in-memory state manager tracks task progress. The configuration is runtime-editable and stored in a TOML file. The WebUI is a large Streamlit single-page application with dozens of provider-specific controls, billing confirmations, and i18n support for 12+ languages. The codebase demonstrates strong engineering practices: extensive test coverage (200+ test files), structured error handling with per-provider error classes, atomic file writes, path traversal protection, and security-conscious API key handling. However, it also shows signs of rapid growth: the WebUI is a monolithic 1300+ line script, provider integration logic is occasionally duplicated, and the dependency list is large (many third-party APIs). The repository is actively maintained with recent commits and good issue triage. It is not a research project; there are no datasets, model training, or analytical queries.
chubbyskills is a personal knowledge management CLI toolkit that ingests content from over a dozen platforms (YouTube, Bilibili, X, Xiaohongshu, WeChat, Douyin, Podcasts, etc.), transcribes audio/video locally or via cloud providers, and outputs structured Markdown notes with YAML frontmatter (schema v1). The system is architected as a collection of modular "skills" (one per platform) orchestrated by a unified CLI (`tools/chubby.py`), with shared utilities in `chubby_common/`. It includes an incremental knowledge base indexer using SQLite FTS5 and optional semantic embeddings, an MCP server exposing six tools for AI agent integration, and a deterministic evidence brief generator with integrity validation. Operation is local-first: transcription uses SenseVoice and faster-whisper, video downloading uses yt-dlp, and all processing is designed to work offline. Cloud transcription (Atlas/MuAPI) and LLM enrichment (DeepSeek) are optional extensions. The repository also provides a skill installer, platform smoke testing, output validation, and retry mechanisms with credential redaction. Despite the breadth, the codebase maintains clean separation of concerns, with the shared library standing out as a reusable zero-dependency core.
A Visual Studio extension (VSIX) that integrates SonarQube/SonarCloud static analysis into the IDE. It provides real-time code quality and security analysis for C#, VB.NET, C/C++, JavaScript/TypeScript, CSS, and other languages via an embedded SLCore backend (Java process) and a separate Roslyn analyzer server for .NET. The extension supports both standalone analysis and 'connected mode' that binds to SonarQube/SonarCloud servers, including project binding, credential management, issue visualization with flows and locations, security hotspots, taint vulnerabilities, dependency risks, quick fixes, rule education (HTML-to-XAML rendering), and telemetry. The architecture is modular, heavily relying on MEF for dependency injection, with distinct assemblies for core abstractions, VS integration, connected mode UI, issue visualization, security-specific features, and education. The codebase is well-tested with extensive unit tests and some integration tests, but shows signs of ongoing migration from legacy binding models to a modern SLCore-based protocol, resulting in some technical debt and backward-compatibility code.