This repository provides a Dart FFI binding for llama.cpp, targeting Flutter mobile apps on iOS, Android, and macOS. It wraps the native C API with a modular, isolate-safe architecture: a worker isolate (LlamaEngine) owns the native model and context, while the main isolate communicates via message passing. The binding exposes streaming token generation, chat template rendering, multimodal (vision/audio) support, speculative decoding, context shifting, and session persistence. The codebase is well-structured with clear separation of concerns: FFI bindings, model/context wrappers, sampling, tokenization, and generation logic are each in their own modules. It uses modern Dart features (sealed classes, records, pattern matching) and demonstrates high craftsmanship in error handling, resource management, and cross-isolate communication. The package is a Flutter plugin with native-assets build hooks for Android and Swift Package Manager support for Apple platforms.
This repository provides ESPHome components for integrating the Xiaomi Mijia Bedside Lamp 2 with Home Assistant. It enables full control of the lamp's RGBWW LEDs, front panel touch interface (buttons and slider), and illumination through custom ESPHome components. The firmware replaces the original Xiaomi firmware to eliminate cloud dependencies and enable local-only control. The architecture follows ESPHome's component model with hardware abstraction layers (HAL) for light control and front panel I2C communication. Key features include individually controllable front panel LEDs, light presets with group support, disco mode for immediate updates, night light calibration, and smooth color transitions with hardware-aware handling of low-brightness limitations. The design emphasizes separation of concerns, with distinct modules for color handling, state management, and event processing. Notable technical implementations include a Chain of Responsibility pattern for handling different light modes (off, night, RGB, white), a flexible preset system with group-based activation, and specialized transition handling that avoids attempting smooth color changes at very low brightness levels where the hardware cannot perform them reliably. The firmware maintains compatibility with modern ESPHome versions through active maintenance and timely adaptation to API changes.
Excellent Migrations is an Elixir library that analyzes Ecto migration files to detect potentially dangerous database operations that could cause downtime or data loss. It provides a Credo check, Mix tasks, and direct API usage to identify issues such as adding columns with default values, removing columns, running raw SQL, and more. The tool allows marking operations as safe via special comments or configuration to avoid false positives. The architecture is centered around AST parsing of migration files, with separate modules for detecting dangers, parsing safety comments, and filtering results. Key components like the AST parser and comment parser are well-crafted and reusable in other contexts.
pkg-types is a small, ESM-only TypeScript library that provides utilities and type definitions for Node.js configuration files. In production it is consumed as a dependency rather than run as a service: callers use functions such as readPackageJSON, readTSConfig, findWorkspaceDir, resolveLockfile, parseGitConfig, and sortPackage to locate, parse, serialize, and manipulate package.json, tsconfig.json, and .git/config files. It supports package.json, package.json5, package.yaml, JSONC, JSON5, YAML, and git INI formats, with lazy parser loading for JSON5/YAML to keep cold-start cost low. The code is organized into four small vertical slices — resolve, packagejson, tsconfig, and gitconfig — all re-exported from src/index.ts. Path resolution uses exsolve for module-style imports and pathe for cross-platform paths; parsing is delegated to confbox. The primary data flow is simple: an input path is resolved to the nearest or farthest matching file, the file is read and parsed into a typed object, and optionally written back after mutation. There are no API endpoints, data stores, or background processes. Notable implementations include a boundary-aware nearest/farthest file finder that stops at node_modules, workspace directory detection from pnpm/lerna/turbo/rush/deno markers or git config/lockfiles, a Proxy-based updatePackage that auto-creates common package.json map fields during mutation, deterministic package.json sorting/normalizing, and conversion of git's subsection INI format ([remote "origin"]) into generic dotted INI keys for parsing and back on stringify. The project has solid Vitest coverage, type-level tests, and modern CI using oxlint, oxfmt, and obuild. One minor limitation is that git config subsection round-tripping is regex-based and not fully robust for unusual subsection names containing slashes or other non-word characters.
Camposer is a Kotlin Multiplatform camera library that provides a unified Compose Multiplatform API for camera functionality on Android (via CameraX) and iOS (via AVFoundation). It offers camera preview, photo capture, video recording, flash/torch control, zoom, tap-to-focus, exposure compensation, mirror mode, multi-lens selection, video stabilization, and image analysis (including barcode/QR scanning via the optional code-scanner module). The library is published on Maven Central and includes sample apps demonstrating both Android-only and cross-platform usage. The architecture follows a clean separation of concerns with an internal engine/applier layer: `CameraState` holds reactive state flows for all configurable properties, and platform-specific appliers (`PreviewApplier`, `VideoApplier`, etc.) write changes to the underlying hardware when the state changes. The public API exposes `CameraPreview` composable, `CameraSession` for state/info access, and `CameraController` for capture commands. The codebase is well-structured with expect/actual pattern, version catalogs, ABI validation, and extensive test infrastructure (unit tests, device tests, iOS tests). It is a well-crafted, actively maintained library with high reusability potential.
Wait-for-secrets is a GitHub Action designed to pause workflow execution until a user manually inputs required secrets via a web interface. It leverages GitHub's OIDC tokens to authenticate with the StepSecurity API, which temporarily stores secret metadata and provides a unique URL for user interaction. The action polls the API until secrets are submitted, then makes them available as outputs for subsequent workflow steps. Key features include optional Slack notifications and configurable timeouts. The architecture is straightforward: a TypeScript-based action compiled via ncc into a single JavaScript file, with core logic separated into utility functions for environment parsing, URL generation, and secret handling.
bson4jackson is a Java library that adds BSON (Binary JSON) support to the Jackson JSON processor. It enables seamless serialization and deserialization of data in BSON format using Jackson's familiar API, primarily targeting MongoDB integration since BSON is MongoDB's native format. The library follows Jackson's extension patterns through a custom BsonFactory, BsonGenerator for writing, and BsonParser for reading, complemented by specialized serializers/deserializers for BSON-specific types like ObjectId, Timestamp, and Decimal128. The architecture cleanly separates concerns with dedicated components for factory management, stream handling, type conversion, and buffering optimizations.
https-localhost is a developer tool that runs an HTTPS server on localhost using locally-trusted SSL certificates. It downloads the mkcert binary appropriate for the host OS/architecture, executes it to install a local CA into system/browser trust stores, and generates per-host certificates. The resulting HTTPS server can either serve static files from a directory or act as a reverse proxy to an existing HTTP/HTTPS server, with optional HTTP-to-HTTPS redirection. It ships as both a global CLI (the `serve` binary) and a programmatic API (`createServer()`, `getCerts()`). Architecturally it is a small, well-separated set of ES modules built directly on Node's built-in http/https servers with no framework dependency. The router dispatches between a static file handler and a proxy handler; CORS headers are applied globally; environment/CLI configuration is validated through a zod schema where CLI flags take precedence over environment variables. External runtime dependencies are deliberately minimal: only `mrmime` for MIME lookup and `zod` for config validation, plus the downloaded mkcert binary. The static handler is the strongest piece: it implements path-traversal sanitization, ETag/Last-Modified conditional requests, byte-range responses (206/416), HEAD/OPTIONS support, directory index serving with query-preserving trailing-slash redirects, and custom 404.html lookup. The proxy handler strips hop-by-hop headers, injects X-Forwarded-* headers, tunnels WebSocket upgrades, and cleanly returns 502 when upstreams fail or die mid-response. Test coverage is comprehensive and genuinely exercises real sockets: unit tests for env parsing and certificate lifecycle, integration tests for static serving and proxying, CLI spawn tests, and edge-case tests for malformed percent-encoding, CORS preflight without upstream contact, and client hang prevention. CI runs a matrix across Ubuntu/macOS and Node 24/26 with strict type-aware linting, dead-code detection (knip), and formatting checks. Minor weaknesses: synchronous fs calls in the static handler (fine for a dev tool), a `process.exit()` call inside library code for unsupported platforms, minimal error handling in the mkcert download routine (only follows 302s, ignores non-2xx statuses), and no Windows CI coverage despite Windows-specific code paths.
TelegramRSS is a PHP web service that acts as a proxy between clients and a Telegram API server, providing RSS and JSON feeds of Telegram channel content. Built with AMPHP for asynchronous handling, it features rate limiting, IP-based access control, and middleware-based request processing. The service fetches messages from Telegram via a backend API server and formats them as RSS or JSON responses with media support.
dogpile.cache is a mature Python caching library that provides a generic caching API with a 'dogpile lock' mechanism to prevent thundering herd problems during cache regeneration. The library is structured around a central `CacheRegion` class that configures a pluggable backend (memory, memcached, redis, valkey, dbm, null) and exposes high-level operations like `get_or_create`, `set`, `get`, and `delete`. It also offers function decorators (`cache_on_arguments`, `cache_multi_on_arguments`) that automatically generate cache keys from function arguments and manage expiration and invalidation. The architecture follows a layered design: the `dogpile.lock` module implements the core dogpile lock (a mutex-based coordination pattern), the `dogpile.cache.api` module defines abstract base classes for backends and serialization, and concrete backends in `dogpile.cache.backends` implement the actual storage. The `CacheRegion` orchestrates key mangling, serialization, expiration, and invalidation strategies. External dependencies are minimal (decorator, stevedore, typing_extensions), with optional backends pulling in redis, valkey, pymemcache, etc. Key entry points include `make_region()` to create a region, `CacheRegion.configure()` to set up a backend, and `CacheRegion.get_or_create()` for atomic cache-and-generate operations. The library is well-tested with extensive unit and integration tests, uses modern Python features (type hints, ABCs, context managers), and follows a clean, maintainable code style. It is actively maintained as part of the SQLAlchemy project.
Delocate is a macOS-specific tool for making Python wheels redistributable by bundling required dynamic libraries. It analyzes wheels to find external library dependencies, copies those libraries into the wheel, and updates install names and rpaths to use @loader_path references. The tool consists of a core library with a rich CLI interface providing commands like delocate-wheel, delocate-path, delocate-listdeps, and delocate-merge for combining wheels of different architectures. The architecture separates concerns into modules for dependency analysis (libsana), low-level Mach-O binary manipulation (tools), wheel-specific operations (wheeltools), and directory tree fusing (fuse).
JSON::JWT is a Ruby library implementing the JSON Web Token (JWT) family of standards, including JWS (JSON Web Signature), JWE (JSON Web Encryption), and JWK (JSON Web Key). It provides comprehensive cryptographic operations for signing, verifying, encrypting, and decrypting tokens using various algorithms (HMAC, RSA, ECDSA, RSA-PSS, and symmetric encryption). The library also includes JWK Set management with remote fetching and caching capabilities. The architecture is modular: JSON::JWT serves as the main entry point, delegating to JSON::JWS for signing/verification and JSON::JWE for encryption/decryption. JSON::JWK handles key representation and conversion between OpenSSL keys and JWK format. The JWK Set Fetcher provides a configurable HTTP client with pluggable caching, suitable for OpenID Connect discovery. The code demonstrates strong security practices, including constant-time comparison for MAC verification, protection against padding oracle attacks in CBC mode, and proper validation of algorithm headers to prevent algorithm confusion attacks. The test suite covers interop with other JWT libraries (Nimbus JOSE, jsrsasign) and RFC examples, ensuring correctness. Key external dependencies include OpenSSL for cryptographic primitives, Faraday for HTTP, and ActiveSupport for hash extensions. The library is actively maintained with support for Ruby 3.2-4.0 and recent security fixes.
GPTLint is a linting tool that uses Large Language Models (LLMs) to enforce code quality rules. It employs a two-pass approach: a faster, cheaper LLM model identifies potential violations, and a more accurate, expensive model validates them to reduce false positives. The tool supports both built-in and custom rules defined in markdown files, caches results to avoid reprocessing unchanged files, and provides both a CLI and programmatic API. The architecture is modular, with separate components for configuration, rule parsing, linting, caching, and LLM abstraction.
AudioX-Turbo is a unified framework for efficient anything-to-audio generation (text, video, audio conditions → audio) using Distribution Matching Distillation (DMD) adapted to flow matching and adversarial training. It supports six core tasks: Text-to-Audio, Text-to-Music, Video-to-Audio, Video-to-Music, Text+Video-to-Audio, and Text+Video-to-Music. The system includes training scripts, an interactive Gradio demo, and a modular architecture with interchangeable components (attention, bottlenecks, etc.) via a factory pattern. Key innovations enable high-quality 4-step audio generation, reducing computational cost by up to 25× compared to multi-step diffusion models.
Open Dungeon is a fully local AI roleplay application that generates both narrative text and inline scene images on-device. Built as a Next.js 16 application with a React 19 frontend, it provides a rich interactive fiction experience where users can write actions, dialogue, or narration, and the AI narrator responds with streaming prose. The system supports multiple text backends: local Ollama models (Gemma 4 QAT variants) or any OpenAI-compatible server (LM Studio, llama.cpp, OpenRouter, etc.). Image generation is handled by a separate Python HTTP server that wraps optimized FLUX.2-klein backends (MFLUX/MLX on Apple Silicon, PyTorch SDNQ on CUDA/CPU) or an existing ComfyUI instance. The architecture is a monolith with a Next.js server handling API routes, SQLite for persistence, and an optional sidecar image server. Key entry points include the main chat UI (src/app/page.tsx), the story API (src/app/api/story/route.ts) which handles streaming narration, and the image generation API (src/app/api/images/route.ts). The codebase demonstrates strong engineering practices: thorough streaming support with artifact filtering, robust error handling with timeouts and fallbacks, clean database abstraction, and well-structured Python image server with resident model support.
This is an OpenCode plugin that provides isolated branch workspaces using devcontainers or git worktrees. It enables developers to work on multiple branches simultaneously with full container isolation (devcontainers) or lightweight filesystem isolation (worktrees). The plugin manages the entire lifecycle: creating clones or worktrees, allocating unique ports from a configurable range, copying gitignored secrets, and intercepting bash commands to run inside the target container or worktree directory. It also offers session state persistence, background job tracking for non-blocking container starts, and unified workspace management with cleanup of stale workspaces. The codebase is well-architected with clean separation into core modules (clones, config, devcontainer, git, jobs, paths, ports, worktree, workspaces) and a main plugin entry point that registers tools and bash interception hooks. It uses Node.js built-in test runner with comprehensive unit and integration tests. The plugin demonstrates strong engineering practices including file-based locking for concurrency, atomic writes, security-conscious command construction (shell quoting, Docker filter sanitization), and support for both Docker and Podman runtimes.
The Universal Intel Chipset Device Updater is a Windows PowerShell utility that automates the detection, download, verification, and silent installation of Intel Chipset Device Software (INF files). It scans the system for Intel chipset hardware, matches detected devices against a curated database of INF packages, downloads the correct installer from primary (GitHub) or backup mirrors, verifies integrity via SHA-256 hashes and Intel digital signatures, creates a system restore point, and installs the INF files without unnecessary drivers or bloatware. The tool supports interactive, automatic, and quiet modes, making it suitable for home users, IT technicians, and MDM deployments (Intune, SCCM, Workspace ONE, PDQ Deploy). Architecturally, the system is composed of several PowerShell scripts with batch launchers, distributed as a self-extracting executable (SFX). The main updater script (universal-intel-chipset-device-updater.ps1) orchestrates the workflow: hardware detection via get-intel-hwids.ps1, database parsing from Markdown files, download management, multi-layer security verification, and installation. A companion INF extractor script (universal-intel-chipset-inf-extractor.ps1) handles extraction from modern Intel installers using NanaZip. The codebase emphasizes safety with self-hash verification, automatic update checks, EOL package handling, and cAVS audio conflict safeguards. While primarily an application, the repository also contains data assets: a comprehensive INF database (intel-chipset-infs-latest.md with 6954 entries), download lists with SHA-256 hashes, and extensive documentation including AI audit reports and deployment guides. These data files are tightly coupled to the tool's operation but represent a structured, maintainable knowledge base for Intel chipset support across generations.
TempMail is a self-hosted temporary email service platform that provides disposable email addresses with automatic expiration. The system uses a microservices architecture with separate containers for API (Go/Gin), frontend (Nginx), email reception (Postfix), and email processing (custom LMTP daemon). Key features include domain pool management with automatic MX validation, API Key authentication with rate limiting, OAuth login options, and web-based email management. The platform is optimized for high throughput with innovations like a delivery batcher that groups multiple email writes into single database transactions and a stats flusher that efficiently aggregates in-memory counters.
maxapi is an async Python SDK and bot framework for the MAX messenger platform. It provides a high-level client (`Bot`) for all MAX Bot API endpoints, an event-driven dispatcher with router hierarchy (`Dispatcher`/`Router`), composable filters (magic filter `F`, `Command`, `StateFilter`, etc.), middleware chains, a finite state machine (FSM) with multiple backends (memory, Redis), and webhook integrations for aiohttp, FastAPI, and Litestar. The library is heavily inspired by aiogram but adapted to MAX API specifics such as attachment types, callback payloads, and deep linking. It includes extensive utility code for formatting (HTML/Markdown), inline keyboard building, VCF parsing, and time helpers. The codebase is well-structured with Pydantic v2 models for all API entities, rigorous retry logic, and a comprehensive test suite covering unit, integration, and webhook scenarios. While not a standalone application, it serves as a foundational component for building production-ready MAX bots.
This is a Node.js automation tool that deploys and maintains Cloudflare Web Application Firewall (WAF) custom rules across all zones in a Cloudflare account. It runs as a long-lived daemon (via PM2) with scheduled cron jobs: one fetches the latest expressions from the repository (git pull) and another updates the WAF rules via the Cloudflare Rulesets API. The service parses rule expressions from a markdown file, applies user configuration (PHP/WordPress support, custom allowlist/blocklist per zone, external IP threat feeds from SniffCat and AbuseIPDB), and syncs the resulting custom rules and IP lists to Cloudflare. It includes robust error notification via email and Discord webhooks with batched alert delivery. The codebase is well-structured, thoroughly tested (Jest), and actively maintained with frequent updates.