This is a monorepo containing over 40 independent MCP (Model Context Protocol) servers for AWS services, hosted under awslabs/mcp. Each server exposes tools, resources, and prompts that allow AI agents (coding assistants, chatbots) to interact with AWS APIs, databases, documentation, and best practices. The architecture is a collection of standalone Python packages, each with its own entry point (server.py/main.py), tests, Dockerfile, and PyPI distribution. Servers are built using either FastMCP or the lower-level MCP Python SDK, with consistent patterns for read-only enforcement, error handling, AWS client creation, and user-agent tracking. The codebase demonstrates solid engineering with extensive test coverage, but suffers from significant copy-paste boilerplate across servers (healthcheck scripts, Context singletons, test_init.py) and inconsistent use of MCP frameworks. Notable high-craftsmanship components include the AWS API MCP server's custom CLI-to-IR parser and execution engine, the OpenAPI MCP server's pluggable authentication provider framework, and the DynamoDB MCP server's schema-driven code generation tool. The repository is actively maintained with frequent commits and issue triage, though some servers are marked as deprecated or in initial setup, and the announcement of a successor ('Agent Toolkit for AWS') suggests potential future migration.
python-for-android (p4a) is a build toolchain that packages Python applications into Android APK, AAB, and AAR binaries. It cross-compiles Python interpreters, native libraries, and Python packages for Android architectures (arm64-v8a, armeabi-v7a, x86_64, x86) using the Android NDK and SDK. The system is driven by a recipe system where each Python package or native library defines its download, patching, and build steps. A bootstrap system (SDL2, SDL3, webview, Qt, service-only) generates the Android project structure and Gradle build files. The tool is primarily used via CLI (`p4a` or `python-for-android`) or integrated with Buildozer. It resolves dependencies via a graph algorithm that handles conflicts, alternatives, and optional dependencies. The codebase is large and mature, with hundreds of recipes and extensive CI testing, but carries technical debt from version-specific patches and workarounds for Android's evolving ecosystem.
Pollinations is a large monorepo hosting a multi-modal generative AI platform. It provides a unified API gateway (gen.pollinations.ai) for text, image, video, audio, 3D, and embeddings generation, routing requests to dozens of providers (OpenAI, Anthropic, Google, Fal, Replicate, ElevenLabs, etc.) with fallback chains, caching, rate limiting, and billing. A separate enterprise service (enter.pollinations.ai) handles authentication (GitHub OAuth, API keys, device code flow), user dashboards, API key management, Stripe/Pollen billing, community-hosted model endpoints, and a quest/reward gamification system. The platform also includes a design system UI library (@pollinations/ui), a Node.js SDK, a CLI tool (polli-cli), an MCP server, and over 20 community apps (chat, AI dungeon master, image reimagine, Discord bots, etc.). Infrastructure includes Cloudflare Workers, Durable Objects, D1, R2, Tinybird for analytics, and GPU inference servers (Flux, Dreamshaper, Z-Image) deployed on Vast/IO.net. The architecture is service-oriented within a monorepo: each major service (gen, enter, media, pollinations.ai) is a separate Cloudflare Worker with its own wrangler.toml and dependencies. Shared packages (shared/, packages/) provide cross-cutting concerns like billing logic, model registry, auth middleware, HTTP error handling, and UI components. Data flows from user requests through middleware (auth, balance, cache, rate-limit, safety, model resolution) to provider-specific adapters, with usage events published to Tinybird for analytics and billing. The system is heavily tested with Vitest, including integration tests with Cloudflare Workers test environment and VCR snapshots. External dependencies include Cloudflare (Workers, D1, R2, KV, Durable Objects, Vectorize), Tinybird (analytics), Stripe (payments), GitHub (OAuth, quests), and numerous AI model providers. The codebase is predominantly TypeScript with some Python (Floret agent, GPU servers, social automation) and Rust (polli-core for Discord bot). While the core platform is well-maintained, there is significant drift in community apps and infrastructure scripts, and the sheer number of provider integrations creates maintenance surface area.
git-sim is a CLI tool that generates visualizations of Git operations from within a local repository. It uses the Manim animation library to render commit graphs, branches, tags, and file states as either static images (JPG/PNG) or animated videos (MP4/WEBM). The architecture is built around a base class `GitSimBaseCommand` that extends Manim's `MovingCameraScene` and provides common drawing methods for commits, refs, arrows, and zone tables. Each Git subcommand (add, branch, checkout, merge, etc.) is implemented as a separate Python class that inherits from this base class and overrides the `construct()` method to build the specific scene. The CLI interface is defined using Typer, and the application reads the current Git repository state via GitPython. The animation output is handled by a dedicated `handle_animations` function that manages rendering, format conversion, and file output. The codebase is well-organized, with clean separation between command logic, animation rendering, and configuration management via Pydantic settings. It demonstrates effective use of the Manim library for non-trivial graphics and provides a robust framework for simulating Git commands.
Themis is a cross-platform cryptographic library developed by Cossack Labs, providing high-level security primitives such as Secure Cell (symmetric encryption with three modes: Seal, Token Protect, Context Imprint), Secure Message (asymmetric encryption and digital signing), Secure Session (mutual authentication and key exchange), Secure Comparator (zero-knowledge password comparison), and key generation (EC, RSA, symmetric). The core is implemented in C with a cryptographic engine abstraction layer (Soter) that supports OpenSSL, BoringSSL, and a custom Ed25519 implementation. The library exposes consistent APIs across 13+ language wrappers including Python, Go, Rust, Java, Swift, Objective-C, JavaScript (Node.js and WebAssembly), PHP, Ruby, C++, and React Native. The codebase is well-structured with extensive unit, integration, and fuzz testing, CI/CD pipelines for each wrapper, and cross-language compatibility tests. While the repository has deprecated older APIs and maintains multiple backends, it remains actively maintained with regular releases, security patches, and community engagement.
pythonocc-core is a Python binding for the OpenCASCADE Technology (OCCT) 3D CAD kernel, providing full access to OCCT's geometry, topology, visualization, and data exchange capabilities. It is a library, not a standalone application, designed for engineers and developers building CAD/CAE/PLM tools in Python. The architecture is a C++ extension module generated via SWIG, with a thin Python layer that adds convenience utilities, multiple display backends (Qt, wxPython, Tkinter, WebGL/Jupyter), and high-level helpers for shape creation, topology traversal, and file I/O (STEP, IGES, STL, glTF, OBJ, PLY). The core API mirrors OCCT's class hierarchy, while the `OCC.Extend` and `OCC.Display` packages provide Pythonic wrappers and interactive 3D viewers. The repository is actively maintained, with recent releases (7.9.3) and CI pipelines for conda packages across platforms. It includes extensive type stubs for IDE support, a comprehensive test suite, and support for offscreen rendering and Jupyter notebooks. Key entry points include `OCC.Display.SimpleGui.init_display()` for interactive viewers, `OCC.Extend.DataExchange.read_step_file()` for CAD file import, and `OCC.Core.Tesselator.ShapeTesselator` for mesh generation. The codebase demonstrates strong engineering practices with reusable patterns like the topology explorer, shape factory utilities, and a generic proxy pattern for OCCT objects. However, the SWIG-generated wrapper layer is large and fragile, requiring careful version alignment with OCCT, and some OCCT classes remain unwrapped.
Irodori-TTS is a state-of-the-art Japanese text-to-speech system built on rectified flow diffusion over DACVAE audio latents. It provides a complete training and inference stack: `train.py` supports full fine-tuning, LoRA, speaker inversion, and MeanFlow distillation with distributed training and WSD scheduling; `infer.py` offers a rich CLI with CFG, Sway Sampling, duration scaling, and watermarking; `gradio_app.py` and `gradio_app_voicedesign.py` expose interactive UIs for reference-based cloning and caption-conditioned voice design. The architecture is modular (`irodori_tts` package) with attention dispatch (FA3/SDPA), a hybrid Muon+AdamW optimizer, codec abstraction, and caching-friendly `InferenceRuntime`. While highly engineered for production use (INT4/INT8 quantization, safetensors, HuggingFace Hub integration), it remains a research-oriented codebase with experimental sampling algorithms and many YAML configs for different model sizes/versions. The code is clean and well-factored, though `train.py` is a large monolith and external dependencies (DACVAE, silentcipher) are pinned to specific commits, creating ecosystem fragility.
This repository contains a minimal coding agent designed to work with local Ollama models. The agent operates in a loop, processing user requests by inspecting and modifying the workspace through a set of structured tools (file operations, shell commands, etc.). It implements six core components: live repo context collection, prompt shaping with cache reuse, structured tools with validation and permissions, context reduction and output management, transcript and memory persistence, and bounded delegation to subagents. The agent is implemented in a single Python file (`mini_coding_agent.py`) and is intended to be run via a CLI entry point. It requires an Ollama server running with a compatible model (default: qwen3.5:4b). The code is clean, well-structured, and demonstrates thoughtful agent design patterns, making it a useful reference for building similar agents.
Linkook is a Python CLI tool for OSINT-driven discovery of linked social media accounts and associated emails across multiple platforms. It takes a single username as input, scans a configurable set of websites (defined in a provider.json file), and recursively extracts linked accounts, usernames, and emails from discovered profiles. The tool supports breach checking via HudsonRock's database or Have I Been Pwned, and can export results to a Neo4j-compatible JSON format for graph visualization. The architecture is modular: a Provider class handles URL templates and regex extraction, a SiteScanner performs HTTP requests and keyword-based existence checks, and a ScannerManager manages concurrent scanning via a thread pool with a queue. The code is well-structured, includes unit tests, and demonstrates solid engineering practices with clean separation of concerns, error handling, and logging.
GNews is a Python package that provides an API to search and retrieve news articles from Google News. It features a modular architecture with a core GNews class, backend abstraction (RSS and SearchApi), async support, URL resolution via Playwright, full-text extraction, and export capabilities. The package includes a command-line interface and handles common challenges like rate limiting with exponential backoff, proxy support, and graceful fallbacks. The codebase is actively maintained with regular updates and follows modern Python practices.
AI Vocabulary Builder is a full-stack web application (FastAPI backend, Vue.js frontend) that uses LLM APIs (OpenAI, Gemini, Anthropic, DeepSeek) to help users build and learn English vocabulary. The backend exposes REST APIs for translating text, extracting rare words, managing a vocabulary store (TinyDB), generating stories and quizzes from learned words, and exporting to Anki/CSV. The frontend provides an intuitive single-page interface with build, learn (words list, quiz, story) and settings views. The architecture follows a modular monolith with clear separation of concerns: `builder` for vocabulary extraction, `learn` for quiz/story features, `system` for settings and model configuration, and `infras` for persistence and AI model abstraction. The codebase is well-structured with comprehensive unit tests, CI/CD via GitHub Actions, and rigorous linting/formatting (ruff, mypy).
vimhjkl is a Vim technique trainer that drills advanced Vim skills (dot command, text objects, registers, macros, Ex commands) using real vim/nvim as a subprocess. It features a CLI with interactive menu (j/k navigation, vim-style counts) and four practice modes: Learn, Blind, Practice, Grind, and Review. The application uses a spaced repetition system (Leitner boxes) to schedule skill reviews based on keystroke efficiency. It includes 66 skills across 230 challenges, with a curriculum defined in JSON and generated by a separate build system. The codebase is pure Python stdlib with zero third-party dependencies, and supports internationalization (Chinese shipped) with an AI translation pipeline. The architecture is modular: CLI entry point, engine for session logic, grader for vim subprocess management, store for progress persistence, and TUI for terminal rendering. The repository also contains build scripts for curriculum generation and packaging (Homebrew, AUR).
FluentYTDL is a Windows desktop application built with PySide6 and qfluentwidgets that provides a modern Fluent Design GUI for downloading videos and audio from YouTube and X (Twitter). It wraps yt-dlp as the core download engine and adds a rich layer of task management, error diagnostics, cookie-based authentication, subtitle handling, metadata embedding, and VR/360 video support. The application features a persistent SQLite task database, structured observability with JSONL event tracing, rule-based error classification, and a self-update mechanism with component version management. It is heavily internationalized (English and Simplified Chinese) and includes a comprehensive test suite covering UI, download pipeline, diagnostics, and observability contracts. Architecturally, the codebase is organized into distinct modules: core (config, dependency, update), auth (cookie sentinel, WebView2 provider), download (executor, workers, staging, quality guard), processing (metadata, subtitles, sponsorblock), diagnostics (rule engine, line collector), observability (events, traces, artifacts), and UI (pages, dialogs, delegates, components). The download pipeline is transactional with staging areas and manifests, ensuring atomicity and recovery. The observability layer is best-effort and never blocks business logic, emitting structured events for debugging and bug bundle export. The code demonstrates strong engineering practices: immutable dataclasses for plans, singleton managers with explicit initialization guards, strategy patterns for subtitle selection, and extensive use of signals for async UI updates. Key reusable components include the container compatibility resolver (deterministic output container decision), BCP-47 language matching with tiered resolution, the quality guard pattern for pre/post-download verification, the transactional staging pattern for file management, and the cookie runfile isolation pattern to prevent yt-dlp from corrupting truth-source cookies. The diagnostics rule engine and observability event system are also well-isolated and could be extracted for other yt-dlp wrapper projects.
CyberClaw is an enterprise-grade transparent AI agent architecture that implements a dual-watermark memory system, full behavior auditing, zero-trust execution via two-phase skill invocation, and a heartbeat task engine for scheduled operations. The system provides an interactive conversational interface where users can query time, perform calculations, manage files, execute shell commands within a secure sandbox, and utilize pluggable skills. All agent actions are transparently logged to JSONL files for monitoring and compliance. The architecture features clean separation of concerns with distinct modules for agent logic, skill loading, context management, logging, heartbeat tasks, and tool execution.
Lumen is an open-source agent framework for chatting with data and building interactive dashboards, built on the HoloViz ecosystem (Panel, hvPlot, Bokeh). It provides a dual interface: a YAML-based declarative dashboard specification system for non-AI use cases, and an AI-powered natural language interface that generates SQL queries and Vega-Lite visualizations from user prompts. The architecture is modular, with clear separation between data sources (files, SQL databases, REST APIs, xarray), data pipelines (filters, transforms, SQL rewrites), views (hvPlot, Vega-Lite, DeckGL, tables), and an AI agent layer (Planner, Coordinator, specialized agents for SQL, visualization, validation, and source ingestion). The AI system supports multiple LLM providers (OpenAI, Anthropic, Google, Ollama, etc.), vector stores for RAG (Numpy, DuckDB, ChromaDB), and a tool system for external API integration. The codebase demonstrates high craftsmanship with reusable patterns like spec-driven component creation (MultiTypeComponent), composable SQL transforms, and a consistent agent/tool abstraction. While primarily an application framework, it includes data pipeline components that could be extracted for data engineering use. The project is actively maintained with extensive test coverage, though its large optional dependency surface and rapid feature velocity introduce some maintenance risk.
Harmonica is a Python library for forward modeling, processing, and inversion of gravity and magnetic data, part of the Fatiando a Terra project. It provides high-performance, Numba-accelerated kernels for computing gravitational and magnetic fields of various geometric bodies (prisms, tesseroids, point masses, ellipsoids, dipoles) and supports equivalent source interpolation, frequency-domain grid transformations (upward continuation, reduction to pole, derivatives), spherical harmonic models (IGRF), and Euler deconvolution. The library is designed for geoscientists and educators, integrating tightly with the scientific Python ecosystem (xarray, scikit-learn, pygmt, verde). It features a scikit-learn-like API for its gridder classes, extensive documentation and testing, and is actively maintained as part of the Fatiando project. While its core functionality is domain-specific, the code exhibits strong software engineering practices: modular architecture, Numba JIT compilation, consistent interfaces, and comprehensive CI/CD.
RocksDict is a persistent key-value store for Python, backed by the RocksDB embedded database. It provides a Pythonic dictionary-like interface (`Rdict`) for storing and retrieving data that exceeds memory capacity, with support for arbitrary Python objects via pickle or raw bytes. The system is implemented as a Rust extension (using PyO3/maturin) that interacts with the RocksDB C++ library via FFI, offering high performance and durability. The architecture separates concerns into a safe Rust core and a thin Python binding layer. The Rust core implements resource management (via `DbReferenceHolder` to handle RocksDB reference counts and background work cleanup), centralized FFI error handling (through the `ffi_try` macro), and a flexible encoding layer (`encoder.rs`) that supports both raw bytes and pickled Python objects with type tags for efficient deserialization. Configuration persistence is handled via `RocksDictConfig`, which serializes database options to JSON for automatic reloading, enabling seamless recovery of tuned settings. Key features include column families, snapshots, write batches, checkpoints, and SST file ingestion, making it suitable for a wide range of applications from caching to large-scale data storage. The project is actively maintained with comprehensive CI coverage across multiple Python versions and platforms, demonstrating a commitment to reliability and cross-platform compatibility.
Rubicon-ObjC is a bridge library enabling Python to interact with Objective-C runtime and frameworks. It allows instantiation and invocation of Objective-C objects from Python, subclassing Objective-C classes in Python, and automatic conversion between Python and Objective-C data types. The library is part of the BeeWare suite for cross-platform Python GUI development, targeting iOS/macOS platforms. Architecture centers around modular submodules: api (high-level decorators and wrappers), runtime (low-level Objective-C interactions), types (data conversion), collections (Foundation collection wrappers), and eventloop (asyncio integration). Key features include bidirectional type conversion, decorator-based exposure of Python to Objective-C, block handling, and memory management automation. The library is used in production iOS apps (e.g., Travel Tips) and serves as a dependency for other BeeWare projects like Toga.
This repository provides a Python library for parsing human-readable time expressions into seconds. The core functionality is implemented in the `timeparse` function, which accepts strings in various formats (e.g., '32m', '2h32m', '1w3d2h32m', '4:13:02', '2 days, 4:13:02') and returns the equivalent duration in seconds as an integer or float. The library handles optional signs, fractional values, and ambiguous colon-separated values via a granularity parameter. It has no external dependencies and is designed for simplicity and reliability. The library is structured as a single module with a clear public interface. The implementation uses a series of regular expressions to match different time formats and converts the matched components to seconds using predefined multipliers. The code is well-tested with a comprehensive test suite that covers the various input formats and edge cases. Despite its simplicity, the library demonstrates careful attention to detail in handling the complexities of time expression parsing, such as supporting both written and abbreviated units, handling spaces and commas as separators, and providing control over the interpretation of colon-separated values.
Calysto Scheme is a sophisticated Scheme interpreter implemented in Python that integrates with Jupyter notebooks. It features a multi-stage compilation pipeline (CPS → DS → RM → Python) with JIT compilation, continuation support (call/cc), and seamless Python interoperability. The system includes advanced optimizations like persistent vectors for O(1) lexical address lookup, epoch-based cache invalidation, and escape analysis for JIT inlining. It's used in educational contexts and provides debugging tools like a stepper-debugger, making it both a practical development tool and a platform for teaching programming language concepts.