MicroBin is a self-hosted pastebin and URL shortener implemented as a single Rust binary using Actix-web. In production it accepts text pastes and multi-file uploads via multipart POST /upload, serves them via /upload/{id} (HTML), /raw/{id} (plain text), /file/{id} (streamed via actix-files), /url/{id} (HTTP redirect for URL pastes), and /p/{id}, /u/{id} short-path aliases. It supports server-side encryption (magic-crypt), client-side AES-CTR encryption in the browser, QR code generation, read-count tracking, expiry and burn-after-read deletion, ZIP archive download of all attachments, a password-protected admin panel, and either SQLite (default) or JSON file persistence. All state is held in a global `Mutex<Vec<Pasta>>` and every request performs a linear scan of that vector, which is simple but becomes a performance ceiling as uploads accumulate. IDs are random `u16` values (at most 65,536 unique), so any moderately used instance is vulnerable to collisions; a collision causes the SQLite insert to panic the request thread. The JSON storage path is hard-coded, and `remove_expired` passes `None` to `delete()` which panics in JSON-DB mode. Client-side encryption uses AES-CTR with the password padded to 32 bytes as a fixed key and the library-default counter (no salt, no per-message nonce), which is cryptographically weak. Telemetry and update checks phone home to api.microbin.eu by default but can be disabled via flags.
AutoTrain Advanced is a no-code / low-code machine learning training platform from Hugging Face, providing a unified CLI, web UI (FastAPI + Jinja2), and REST API to fine-tune models on 15+ tasks including LLM SFT/DPO/ORPO, text/image classification, object detection, token classification, seq2seq, sentence transformers, and VLM. It wraps the Hugging Face ecosystem (transformers, datasets, peft, accelerate, trl) and supports multiple backends: local GPU, Hugging Face Spaces, NVIDIA NGC, and NVCF. The project is officially deprecated as of late 2024, with the README explicitly stating no new features will be added and recommending alternatives. The codebase is architecturally well-organized with a modular trainer structure, Pydantic parameter models, abstract backend patterns, and reusable callbacks, but suffers from duplicated dataset loading logic, heavy coupling to Hugging Face API, and a sprawling single-file UI. It remains functional for users who need a turnkey fine-tuning solution with existing dependencies, but is not suitable for new projects requiring long-term maintenance.
Polynote is a polyglot notebook server that enables interactive data analysis in Scala, Python, SQL, and Vega. It provides a web-based IDE with Monaco editor, real-time kernel execution, rich visualizations (Vega-Lite, Matplotlib), and tight Apache Spark integration. The architecture is a monolith: a Scala server using ZIO effects, a custom binary WebSocket protocol, and a TypeScript frontend with its own reactive state management. Key entry points are the server main class (Main.scala) and the headless notebook runner (NotebookRunner.scala). Data flows from cell edits through WebSocket messages to the kernel, which executes code and streams results back. External dependencies include JVM, Python via Jep, Spark, Coursier for dependency resolution, and various frontend libraries. While primarily an application, it also serves as a data exploration platform with streaming data abstractions and visualization capabilities.
This repository is a large model zoo (500+ models) focused on converting and quantizing pre-trained computer vision models for edge deployment across multiple frameworks (TFLite, ONNX, OpenVINO, CoreML, TensorRT). It provides hundreds of download scripts for pre-converted model files, conversion pipelines (PyTorch/ TensorFlow to ONNX/TFLite), and a modest set of inference demo scripts for real-time video processing. The repo is primarily a research/engineering resource for edge AI practitioners, not a software application. It exhibits heavy duplication across model directories, relies on an external S3 bucket for model distribution, and uses older framework versions (TensorFlow 1.x/2.x, ONNX opset 11-13). Despite these maintenance concerns, the systematic conversion workflow and post-processing ONNX subgraph generators (e.g., NMS, box decoding) represent reusable patterns for anyone building end-to-end inference pipelines. The repo is hybrid in nature: while classified as data_research due to its focus on pre-trained model artifacts and lack of original training code, it also contains application-like demo code and CLI entry points.
This repository is a comprehensive collection of ad-blocking filter lists, CSS styles, JavaScript scripts, and configuration files for proxy tools (Surge, QuantumultX, Shadowrocket, Clash) and browser extensions (uBlock Origin Lite, Tampermonkey). It targets both international and Chinese websites, with a strong focus on streaming, adult, and content farm sites. The core functionality is delivered through platform-specific rule lists (`.list`, `.conf`, `.sgmodule`, `.yaml`) and a set of userscripts that inject CSS/JS into web pages via MITM proxy or browser script managers. The repository also includes a large set of pre-compiled uBlock Origin Lite scriptlets that implement advanced ad-blocking techniques such as cookie removal, attribute removal, eval blocking, and timer manipulation. The project is community-driven, with issue templates for reporting new ad locations. While primarily a configuration/rule collection, it contains substantial JavaScript code that demonstrates client-side ad-blocking patterns.
This repository contains a suite of 152 installable Claude skills (76 builder and 76 checklist-reviewer pairs) that automate the generation and auditing of structured Excel workbooks for automotive engineering domains including functional safety (ISO 26262), cybersecurity (ISO/SAE 21434), SOTIF (ISO 21448), quality (IATF 16949, ASPICE), AUTOSAR, diagnostics (UDS), and systems engineering (SysML, MBSE). Each skill is a self-contained archive with a SKILL.md describing trigger phrases, Python scripts for workbook generation (using openpyxl), and reviewers that probe and validate outputs against checklists. The repository also includes supporting automation (chain-contract auditing, status regeneration) and extensive documentation of polish passes and weekly planning. The architecture is a collection of independent skill archives rather than a monolithic application. While the concept is robust and domain coverage is broad, the codebase exhibits significant quality issues: silent data corruption from column shifts and truncated files, incomplete check implementations in many reviewers, duplicated utility scripts, and a pattern of bugs surviving months before detection. The suite is actively maintained with weekly polish cycles.
This repository is a full-stack web application, 'BypassAIGC', functioning as an AI-powered academic writing assistant. Its core operational logic, confirmed by code analysis of its two-stage 'polish' and 'enhance' pipeline and configuration flags, is to rewrite text to evade detection by AIGC classifiers like GPTZero. It is a monolithic application built with a FastAPI backend and a React frontend, packaged for standalone deployment via PyInstaller. The codebase features a highly sophisticated, deterministic Word document formatting engine (`word_formatter`) that parses markdown into a Document AST, renders it to OOXML, validates the structure against a StyleSpec, and applies patch-based auto-corrections. This module demonstrates exceptional craftsmanship in document engineering. While functionally complete, the application suffers from significant technical debt. This includes critical security vulnerabilities (default secrets, permissive CORS, plaintext API key storage), ad-hoc database migrations, duplicated application entry points, and incomplete features (e.g., PDF export is a stub). The project is maintained but carries a high risk of technical drift, making it unsuitable for production deployment without substantial hardening.
This repository contains the ESP-Drone firmware, an open-source flight control system for quadcopters based on Espressif ESP32/ESP32-S2/ESP32-S3 microcontrollers. It is a port of the Bitcraze Crazyflie firmware (GPLv3) and is primarily designed for STEAM education. The system provides a complete drone control stack including sensor drivers (IMU, barometer, magnetometer, ToF range finders, optical flow), state estimation (complementary filter, extended Kalman filter), attitude/position controllers (PID, Mellinger, INDI), and communication via Wi-Fi (UDP, ESP-NOW) and CRTP protocol. It supports modular expansion 'decks' for additional hardware like flow sensors, thermal cameras, and localization systems (Lighthouse V1/V2, UWB TDoA). The firmware runs on FreeRTOS and uses ESP-IDF v5.0 as its development framework.
Evil Limiter is a Linux command-line tool that uses ARP spoofing and traffic shaping (tc/iptables) to monitor, analyze, and limit the bandwidth of devices on a local network without requiring administrative access to those devices. It provides an interactive shell with commands to scan for hosts, limit or block bandwidth per host (upload/download), monitor real-time bandwidth usage, and detect host IP changes. The architecture is a single-threaded interactive menu with background threads for ARP spoofing, bandwidth monitoring, and reconnection watching. It depends on scapy for packet manipulation, netifaces for network interface info, and standard Linux utilities (tc, iptables, sysctl). The codebase is well-organized into modules for console UI, networking operations, and command parsing, but lacks tests and has limited reusability outside its specific purpose.
This repository is a curated collection of Jupyter notebooks and Python scripts that demonstrate fine-tuning, inference, quantization, and optimization techniques for state-of-the-art vision and multimodal models from the Hugging Face ecosystem. It covers a wide range of tasks including visual question answering, object detection, image classification, video understanding, multimodal retrieval, and preference alignment (DPO). The code is primarily educational, designed to be run interactively in Google Colab, with each notebook or script serving as a self-contained tutorial for a specific model or technique. Notable patterns include LoRA/QLoRA fine-tuning, knowledge distillation via a custom ImageDistilTrainer, any-to-any embedding for retrieval, and video frame downsampling for multimodal input. The repository is not a production library or application; it is a learning resource for engineers and researchers exploring vision-language models.
This repository is a curated 'awesome list' of third-party services for static websites. It contains no application code, build system, or tests—the entire content is a single markdown file (readme.md) organized into categories such as audio, images, maps, comments, forms, live chat, e-commerce, payments, search, analytics, authentication, and FaaS. The list is maintained through community pull requests following contributing.md guidelines. It functions as a decision-support directory for engineers building static or JAMstack sites, pointing to hosted services and embeddable widgets rather than providing implementations. The repo's value is in curation and categorization, not software craftsmanship. Some listed services are outdated or dead (e.g., LiveFyre, Zopim, Vevo), so users should verify current availability before relying on an entry. Maintenance is active (last commit 2026-04-20), and the list continues to receive additions. Technical drift is medium because the content references many third-party services whose status changes over time, but the repository itself has no dependencies to rot.
This repository is a curated awesome-list of research papers, projects, datasets, and industry demos related to digital humans: 3D avatar generation and animation, clothed human reconstruction, head avatars, human motion generation, virtual try-on, cloth simulation, and related topics. It contains only a README.md plus LICENSE; there is no application code, no executable pipeline, no tests, no CI, and no dependency manifests. Despite the data_research classification, it is best understood as a manually maintained literature index rather than a data pipeline or software project.
Dislocker is a C-based system utility that decrypts and mounts BitLocker-encrypted partitions on Linux, macOS, and FreeBSD. It provides read/write access to BitLocker volumes (Vista/7/8/8.1/10) via FUSE (virtual NTFS mount) or flat-file decryption. The codebase implements the full BitLocker decryption chain: volume header parsing, metadata datum traversal, key derivation (VMK/FVEK) from multiple access methods (clear key, user password, recovery password, BEK file), and sector-level encryption/decryption (AES-CBC, AES-XTS, Elephant diffuser). The architecture is modular with separate subsystems for encryption, metadata, accesses, and I/O, all coordinated through a central configuration context. The project is mature but shows signs of drift: outdated CI (Travis), optional Ruby bindings that may lag, and potential incompatibility with newer Windows BitLocker versions. Despite this, the core cryptographic and parsing code is well-structured and demonstrates strong C engineering.
Awesome Video is a curated resource catalog for streaming video technologies, maintained as structured JSON (contents.json) and rendered into both a README and a static documentation site (MkDocs Material) deployed to awesome.video. The catalog spans ~50 hierarchical categories covering HLS, DASH, FFmpeg, codecs (HEVC/VP9/AV1), transport protocols (RIST/SRT/RTMP), players, DRM, QoE, testing, CDN, HDR, subtitles, and learning resources. It is not a runtime application — it is a content curation pipeline: contents.json is the source of truth, convert.rb generates the README, database.json feeds the site's lunr search index, and loader.py imports bookmarks from Raindrop.io. The repo is hybrid in nature: primarily a curated data catalog with a static-site build system around it, rather than a software application or research pipeline. The build pipeline (GitHub Actions) validates the JSON against schema.json, checks external links with awesome_bot, and deploys to GitHub Pages. The curation quality is strong — entries include authoritative references (Netflix TechBlog, DASH-IF specs, Apple WWDC sessions, Xiph test media) — but the supporting code is simple utility scripting with notable weaknesses: hardcoded Raindrop.io credentials in loader.py, inconsistent category identifiers and malformed tag entries in the data, and aging CI dependencies (Python 3.7, jinja2 3.0.3). The repo is actively maintained and the data model (hierarchical categories + multi-tagged project records) is a clean, portable schema for any awesome-list-style content site.
This repository is a curated knowledge base and learning guide for embodied AI and robotics. It aggregates documentation, paper summaries, interview notes, and learning roadmaps rather than providing executable software. The content is organized into sections covering foundations, classical methods, state-of-the-art techniques, simulation tools, and real-world robotics. Helper scripts for PDF processing support the documentation workflow by converting research papers into text for summarization. The repository serves as an educational resource and reference guide for practitioners and researchers in the field.
This is a Chinese-language educational companion to DeepLearning.AI's 'Agentic AI' course by Andrew Ng, maintained by the Datawhale community. The repository is a hybrid learning resource: ~60 translated markdown files across 5 modules (agentic workflow intro, reflection design patterns, tool use, practical building tips, highly autonomous agents) plus executable Jupyter notebooks and supporting Python modules that implement the lab exercises. It is not a production application or service — it is courseware with real, runnable agent code demonstrating current LLM agent patterns.
Firemote is a Home Assistant custom dashboard card that emulates physical remote controls for a wide range of media devices including Amazon Fire TV, Apple TV, Chromecast, NVIDIA Shield, Roku, Xiaomi, and more. The card provides realistic visual remotes with extensive customization: over 330 app launcher shortcuts, button overrides via Home Assistant scripts or actions, multiple remote styles, and support for device-specific compatibility modes. It integrates with Home Assistant's Android Debug Bridge, Android TV Remote, Apple TV, and Roku integrations. The repository contains the compiled JavaScript card and supporting data files for device definitions and translations, but the source is distributed as a single minified bundle. It is actively maintained with a large community and frequent updates.
This repository is a curated 'awesome list' of resources related to status badges, maintained by the Shields core team. It is not a software application or data pipeline; it is a markdown file (README.md) that catalogs dynamic badge services, badge tools, static badge collections, badge-rendering software, icons, academic articles, and more. The repository includes configuration for automated link checking via GitHub Actions and Dependabot for dependency updates. While it contains no application code, it serves as a community-maintained directory for badge-related projects and is actively kept up-to-date with regular link checks and contributions.
This repository implements a hierarchical controller for quadruped locomotion that combines a high-level reinforcement learning policy with a low-level model predictive controller (MPC). The RL policy dynamically predicts weight parameters for the MPC, which computes optimal ground reaction forces and joint torques. The codebase includes Isaac Gym training environments for the Unitree A1, Go1, and Aliengo robots, along with a custom C++ MPC solver (OSQP-based) and a finite state machine for robust transient handling. While primarily a research project that bridges RL and MPC for legged robots, it also provides reusable robotics components such as state estimators, leg controllers, gait pattern generators, and foot swing trajectory planners. The work is based on the MIT Cheetah software and includes detailed development logs and configuration files for reproducible experiments.
Procursus is a cross-compilation build system and Debian package repository infrastructure targeting jailbroken iOS and macOS devices. It does not ship a final application itself, but instead provides a comprehensive framework to compile, patch, and pack hundreds of open-source tools (e.g., bash, git, vim, ffmpeg, openssh) into .deb packages suitable for installation on Apple platforms. The architecture is a monolithic Makefile orchestration with per-package .mk recipes that define download, extract, patch, build, and package steps. A large collection of custom patches in build_patch/ adapts upstream sources for Darwin/iOS quirks, and build_info/ holds Debian control metadata and maintainer scripts. The system also manages GPG keyrings and launchd service plists, acting as a complete distribution toolchain for the jailbreak community. Despite being a build system rather than a runtime software product, it is actively maintained with thousands of packages and a significant community following. The codebase is massive and well-structured for its niche purpose, but the component are tightly coupled to the Procursus build environment, limiting their reusability outside the project.