Preset
Background
Text
Font
Size
Width
Account Monday, March 23, 2026

The Git Times

“Everywhere we remain unfree and chained to technology, whether we passionately affirm or deny it.” — Martin Heidegger

AI Models
Claude Sonnet 4.6 $15/M GPT-5.4 $15/M Gemini 3.1 Pro $12/M Grok 4.20 $6/M DeepSeek V3.2 $0.89/M Llama 4 Maverick $0.60/M
Full Markets →

OpenViking Applies Filesystem Model to AI Agent Context 🔗

New context database unifies memory, resources and skills in a hierarchical structure, giving agents an observable and evolvable knowledge layer.

volcengine/OpenViking · Python · 18.2k stars 2mo old · Latest: v0.2.9

OpenViking is an open-source context database built specifically for AI agents. Rather than bolting additional vector stores onto existing agent frameworks, the project replaces fragmented context management with a single filesystem-inspired paradigm.

Traditional agent development scatters critical information.

OpenViking is an open-source context database built specifically for AI agents. Rather than bolting additional vector stores onto existing agent frameworks, the project replaces fragmented context management with a single filesystem-inspired paradigm.

Traditional agent development scatters critical information. Memories live in application code or chat history, resources sit in separate vector databases, and skills are implemented as disconnected tool calls. The result is brittle retrieval, lost information during long-running tasks, and retrieval chains that function as black boxes. OpenViking addresses these problems by treating an agent’s entire context—memories, documents, learned skills—as files and directories.

The system organizes data hierarchically. Developers can create folders for specific tasks, projects or time periods, place resources inside them, and let the agent navigate the structure much as a human would browse a local drive. This delivers context in a structured, observable way. When an agent needs information, the retrieval path is explicit rather than hidden behind opaque similarity scores. The project also supports self-evolving behavior: agents can write new memories or update skill definitions directly into the filesystem, allowing incremental improvement without external orchestration.

Recent work on version 0.2.9 shows the team hardening the system for production use. Changes include enforcing agent-level watch task isolation, using file summaries for more effective embedding in the semantic pipeline, fixing incremental directory updates in the vector store, and sharing a single RocksDB adapter across backends to prevent lock contention. The bot component gained mode configuration, a debug mode, and a /remember command that lets users explicitly add information to the agent’s persistent memory. Documentation updates now include Docker Compose instructions, making local deployment straightforward.

Built in Python, OpenViking integrates with agent frameworks such as OpenClaw. It abandons the “flat storage” model of conventional RAG in favor of a global, hierarchical view that preserves relationships between pieces of context. For developers building agents that must maintain coherence over weeks or months of operation, this approach removes much of the custom glue code previously required to stitch together memory, tools and knowledge bases.

The project is still early, but its minimalist interaction model offers a cleaner mental model for context engineering. Builders no longer need to decide whether a piece of information belongs in short-term memory, long-term vector storage or a skill registry. They simply put it in the right directory.

**

Use Cases
  • Long-running autonomous agents maintaining task memory
  • Hierarchical organization of agent skills and resources
  • Observable retrieval debugging for complex agent workflows
Similar Projects
  • MemGPT - manages persistent agent memory but uses fixed token windows instead of a navigable filesystem
  • LlamaIndex - provides strong data indexing for RAG yet lacks unified handling of skills and agent memories
  • LangGraph - orchestrates agent workflows effectively but treats context as external state rather than a native filesystem

More Stories

PicoClaw Delivers Efficient AI Assistance on Tiny Hardware 🔗

Go implementation enables AI automation with minimal memory and fast startup times

sipeed/picoclaw · Go · 25.8k stars 1mo old

PicoClaw is an AI assistant written entirely in Go that runs on hardware costing around $10. The application uses approximately 10MB of RAM and achieves millisecond boot times, allowing deployment on resource-constrained single-board computers.

It connects to multiple model providers including Kimi, Minimax and Avian, with built-in support for model routing.

PicoClaw is an AI assistant written entirely in Go that runs on hardware costing around $10. The application uses approximately 10MB of RAM and achieves millisecond boot times, allowing deployment on resource-constrained single-board computers.

It connects to multiple model providers including Kimi, Minimax and Avian, with built-in support for model routing. Communication channels include Matrix, IRC, WeCom and Discord proxy functionality. Version v0.2.3 added MCP protocol compatibility, a vision processing pipeline and JSONL-based memory storage.

The system provides a web UI launcher and Docker Compose configuration for deployment. A cross-platform system tray interface is available for Windows and Linux. Additional tools include SpawnStatusTool for monitoring sub-agents, configurable cron execution controls and gateway hot-reloading.

Security measures focus on command whitelisting, execution gating and path validation for symlinks. The architecture emphasizes low overhead while supporting persistent memory, multi-channel orchestration and provider failover.

By running locally on minimal devices, PicoClaw enables automation without reliance on cloud infrastructure or high-end servers. Recent updates addressed provider input handling, WebSocket proxying and UI fallback implementations for non-CGO environments. The project remains in active development ahead of a v1.0 release.

Use Cases
  • Hobbyists automating routines on low-cost single-board computers
  • Developers integrating AI agents across Matrix and Discord channels
  • Engineers deploying vision pipelines on constrained embedded hardware
Similar Projects
  • NanoBot - similar lightweight assistant but implemented outside Go
  • Auto-GPT - autonomous agent framework with significantly higher RAM needs
  • LangChain - Python-based agent tools lacking PicoClaw's minimal footprint

Plugin Reuses Claude Credentials for OpenCode Authentication 🔗

Tool intercepts Anthropic requests using existing OAuth tokens from Keychain or credentials file

griffinmartin/opencode-claude-auth · TypeScript · 350 stars 3d old

OpenCode users no longer need separate credentials to access Anthropic models. The opencode-claude-auth plugin leverages existing Claude Code authentication to handle all Anthropic API interactions.

Created in TypeScript, it installs as an OpenCode plugin that registers a dedicated auth provider.

OpenCode users no longer need separate credentials to access Anthropic models. The opencode-claude-auth plugin leverages existing Claude Code authentication to handle all Anthropic API interactions.

Created in TypeScript, it installs as an OpenCode plugin that registers a dedicated auth provider. A custom fetch handler intercepts requests, reading OAuth tokens from the macOS Keychain on preferred platforms or the ~/.claude/.credentials.json file elsewhere. The system maintains an in-memory cache with 30-second TTL.

Automatic refresh occurs via the Claude CLI when tokens near expiry. Background re-sync runs every five minutes. It also populates OpenCode's auth.json for compatibility, accommodating multiple installation paths on Windows.

Developers require both Claude Code and OpenCode installed, with initial authentication completed through the Claude CLI. Installation methods range from Homebrew taps to npm global packages and AI-assisted setups that fetch detailed instructions.

Once activated in opencode.json, the plugin operates transparently. This eliminates the builtin Anthropic auth plugin requirement and streamlines workflows for teams using multiple Anthropic products.

Use Cases
  • OpenCode developers accessing Claude models without new API keys
  • Mac users utilizing Keychain for Anthropic authentication in OpenCode
  • Cross-platform teams sharing credentials between Claude and OpenCode
Similar Projects
  • anthropic-opencode-plugin - requires separate API key management
  • claude-credentials-sync - focuses only on credential file sharing
  • custom-auth-provider - lacks automatic token refresh capabilities

macOS App Organizes AI Coding Agent Skills 🔗

Unified interface supports editing and management across Claude Code Cursor and similar platforms

Shpigford/chops · Swift · 429 stars 4d old

Chops is a macOS application that brings order to AI agent skills scattered across multiple coding tools. Written in Swift with SwiftUI, the app scans for skill files used by Claude Code, Cursor, Codex, Windsurf, Copilot, Aider and Amp, presenting them in a three-column interface.

Instead of hunting through dotfiles, users browse, search and edit skills from one location.

Chops is a macOS application that brings order to AI agent skills scattered across multiple coding tools. Written in Swift with SwiftUI, the app scans for skill files used by Claude Code, Cursor, Codex, Windsurf, Copilot, Aider and Amp, presenting them in a three-column interface.

Instead of hunting through dotfiles, users browse, search and edit skills from one location. The built-in monospaced editor supports Cmd+S saving and frontmatter parsing. Collections let developers group skills logically without modifying the original source files. FSEvents-based real-time watching updates the display instantly when files change on disk, while full-text search covers names, descriptions and content.

The app generates new skills with correct boilerplate for each supported tool. It also connects to remote skill servers such as OpenClaw, enabling discovery, browsing and installation of additional skills. Version 1.6.0 fixed a layout freeze on skill selection, added Enter-key collection creation and enabled renaming via the right-click menu.

Chops requires macOS 15 Sequoia or later. The project uses SwiftData for its model layer, Sparkle for updates, and generates its Xcode project from a project.yml file using xcodegen.

Use Cases
  • Developers manage skills across Claude Code and Cursor simultaneously
  • Engineers edit AI agent skills using built-in monospaced editor
  • Mac users install skills from remote servers like OpenClaw
Similar Projects
  • OpenClaw - supplies remote skill repositories that Chops integrates with
  • Aider - supports skills but requires manual dotfile management
  • Cursor - offers built-in AI tools without cross-platform organization

Go CLI Automates App Store Connect Tasks 🔗

Scriptable tool simplifies automation of TestFlight builds and app submissions for Apple developers

rudrankriyam/App-Store-Connect-CLI · Go · 3.2k stars 2mo old

A new command-line tool offers Apple developers a fast way to interact with the App Store Connect API. The project, written in Go, focuses on scriptability and automation without interactive elements.

The CLI supports key operations such as uploading builds to TestFlight, submitting apps for review, managing code signing, pulling analytics, generating screenshots, and handling in-app subscriptions.

A new command-line tool offers Apple developers a fast way to interact with the App Store Connect API. The project, written in Go, focuses on scriptability and automation without interactive elements.

The CLI supports key operations such as uploading builds to TestFlight, submitting apps for review, managing code signing, pulling analytics, generating screenshots, and handling in-app subscriptions. It defaults to JSON output for machine readability while using tables for human interaction in terminals.

Installation is straightforward with Homebrew using the command brew install asc. Authentication requires an API key from Apple, configured with key ID, issuer ID and private key file. The tool includes commands to validate the setup and diagnose issues.

Version 0.45.2 brings refinements to submission verification, adds privacy advisory checks, and updates related tools for screenshots.

This CLI fits into existing developer workflows, allowing integration with IDEs, terminals and CI/CD systems. By removing the need for manual web portal interactions, it reduces errors in the release process for iOS, macOS, tvOS and visionOS applications.

Typical usage involves commands like asc apps list --output json to retrieve data in script-friendly format.

The JSON-first design makes it particularly suitable for automated pipelines where reliability and speed matter.

Use Cases
  • iOS developers automate TestFlight builds and beta submissions
  • DevOps engineers manage app metadata in CI environments
  • Mobile teams handle app signing and analytics retrieval
Similar Projects
  • fastlane - Ruby toolkit for broader iOS deployment automation
  • altool - Apple's former tool for App Store uploads
  • Transporter - desktop application for metadata and build delivery

OpenYak Delivers Local AI Agent for Desktop 🔗

Open-source application runs powerful AI assistant entirely on user hardware

openyak/desktop · Python · 371 stars 2d old

OpenYak is an open-source desktop AI assistant that executes entirely on the user's machine. Built in Python, the application lets users manage files, analyze data and automate office workflows without transmitting any information to the cloud.

It connects to more than 100 large language models through OpenRouter, including `Claude Opus 4.

OpenYak is an open-source desktop AI assistant that executes entirely on the user's machine. Built in Python, the application lets users manage files, analyze data and automate office workflows without transmitting any information to the cloud.

It connects to more than 100 large language models through OpenRouter, including Claude Opus 4.6, DeepSeek V3.2, Gemini 3 Flash and GPT-4.1. Users with existing ChatGPT subscriptions can link them directly, avoiding extra API costs. The system provides 16 built-in tools such as file read/write/edit, bash execution, glob search and web fetching.

Seven specialized agent modes support multi-step tool calling and sub-agent nesting for tasks ranging from planning to exploration. The application maintains complete privacy with zero telemetry and no cloud storage.

In operation, it performs batch file renaming and sorting across folders while generating auditable change logs. Data analysis features parse spreadsheets and CSVs locally to detect anomalies and export reports. Writing tools convert rough notes into polished documents that match company tone, while team functions merge PDFs, DOCX files and CSVs into structured briefings that extract action items and track deadlines.

A one-click secure tunnel with QR code access allows sessions to continue on mobile devices. Installers are available for Windows and macOS. The project offers one million free tokens weekly on basic models, with pay-as-you-go access at standard OpenRouter rates.

Use Cases
  • Office administrators automate file renaming with audit logs locally
  • Business analysts parse spreadsheets and generate reports privately
  • Team coordinators merge documents into action-item briefings offline
Similar Projects
  • Ollama - runs local models but lacks OpenYak's office automation tools
  • OpenInterpreter - supports code execution without native desktop interface
  • PrivateGPT - focuses on document querying with fewer agent modes

Open Source Standardizes Reusable Skills for Autonomous AI Agents 🔗

Developers are creating modular skill libraries and agent harnesses to enable sophisticated, self-improving AI systems across domains

An emerging pattern in open source reveals a decisive shift toward modular, skill-centric AI agent infrastructure. Rather than building monolithic AI applications, developers are decomposing agent capabilities into standardized, reusable skills that can be mixed, extended, and shared across platforms.

This cluster demonstrates several consistent technical themes.

An emerging pattern in open source reveals a decisive shift toward modular, skill-centric AI agent infrastructure. Rather than building monolithic AI applications, developers are decomposing agent capabilities into standardized, reusable skills that can be mixed, extended, and shared across platforms.

This cluster demonstrates several consistent technical themes. First is the formalization of skills as first-class primitives. Repositories such as anthropics/skills, mukul975/Anthropic-Cybersecurity-Skills (with 734+ MITRE-mapped skills), coreyhaines31/marketingskills, and kepano/obsidian-skills treat skills as composable units that agents can discover and invoke. VoltAgent/awesome-openclaw-skills has already cataloged over 5,400 filtered skills, showing the rapid maturation of skill registries.

Memory and context management represent another focal point. volcengine/OpenViking introduces a context database that uses a file-system paradigm for hierarchical delivery of memory, resources, and skills, enabling self-evolving behavior. vectorize-io/hindsight focuses on memory systems that learn from experience, while thedotmack/claude-mem automatically captures, compresses, and reinjects session context. wangziqi06/724-office combines three-layer memory with self-repair mechanisms in just 3500 lines of pure Python, supported by 26 tools and MCP/Skill plugins.

The pattern also emphasizes harnesses and orchestration. langchain-ai/deepagents provides an agent harness with planning tools, filesystem backends, and subagent spawning. bytedance/deer-flow acts as a SuperAgent research-and-code engine using sandboxes, memories, and subagents. msitarzewski/agency-agents and Donchitos/Claude-Code-Game-Studios demonstrate multi-agent coordination with specialized personalities and workflow hierarchies. Domain-specific implementations appear in TauricResearch/TradingAgents for financial trading and vxcontrol/pentagi for autonomous penetration testing.

Tooling for visibility, testing, and interoperability rounds out the ecosystem. mlflow/mlflow and Arize-ai/phoenix deliver observability and evaluation, while promptfoo/promptfoo brings systematic testing and red-teaming to agent workflows. Projects like Shpigford/chops and jarrodwatts/claude-hud help developers manage skills across Claude Code, Cursor, and similar environments.

Collectively, these efforts signal that open source is moving beyond prompt engineering toward production-grade agent operating systems. The future lies in composable architectures where agents dynamically acquire skills, maintain persistent memory, spawn specialized subagents, and self-repair—lowering the cost of building sophisticated autonomous systems while creating shared standards for safety, observability, and capability reuse.

Use Cases
  • Engineers assembling self-evolving agents for software development
  • Security teams deploying specialized agents for penetration testing
  • Analysts creating multi-agent systems for financial trading
Similar Projects
  • langchain-ai/deepagents - Delivers planning and subagent capabilities that mirror the modular skill pattern
  • mlflow/mlflow - Adds observability and evaluation layers to the emerging agent skill ecosystem
  • promptfoo/promptfoo - Provides systematic testing tools that support production-grade agent development

Open Source Tooling Layer Supercharges Commercial LLM Agents 🔗

From specialized skills to token-efficient proxies and unified memory systems, developers are building the middleware that turns proprietary models into production-ready agents.

The open source community is no longer primarily racing to build new foundational LLMs. Instead, a clear pattern has emerged: the rapid construction of a rich tooling and orchestration layer that augments existing commercial models like Claude, Gemini, and GPT with agentic capabilities, domain-specific skills, and practical engineering infrastructure.

This cluster reveals a maturing ecosystem focused on agent infrastructure.

The open source community is no longer primarily racing to build new foundational LLMs. Instead, a clear pattern has emerged: the rapid construction of a rich tooling and orchestration layer that augments existing commercial models like Claude, Gemini, and GPT with agentic capabilities, domain-specific skills, and practical engineering infrastructure.

This cluster reveals a maturing ecosystem focused on agent infrastructure. Repositories such as anthropics/skills and mukul975/Anthropic-Cybersecurity-Skills demonstrate the trend toward standardized, reusable capabilities. The latter offers over 734 structured cybersecurity skills mapped to MITRE ATT&CK, compatible across Claude Code, Gemini CLI, Cursor, and more than 20 platforms. Similarly, volcengine/OpenViking introduces a file-system-based context database that unifies memory, resources, and skills for agents, enabling hierarchical context delivery and self-evolving behavior.

Efficiency and integration form another major pillar. rtk-ai/rtk delivers a lightweight Rust CLI proxy that cuts LLM token usage by 60-90% on common developer commands. Tools like router-for-me/CLIProxyAPI and sub2api wrap multiple vendor CLIs into unified OpenAI-compatible endpoints, while qixing-jk/all-api-hub and QuantumNous/new-api provide centralized management for keys, usage dashboards, auto check-ins, and model redirection.

Agent frameworks showcase increasingly sophisticated orchestration. bytedance/deer-flow functions as a SuperAgent that researches, codes, and executes tasks over extended periods using sandboxes, memories, tools, and subagents. TauricResearch/TradingAgents applies multi-agent collaboration to financial trading, while mlflow/mlflow offers a full AI engineering platform for debugging, evaluating, and monitoring production agents. Security-conscious designs appear in nanoclaw, a containerized lightweight alternative to OpenClaw that connects to messaging platforms with built-in memory and scheduled jobs.

Testing and observability are also maturing, with promptfoo/promptfoo enabling systematic evaluation, red teaming, and performance comparison across models, and arize-ai/phoenix providing observability for RAG and agent systems.

Collectively, these projects signal where open source is heading: toward composable AI systems. The intelligence remains with proprietary frontier models, but the surrounding stack — tool calling interfaces, memory architectures, plugin systems, cost optimizers, and domain skill libraries — is being built in the open. This approach allows rapid specialization, better security boundaries, cost control, and integration with existing workflows without the expense of training new models.

The pattern suggests a maturing division of labor in AI development: closed models provide raw capability while open source supplies the engineering glue that makes them reliable, observable, and domain-adaptable.

Use Cases
  • Security teams automating penetration testing with MITRE-mapped skills
  • Developers reducing token costs via CLI proxies in daily workflows
  • Organizations deploying multi-agent trading or analysis systems
Similar Projects
  • LangChain - provides similar agent orchestration and tool abstractions but focuses more on Python-based chains than CLI integrations
  • CrewAI - specializes in multi-agent collaboration patterns comparable to deer-flow and TradingAgents
  • AutoGen - offers multi-agent conversation frameworks that complement the skills and memory systems seen here

AI Agents Reshape Open Source Web Frameworks 🔗

New tools blend browser protocols, in-page scripting and LLMs to make web interfaces directly controllable by artificial intelligence

An emerging pattern in open source reveals web frameworks evolving from static rendering engines into platforms that support autonomous AI agents. Rather than simply delivering pages to human users, these projects expose web interfaces for natural language control, automated manipulation, and intelligent orchestration.

The technical shift centers on combining established browser technologies with agentic capabilities.

An emerging pattern in open source reveals web frameworks evolving from static rendering engines into platforms that support autonomous AI agents. Rather than simply delivering pages to human users, these projects expose web interfaces for natural language control, automated manipulation, and intelligent orchestration.

The technical shift centers on combining established browser technologies with agentic capabilities. alibaba/page-agent demonstrates this by creating a JavaScript in-page GUI agent that lets LLMs control web interfaces through natural language instructions. Similarly, eze-is/web-access equips Claude Code with production-grade networking using three-layer channel scheduling, browser CDP, and parallel divide-and-conquer strategies, effectively giving AI models complete web browsing abilities.

This pattern appears across multiple implementations. D4Vinci/Scrapling offers an adaptive web scraping framework that scales from single requests to full crawls, while Leonxlnx/taste-skill focuses on "high-agency frontend" techniques that prevent AI-generated interfaces from becoming generic slop. tw93/Pake takes a different but related approach, turning any webpage into a desktop application with a single command, showing how web boundaries are being redefined.

Traditional web frameworks are also being reimagined for this new reality. UI5/openui5 continues providing enterprise-ready, responsive applications that run across devices and browsers, while digitallyinduced/ihp delivers a batteries-included Haskell framework optimized for long-term productivity and type safety. These projects coexist with AI-centric tools like langchain-ai/deepagents, which provides planning tools and subagent spawning for complex web tasks, and badlogic/pi-mono, an agent toolkit that includes web UI libraries.

Supporting infrastructure reveals another dimension of the trend. Multiple projects such as qixing-jk/all-api-hub, QuantumNous/new-api, and router-for-me/CLIProxyAPI create unified gateways that translate between different LLM APIs, making it easier to route web applications to various models. onyx-dot-app/onyx functions as an open source AI platform that works with every LLM, while DayuanJiang/next-ai-draw-io integrates natural language commands directly into diagram editing within a Next.js application.

Collectively, this cluster signals open source moving toward "agent-native" web architectures. The technical emphasis has shifted from pure UI rendering to exposing controllable surfaces, standardized protocols for AI observation and action, and frameworks that treat the web as both interface and execution environment. Developers are building systems where AI doesn't just consume web content but actively shapes and operates within it, pointing to a future where web applications are co-created and maintained by both humans and autonomous agents.

This evolution prioritizes practical interoperability—CDP integration, in-page execution, model-agnostic APIs, and tasteful frontend generation—over isolated innovation. The result is a more dynamic web stack where frameworks must accommodate both human users and AI collaborators from the ground up.

Use Cases
  • Developers building natural language controllable web interfaces
  • Teams creating AI-augmented enterprise responsive applications
  • Researchers automating adaptive web scraping and data tasks
Similar Projects
  • Playwright - Modern browser automation library that provides low-level control but lacks the natural language agent layer these projects add
  • Next.js - Popular React framework for web applications that excels at rendering but does not natively support in-page GUI agents or CDP orchestration
  • Selenium - Long-standing web testing tool focused on scripted automation rather than LLM-driven autonomous web interaction

Deep Cuts

Autonomous AI Research Skill Changes Everything 🔗

This under-the-radar project equips agents with powerful independent research abilities

olelehmann100kMRR/autoresearch-skill · Unknown · 476 stars

Deep in the GitHub archives lies autoresearch-skill, a project that could revolutionize how we build intelligent AI systems. This clever skill empowers AI agents to take control of the research process from start to finish.

Using sophisticated prompting and tool integration, autoresearch-skill allows models to break down queries, identify knowledge gaps, seek out reliable sources, and compile findings into actionable insights.

Deep in the GitHub archives lies autoresearch-skill, a project that could revolutionize how we build intelligent AI systems. This clever skill empowers AI agents to take control of the research process from start to finish.

Using sophisticated prompting and tool integration, autoresearch-skill allows models to break down queries, identify knowledge gaps, seek out reliable sources, and compile findings into actionable insights. It's like giving your AI a built-in research assistant that never sleeps.

Builders should pay close attention because the capabilities extend far beyond basic search functionality. The skill emphasizes critical analysis and synthesis, producing outputs that feel remarkably human-like in their depth and coherence.

In practice, this means faster iteration cycles for anyone working with emerging technologies or complex subject matter. The modular nature makes it simple to adapt for specific industries or use cases.

What makes this project particularly exciting is its potential to democratize access to high-quality research. Small teams can now leverage enterprise-level information gathering without massive overhead.

As we move toward more autonomous AI applications, autoresearch-skill represents an important building block. Its existence highlights the creative ways developers are extending large language model capabilities.

If you're working on agent-based systems or knowledge tools, this is one to explore immediately. You might just find the key to unlocking your project's full potential.

Use Cases
  • AI developers integrating advanced research into autonomous agents
  • Academic researchers accelerating literature reviews with automated synthesis
  • Businesses performing market analysis with intelligent data gathering
Similar Projects
  • Auto-GPT - delivers general autonomy rather than focused research skills
  • LangChain - offers research tools but lacks this dedicated skill module
  • MetaGPT - provides multi-agent frameworks compared to this specialized approach

Quick Hits

weixin-agent-sdk TypeScript SDK for building WeChat agents with simplified APIs for rapid intelligent automation development. 620
skills C# library from MiniMax AI that adds custom skills and advanced capabilities to your AI applications. 2.5k
claude-plugin-weixin TypeScript plugin that integrates Claude AI with Weixin for intelligent messaging and automation features. 477
tiktok-views-likes-followers-shares-bot-2026 Windows bot automates TikTok views, likes, shares and followers using CAPTCHA solving and proxy rotation. 408
claude-peers-mcp TypeScript tool lets multiple Claude AI instances message each other ad-hoc for seamless collaboration. 704
filmkit Create and manage Fujifilm camera profiles instantly by connecting your camera with this speedy toolkit. 316
MicroWARP 🚀 An 800KB RAM ultra-lightweight Cloudflare WARP SOCKS5 proxy in Docker. 仅需 800KB 内存的纯内核态 Cloudflare WARP 代理 - Docker 529

Deep-Live-Cam 2.7 Beta Adds Real-Time Face Enhancement 🔗

Major update delivers new enhancers, selective masks and multi-GPU support for live face swapping workflows

hacksider/Deep-Live-Cam · Python · 80.3k stars Est. 2023 · Latest: 2.7-beta

Deep-Live-Cam has released version 2.7 beta, introducing a suite of performance and quality improvements to its real-time face swap and one-click video deepfake pipeline. The Python project, which has matured over two years, continues to focus on a core technical constraint: performing convincing face replacement using only a single source image rather than large training datasets.

Deep-Live-Cam has released version 2.7 beta, introducing a suite of performance and quality improvements to its real-time face swap and one-click video deepfake pipeline. The Python project, which has matured over two years, continues to focus on a core technical constraint: performing convincing face replacement using only a single source image rather than large training datasets.

The update centers on several new components. A realtime face enhancer now runs at 27 frames per second on suitable hardware, paired with an inswapper optimizer that reduces latency. Two additional models, GPEN 512 and GPEN 256, expand the enhancement options, while a face enhancer scaler gives developers precise control over output resolution. Interpolation has been added to smooth frame transitions, and the introduction of quick lip mask, lip mask, chin mask and eyes mask allows selective editing of facial regions without affecting the entire frame.

Multi-GPU support arrives through a new GPU changer, enabling workload distribution across discrete NVIDIA or AMD cards. LUTs provide straightforward color grading, and window projection combined with in-window preview creates a more responsive workflow for monitoring live output. Other practical additions include camera refresh, resolution changer, and full-screen realtime video watching mode.

The software processes webcam input or video files by combining face detection, landmark alignment, and generative models based on GAN architectures. Because it requires only one reference image, it lowers the barrier for rapid iteration compared with traditional deepfake pipelines that demand hours of training. Pre-built binaries for Windows, Mac Silicon, and CPU-only systems allow developers to bypass complex local compilation.

Ethical constraints remain embedded in the codebase. Built-in checks block processing of nudity, graphic content, or sensitive material such as war footage. The maintainers emphasize that users must obtain consent when swapping real individuals and label outputs as deepfakes when published. The project explicitly states it is intended for legitimate creative work in the AI-generated media industry, including character animation and content prototyping.

For builders working on interactive AI applications, virtual production, or experimental media tools, the 2.7 release addresses previous pain points around speed, visual quality, and fine-grained control. The combination of real-time performance, selective masking, and multi-GPU scaling makes the tool more suitable for integration into larger creative pipelines while maintaining the single-image simplicity that defined its original value proposition.

As regulatory and ethical debates around synthetic media intensify, Deep-Live-Cam's transparent approach to safeguards and responsible-use requirements offers a practical reference for developers navigating this complex space.

Use Cases
  • Video artists animating custom characters in real time
  • Developers prototyping interactive face-swap webcam apps
  • Filmmakers testing quick deepfake concepts with single images
Similar Projects
  • Roop - simplifies one-click deepfakes but lacks Deep-Live-Cam's realtime enhancer and selective masking tools
  • DeepFaceLab - requires extensive model training unlike Deep-Live-Cam's single-image realtime approach
  • FaceFusion - provides strong post-processing but offers less optimized live webcam performance than version 2.7

More Stories

OpenBB Ships Stable Open Data Platform 🔗

Version 1.0.1 unifies data integration across Python, Workspace and AI agent workflows

OpenBB-finance/OpenBB · Python · 63.4k stars Est. 2020

OpenBB has released v1.0.1, the first stable version of its Open Data Platform.

OpenBB has released v1.0.1, the first stable version of its Open Data Platform. The update solidifies the project's position as essential infrastructure for financial data workflows.

The ODP functions as a consolidation layer for proprietary, licensed and public data sources. It exposes unified interfaces to Python environments favored by quants, OpenBB Workspace for analysts, MCP servers powering AI agents, and REST APIs for other systems.

Getting started requires minimal effort. After pip install "openbb[all]", users launch the backend with openbb-api, creating a FastAPI server on port 6900. Connection to OpenBB Workspace involves entering the local URL in the Apps section.

Data access follows intuitive patterns. The example obb.equity.price.historical("AAPL") returns structured output convertible to pandas DataFrames for further analysis or machine learning models.

This release matters amid growing demand for AI-assisted financial research. By reducing data integration friction, ODP allows teams to focus on insights rather than plumbing. Support spans equities, fixed-income, derivatives, options and cryptocurrencies.

Use Cases
  • Quants building models with unified access to market data sources
  • Analysts integrating backends into OpenBB Workspace dashboards
  • Developers deploying AI agents via standardized MCP data feeds
Similar Projects
  • yfinance - simpler stock data retrieval without multi-surface integration
  • pandas-datareader - limited sources and no AI agent or Workspace support
  • QuantLib - focuses on derivatives pricing rather than data consolidation

OpenClaw Release Tightens Session and Platform Stability 🔗

Recovery version fixes session resets, Telegram security and Discord metadata handling

openclaw/openclaw · TypeScript · 331k stars 3mo old

OpenClaw has shipped v2026.3.13-1, a targeted recovery release that corrects several stability problems in its self-hosted personal AI assistant.

OpenClaw has shipped v2026.3.13-1, a targeted recovery release that corrects several stability problems in its self-hosted personal AI assistant.

The update preserves lastAccountId and lastThreadId across session resets, preventing context loss for long-running conversations. Telegram media handling now correctly threads transport policy into SSRF checks, while Discord gateway metadata fetches gracefully manage transient failures. Post-compaction logic now uses full-session token counts for accurate sanity validation, and delivery deduplication logic has been refined.

These changes arrive alongside an updated default test model (gpt-5.4) and improved CLI help text for high-reasoning modes. The project continues to run on Node 24, with openclaw onboard guiding users through workspace, channel and skill configuration across macOS, Linux and Windows via WSL2.

OpenClaw positions itself as a single-user, always-on assistant that meets users on their existing channels—WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Matrix, Teams and more than a dozen others. It adds native speech on macOS, iOS and Android plus a controllable live Canvas. The gateway functions strictly as control plane; the assistant itself remains the product.

For developers and power users who want local speed, data ownership and minimal prompt-injection surface, the release reinforces reliability without altering the core architecture.

Use Cases
  • Professionals routing AI help across Slack and WhatsApp
  • Developers debugging via persistent Discord threads
  • Mobile users accessing voice AI on iOS and Android
Similar Projects
  • Leon - offers multi-channel bots but lacks OpenClaw's session resilience
  • Open Interpreter - excels at code execution yet provides no native messaging integration
  • Auto-GPT - focuses on autonomous agents rather than always-on chat presence

Gemini CLI Release Refines UI and Agent Tools 🔗

Version 0.34.0 adds tracker visualization and extends operation timeouts

google-gemini/gemini-cli · TypeScript · 98.7k stars 11mo old

Gemini CLI's v0.34.0 release delivers targeted improvements to its terminal-based AI agent.

Gemini CLI's v0.34.0 release delivers targeted improvements to its terminal-based AI agent. The update introduces a chat resume footer displayed on session quit, helping users reconnect to prior context without manual reconstruction.

User interface changes standardize semantic focus colors and increase history visibility. SVG snapshot rendering now supports bold and additional text styles. Core stability received attention through extra safety checks against proto pollution and an increased A2A agent timeout set to 30 minutes.

New tracker CRUD tools with built-in visualization give developers direct control over structured data during extended sessions. These additions sit alongside existing capabilities including Google Search grounding, file system operations, shell command execution and web fetching.

The CLI continues to offer access to Gemini 3 models with a 1M token context window. Free-tier users retain 60 requests per minute and 1,000 requests daily using a personal Google account. MCP support allows custom extensions through the Model Context Protocol.

Weekly release cadence persists, with preview builds published Tuesdays and stable versions following. Installation remains straightforward via npx @google/gemini-cli, global npm, Homebrew or MacPorts. The changes reflect incremental refinement rather than wholesale redesign, addressing practical friction points reported by terminal-heavy users.

(178 words)

Use Cases
  • Developers executing shell commands through natural language prompts
  • Engineers managing structured data with tracker CRUD visualization
  • Teams building custom integrations using MCP protocol extensions
Similar Projects
  • shell-gpt - Supports multiple LLMs but lacks native Gemini 1M context
  • aider - Focuses on git-based code editing rather than general terminal tasks
  • claude-cli - Provides Anthropic models with comparable tools but no MCP support

Quick Hits

AutoGPT AutoGPT gives builders tools to create autonomous AI agents that handle complex tasks independently so you can focus on innovation. 182.7k
open-webui Open-WebUI delivers a polished, user-friendly interface for running local LLMs with seamless Ollama and OpenAI API support. 128.3k
dify Dify provides a production-ready platform to design, orchestrate, and scale sophisticated agentic AI workflows. 134.1k
firecrawl Firecrawl turns any website into clean LLM-ready markdown or structured data through a simple, powerful web scraping API. 96.7k
prompts.chat Prompts.chat lets you discover, share, and self-host a library of battle-tested AI prompts with complete organizational privacy. 153.9k

MapAnything v1.1.1 Expands Metric 3D Reconstruction Pipeline 🔗

Latest release adds Pixio encoder and MPS inference support to universal feed-forward model handling twelve distinct 3D tasks

facebookresearch/map-anything · Python · 3k stars 6mo old · Latest: v1.1.1

MapAnything has received its first significant update since launch with the release of v1.1.1.

MapAnything has received its first significant update since launch with the release of v1.1.1. The new version introduces a Pixio encoder implementation and adds MPS inference capabilities, allowing the framework to run efficiently on Apple silicon hardware.

At its core, MapAnything provides a single transformer model trained end-to-end to regress factored metric 3D geometry from varied inputs including images, calibration data, poses, or depth maps. Unlike traditional pipelines that require separate specialized systems for different tasks, this feed-forward approach supports more than twelve distinct 3D reconstruction problems through one unified architecture. These include multi-image structure-from-motion, multi-view stereo, monocular metric depth estimation, registration, and depth completion.

The framework distinguishes itself through its complete stack. It supplies modular components for data processing, training, inference, and profiling, enabling researchers to swap between different underlying models through a consistent interface. Supported backends now include VGGT, DUSt3R, MASt3R, MUSt3R, Pi3-X, and additional architectures. This interoperability lets teams experiment with multiple approaches without rewriting downstream code.

The v1.1.1 release specifically adds the Pixio encoder contribution from new community member xlei77, alongside MPS support implemented by matthewijordan. These changes address practical deployment needs for developers working on resource-constrained or Apple-based development environments. The project maintains its focus on metric accuracy rather than scale-ambiguous reconstruction, producing outputs suitable for robotics and downstream graphics applications.

Integration features continue to mature. The framework exports directly to COLMAP format, supports visualization through Rerun, and includes pathways for Gaussian Splatting pipelines. A unified output format ensures consistency regardless of the chosen backend model or input modality. Input requirements remain clearly documented, with the system handling both single-image and multi-view scenarios while respecting provided calibration when available.

For builders working at the intersection of computer vision and robotics, these updates reduce friction when deploying learned 3D reconstruction across heterogeneous hardware. The modular design particularly benefits teams that need to benchmark multiple state-of-the-art models against the same evaluation suite or production requirements. Profiling tools included in the repository allow quantitative comparison between the native MapAnything transformer and external models on identical workloads.

The project, a collaboration between Meta researchers and Carnegie Mellon University, reflects an ongoing shift toward universal, task-agnostic 3D perception models. Rather than fragmenting efforts across specialized repositories, MapAnything consolidates the infrastructure required to train, evaluate, and deploy metric reconstruction systems in one coherent framework.

Use Cases
  • Robotics engineers reconstructing metric scenes from multi-view images
  • AR developers generating accurate depth from monocular camera input
  • Autonomous teams completing sparse depth maps for navigation
Similar Projects
  • DUSt3R - serves as an interchangeable backend within MapAnything's unified interface rather than a standalone solution
  • MASt3R - focuses on multi-view stereo while MapAnything generalizes across twelve reconstruction tasks with consistent outputs
  • COLMAP - provides classical SfM methods that MapAnything complements through learned metric regression and direct export compatibility

More Stories

OM1 v1.0.1 Unifies Single and Multi-Agent Modes 🔗

Update improves switching stability and adds full autonomy for Unitree G1

OpenMind/OM1 · Python · 2.7k stars Est. 2025

OpenMind has released version 1.0.1 of OM1, its modular AI runtime for robots.

OpenMind has released version 1.0.1 of OM1, its modular AI runtime for robots. The update unifies single and multi-agent operation and significantly improves the stability of mode switching, addressing long-standing friction for developers moving between configurations.

Written in Python, OM1 lets teams build multimodal agents that consume camera feeds, LIDAR, web data and social media, then issue movement, speech and navigation commands. Hardware connections are handled through plugins for ROS2, Zenoh and CycloneDDS, with the project recommending Zenoh for new work. A local WebSim interface at http://localhost:8000/ provides real-time visual debugging of agent decisions and timing.

This release incorporates full autonomy for the Unitree G1 humanoid, updates to the LLM and simulator stacks, improved TTS interruption handling, and a new configuration validation CLI. Background processes were refactored with Pydantic, multiple documentation and code typos were fixed, and CI checks now include Vulture and automated typo detection.

The changes matter because robot platforms are proliferating across different physical forms. OM1’s plugin architecture and pre-configured endpoints for OpenAI, Anthropic, xAI, Ollama and multiple VLMs allow the same high-level agent logic to be redeployed with minimal rework when hardware changes.

**

Use Cases
  • Developers deploying autonomous agents on Unitree G1 humanoids
  • Engineers testing multimodal models in Gazebo and Isaac Sim
  • Teams adding new sensors to ROS2 robots via OM1 plugins
Similar Projects
  • ROS2 - core middleware that OM1 layers AI agents and LLMs on top of
  • LangGraph - multiagent orchestration framework without native robot hardware plugins
  • NVIDIA Isaac - strong simulation tools but less modular LLM integration than OM1

HORUS 0.1.9 Refines Scheduler for Real-Time Robotics 🔗

Major cleanup removes 77,000 lines of dead code and makes real-time features explicitly opt-in

softmata/horus · Rust · 287 stars 5mo old

HORUS 0.1.9, released in February 2026, represents a substantial refinement of the real-time robotics middleware rather than a new project.

HORUS 0.1.9, released in February 2026, represents a substantial refinement of the real-time robotics middleware rather than a new project. The update eliminates 77,000 lines of dead code, resolves 43 bugs and security issues, and redesigns the Scheduler API for clarity and zero-cost defaults.

Scheduler::new() no longer applies real-time priority, memory locking or CPU pinning automatically. Developers must now opt in explicitly or use one of the new presets:

  • Scheduler::deploy() for production
  • Scheduler::safety_critical() for watchdog-protected execution
  • Scheduler::deterministic() for reproducible timing

This change removes unnecessary syscalls from non-RT code paths while preserving HORUS's core architecture: shared-memory ring buffers that deliver 11–196 ns IPC latency, compared with ROS2's 50–500 µs using DDS. The system maintains five deterministic execution classes, built-in budget and deadline enforcement, and graduated watchdog protection with automatic safe-state transitions.

Python bindings via pip install horus-robotics continue to support DLPack zero-copy tensors, allowing PyTorch or JAX models to share memory directly with real-time motor control loops written in Rust. Configuration remains a single horus.toml file, and the horus new + horus run workflow avoids ROS2's multi-toolchain complexity.

The breaking scheduler changes require updates for existing users but deliver a cleaner, more predictable foundation for production robotics.

Use Cases
  • Robotics programmers implementing 1kHz control loops on humanoid robots
  • Engineers developing real-time AI systems for autonomous drone fleets
  • Teams creating safety-critical systems for autonomous vehicle platforms
Similar Projects
  • ROS2 - uses DDS resulting in 575x higher IPC latency
  • Apollo Cyber - lacks HORUS's lock-free shared memory and Rust core
  • OROCOS - provides real-time features but requires more complex setup

Zenoh 1.8.0 Refines Unified Data Protocol Features 🔗

TOML configuration and connectivity API enhance usability across distributed edge deployments

eclipse-zenoh/zenoh · Rust · 2.6k stars Est. 2020

Eclipse Zenoh released version 1.8.0 this week, delivering targeted improvements to its protocol for unifying data in motion, data at rest, and computations.

Eclipse Zenoh released version 1.8.0 this week, delivering targeted improvements to its protocol for unifying data in motion, data at rest, and computations. The update adds native support for TOML configuration files, simplifying zenohd router setup in heterogeneous environments. A new Connectivity API exposes link and transport details through Session::info(), giving applications better visibility into their network state.

Other changes include stabilization of accept_replies, enforcement of original query QoS on responses, and the removal of the internal_config feature. Bug fixes address a liveliness token memory leak and duplicated delete link events on transport closure.

Written in Rust, Zenoh blends traditional pub/sub semantics with geo-distributed storage, queries, and computations while maintaining minimal overhead. The core zenoh crate serves as the reference implementation, with zenoh-ext providing advanced publishers, subscribers, and lightweight serialization. The standalone zenohd router and plugin system support infrastructure at scales from embedded devices to cloud backends.

For teams operating across edge, IoT, and robotics domains, these refinements reduce configuration complexity and improve operational reliability without sacrificing the protocol's signature efficiency. The project, now over six years old, continues to focus on concrete performance gains for production distributed systems.

(178 words)

Use Cases
  • Robotics teams route ROS2 traffic with geo-distributed queries
  • Edge devices store sensor data and execute local computations
  • IoT platforms unify messaging across embedded and cloud nodes
Similar Projects
  • MQTT - basic pub/sub without native geo-storage or queries
  • Apache Kafka - stream processing at scale with heavier resource use
  • DDS - real-time messaging for robotics but limited compute integration

Quick Hits

foxglove-sdk Rust SDK for building high-performance multimodal visualization platforms, delivering fast interactive tools for robotics data exploration. 205
generate_parameter_library Python library that declaratively generates ROS 2 parameters, simplifying configuration and boosting reliability in robot systems. 357
autoware_universe Modular C++ autonomous driving framework with production-ready perception, planning, and control stacks for self-driving vehicles. 1.6k
chrono High-performance C++ library for multiphysics and multibody dynamics simulations, enabling accurate modeling of robots and vehicles. 2.8k
ros-noetic Vinca config files for building ROS Noetic via RoboStack, bringing easy conda-based ROS to any operating system. 515

Microsoft Sentinel Unifies Defender Content for Cross-Platform Hunting 🔗

Repository merges SIEM detections and Microsoft 365 XDR queries to give builders consistent analytics across cloud and productivity environments.

Azure/Azure-Sentinel · Python · 5.5k stars Est. 2018

Microsoft Sentinel's GitHub repository has become the single source for security content across both its cloud-native SIEM and Microsoft 365 Defender platforms. The unification delivers a tightly integrated collection of out-of-the-box detections, exploration queries, hunting queries, workbooks, and playbooks that security teams can deploy immediately rather than building from scratch.

The repository's hunting queries now cover advanced scenarios that execute in both Sentinel and Microsoft 365 Defender.

Microsoft Sentinel's GitHub repository has become the single source for security content across both its cloud-native SIEM and Microsoft 365 Defender platforms. The unification delivers a tightly integrated collection of out-of-the-box detections, exploration queries, hunting queries, workbooks, and playbooks that security teams can deploy immediately rather than building from scratch.

The repository's hunting queries now cover advanced scenarios that execute in both Sentinel and Microsoft 365 Defender. This removes the previous need to maintain separate logic for each console. Security engineers can write a single KQL query that surfaces anomalies in email, endpoint, identity, and cloud logs without translation or duplication. Workbooks provide pre-built visualizations of the resulting data, while playbooks automate response actions through native Azure integration.

For builders, the value lies in the reusable components. A detection rule written here can trigger analytics rules in Sentinel, surface in advanced hunting within the Microsoft 365 Defender portal, and feed into automated investigation workflows. The project solves the persistent problem of fragmented security content that forces teams to reinvent detection logic for every new data source they onboard.

Contribution remains central to the project's direction. Microsoft requires a Contributor License Agreement for most submissions and provides clear guidelines for adding content through GitHub. Builders can propose new hunting queries, update existing playbooks, or contribute workbooks that address emerging attack techniques. Issues filed against the repository are actively triaged, with templates for both bugs and feature requests.

The repository continues to see regular updates, with the most recent commits in March 2026 adding fresh content for current threat patterns. This steady evolution reflects Microsoft's broader strategy of converging SIEM and XDR capabilities into a more cohesive security operations platform.

Security teams no longer need to choose between depth in Sentinel analytics or breadth in Microsoft 365 visibility. The unified repository gives them both from one well-maintained source. Documentation links, community forums for SIEM/SOAR and XDR discussions, and direct feedback channels to the product team complete the support ecosystem.

The result is a practical, production-grade resource that lets developers focus on custom logic and threat modeling instead of maintaining basic detection coverage. In an environment where attack surfaces span Azure, Microsoft 365, and hybrid infrastructure, shared content that works across both platforms has become essential rather than optional.

(Word count: 378)

Use Cases
  • Security engineers deploying shared detection rules
  • Analysts running cross-platform advanced hunting queries
  • Developers automating responses with reusable playbooks
Similar Projects
  • elastic/detection-rules - Maintains community detection rules for the Elastic SIEM with similar query-based approach but focused on open-source observability data
  • splunk/splunk-security-content - Delivers pre-built analytic stories and searches specifically engineered for the Splunk platform and its data schema
  • SigmaHQ/sigma - Offers vendor-neutral detection signatures that convert to multiple SIEM languages including Sentinel's KQL

More Stories

x64dbg Release Strengthens Stability for Analysts 🔗

August update resolves migration bugs and relocates documentation to repository

x64dbg/x64dbg · C++ · 47.9k stars Est. 2015

x64dbg's August 2025 release focuses on long-term stability after its migration to Visual Studio 2022. The update corrects several regressions that slipped through, including crashes on systems with older Visual C++ Redistributable packages, completely broken pattern finding, and AVX-512 compatibility failures in the 32-bit debugger. Register display for XMM components on AVX-enabled CPUs has also been fixed.

x64dbg's August 2025 release focuses on long-term stability after its migration to Visual Studio 2022. The update corrects several regressions that slipped through, including crashes on systems with older Visual C++ Redistributable packages, completely broken pattern finding, and AVX-512 compatibility failures in the 32-bit debugger. Register display for XMM components on AVX-enabled CPUs has also been fixed.

The project is expanding its automated testing efforts, with the recently introduced headless mode serving as an early building block. Support for AddressSanitizer has been added to detect memory safety problems before they reach users.

Documentation has been moved into the repository's docs folder. The change simplifies maintenance and makes technical content more accessible to modern query tools.

As a mature open-source user-mode debugger for Windows, x64dbg continues to provide disassembly, tracing, and breakpoint capabilities optimized for reverse engineering and malware analysis. It supports both x86 and x64 binaries through a modular architecture with an extensive plugin system. Users launch x32dbg.exe or x64dbg.exe after extracting a snapshot, with x96dbg.exe offering architecture selection.

The community-driven project accepts pull requests and maintains active development through contributions from multiple long-term maintainers.

(178 words)

Use Cases
  • Malware analysts debugging Windows executables without source code
  • Security researchers tracing exploits during dynamic binary analysis
  • CTF participants reverse engineering challenge binaries in competitions
Similar Projects
  • OllyDbg - predecessor limited to 32-bit Windows debugging
  • Ghidra - NSA tool focused on static decompilation
  • WinDbg - Microsoft debugger with deep kernel integration

PhoneSploit Pro Update Improves Audio Feature Stability 🔗

Version 1.61 adds Android version detection to prevent crashes on older devices

AzeemIdrisi/PhoneSploit-Pro · Python · 5.7k stars Est. 2022

PhoneSploit Pro has released version 1.61, refining stability for its newer capabilities. The Python tool now detects the target device's Android version before attempting audio streaming or recording.

PhoneSploit Pro has released version 1.61, refining stability for its newer capabilities. The Python tool now detects the target device's Android version before attempting audio streaming or recording. This feature works only on Android 11 and above. Incompatible devices return control to the main menu instead of triggering a crash.

The project combines ADB and Metasploit-Framework to automate remote exploitation. When a device exposes TCP port 5555, it automatically fetches the tester's IP address, generates a payload with msfvenom, installs the APK, and launches msfconsole to deliver a Meterpreter session. This end-to-end process requires no manual command entry.

Beyond exploitation, the tool functions as a full ADB utility over Wi-Fi or USB. It supports taking screenshots that transfer automatically, recording the screen for a set duration, uploading and downloading files and folders, installing APKs, listing installed applications, accessing the shell, and rebooting into System, Recovery, Bootloader or Fastboot modes.

The update, released this month, addresses practical issues encountered during testing. By handling version compatibility checks, it reduces interruptions for security teams evaluating Android device posture. The project continues to simplify penetration testing and vulnerability assessment by managing the complex steps of payload delivery and session establishment.

PhoneSploit Pro remains focused on making mobile security evaluation accessible without requiring deep expertise in individual tool arguments.

Use Cases
  • Penetration testers gaining Meterpreter sessions on exposed Android devices
  • Security auditors transferring files and recording screens via ADB remotely
  • Developers managing APK installation and device reboots over WiFi connections
Similar Projects
  • aother/PhoneSploit - original version lacking automated Metasploit integration
  • AhMyth - GUI-based Android RAT without deep ADB and msfvenom automation
  • rapid7/metasploit-framework - requires manual payload creation and setup steps

Cilium 1.19.1 Fixes Cluster Mesh and SR-IOV Bugs 🔗

Patch release resolves permissions, panics and mutex contention in production Kubernetes environments

cilium/cilium · Go · 24k stars Est. 2015

Cilium has released v1.19.1, delivering several targeted bug fixes to its eBPF dataplane.

Cilium has released v1.19.1, delivering several targeted bug fixes to its eBPF dataplane.

The update corrects CRD update permissions for the MCS-API in clustermesh, eliminating installation failures in multi-cluster deployments. It also prevents a datapath reinitialization panic that occurred when a required DirectRouting device was missing.

Helm users with operator.enabled=false no longer encounter RBAC errors after conditional Role and RoleBinding logic was aligned. On SR-IOV nodes, the release reduces rtnl_mutex contention by skipping unnecessary VF information requests during netlink RTM_GETLINK operations, improving performance under load.

CI workflow changes simplify kernel testing job names and refine Ariane automation to skip irrelevant runs during LVH kernel updates. Dependency updates to GitHub Actions and other packages keep the Go codebase current.

These fixes matter for operators running Cilium at scale. The project continues to provide identity-based L3-L7 network policies, kube-proxy replacement via eBPF hash tables, and integrated observability while maintaining the last three minor versions as stable.

v1.19.1 represents the type of incremental hardening required for production networking infrastructure built on kernel-level technology.

Use Cases
  • Platform teams securing multi-cluster Kubernetes networks
  • Infrastructure engineers optimizing SR-IOV node performance
  • Operators replacing kube-proxy in large-scale deployments
Similar Projects
  • Calico - offers Kubernetes CNI with iptables or eBPF but simpler policy model
  • Antrea - provides OVS-based networking without Cilium's L7 awareness
  • Istio - delivers service mesh capabilities but relies on separate dataplane

Quick Hits

sherlock Sherlock hunts social media accounts by username across hundreds of networks, streamlining OSINT reconnaissance for investigators. 74k
hacktricks HackTricks delivers battle-tested techniques from CTFs, real-world apps, and security research to level up your hacking skills. 11.1k
authelia Authelia provides OpenID Certified single sign-on with multi-factor authentication, simplifying secure access for web apps. 27.3k
nDPI nDPI equips developers with open-source deep packet inspection to detect protocols and analyze network traffic in real time. 4.4k
HackBrowserData HackBrowserData extracts and decrypts passwords, cookies, and history from browsers across Windows, macOS, and Linux. 13.6k

Gitea 1.25.5 Hardens Security Across Core Functions 🔗

Latest release patches redirect bypasses, permission flaws and OAuth2 handling

go-gitea/gitea · Go · 54.4k stars Est. 2016 · Latest: v1.25.5

Gitea v1.25.5 ships with an extensive security update that addresses 15 distinct issues in the self-hosted Git platform.

Gitea v1.25.5 ships with an extensive security update that addresses 15 distinct issues in the self-hosted Git platform.

The release advances the Go toolchain to 1.25.8 and closes a redirect bypass that exploited backslash-encoded paths. Permission checks have been corrected for release drafts, track time entries and pull request branch updates. A flaw that let users change another account’s primary email has been fixed, while OAuth2 authorization codes now enforce proper expiry and prevent reuse.

Repository creation fields face new validation constraints. Mirror LFS synchronization uses the correct HTTP transport, and organization API endpoints tighten visibility rules for hidden members and private groups. Path resolution, release asset dumping and forwarded proto detection for public URLs also received fixes. A timeout now limits git grep operations.

On the enhancement side, the security-check target is now informational only and several dependencies, including github.com/cloudflare/circl 1.6.3, have been upgraded.

Written in Go, the platform delivers Git hosting, code review, team collaboration, npm and Maven package registries, and CI/CD pipelines in a single binary that runs on Linux, macOS and Windows across x86, ARM and PowerPC. For teams managing their own infrastructure, these patches reduce exposure in production environments at a time when self-hosted tools face sustained scrutiny.

(178 words)

Use Cases
  • Development teams hosting private Git repositories on-premises
  • Organizations running internal npm and Maven package registries
  • DevOps groups implementing self-managed CI/CD pipelines
Similar Projects
  • GitLab - heavier self-hosted option with broader built-in DevOps tools
  • Gogs - minimalist predecessor fork with far less active maintenance
  • Forgejo - community fork focused on governance and independence

More Stories

Geth v1.17.1 Bolsters Ethereum Node Security and Stability 🔗

Bug fixes target snap sync regression and prepare for upcoming Amsterdam fork

ethereum/go-ethereum · Go · 50.9k stars Est. 2013

Geth v1.17.1, released this week, is a recommended bug-fix update for all Ethereum node operators.

Geth v1.17.1, released this week, is a recommended bug-fix update for all Ethereum node operators. The maintenance release resolves a snap-sync regression introduced in v1.17.0 that caused failures for users running --history.chain=postmerge. It also patches several security issues.

Core engine API changes disable plain-text HTTP2 following compatibility reports with Teku clients. Operators gain a new --metrics.influxdb.interval flag for tuning reporting cadence, while the inspect-trie subcommand now prints node counts at every state trie depth.

Amsterdam fork preparations dominate the core updates. These include the EIP-7834 SLOTNUM opcode, activation of EIP-8024, and multiple fixes to EIP-7928 block-level access lists. Payload building logic was corrected to ensure the consensus layer always receives the latest payloads.

Networking improvements drop support for the eth/68 p2p protocol version, fix a rare crash in request tracking, and optimize the blobpool to reduce relay of transactions unlikely to be included. The transaction pool now filters recent chain inclusions more effectively, lowering unnecessary network traffic.

As the reference Go implementation of Ethereum's execution layer, go-ethereum supplies the geth client, clef signer, abigen bindings generator and other utilities. Binaries are available at geth.ethereum.org. Node operators should update promptly to maintain sync reliability and network compatibility.

(178 words)

Use Cases
  • Node operators deploying full Ethereum nodes on mainnet infrastructure
  • Developers integrating dApps with JSON-RPC endpoints from geth clients
  • Protocol engineers implementing Amsterdam fork changes in execution clients
Similar Projects
  • Nethermind - .NET Ethereum client popular in Windows environments
  • Hyperledger Besu - Java-based client focused on enterprise modularity
  • Erigon - Go implementation optimized for storage efficiency and speed

gRPC 1.78.1 Fixes Python Stability Issues 🔗

Maintenance release reverts changes causing multithreaded hangs and adds Ruby 4.0 support

grpc/grpc · C++ · 44.5k stars Est. 2014

The gRPC team has released version 1.78.1, delivering targeted stability fixes for its Python implementation and expanded language support.

The gRPC team has released version 1.78.1, delivering targeted stability fixes for its Python implementation and expanded language support. The update reverts two recent changes that created production issues.

An unintentional log warning directing messages to STDERR before absl::InitializeLog() initialization has been restored. More critically, inconsistent GRPC_ENABLE_FORK_SUPPORT defaults between gRPC Core and Python were corrected. This resolves request processing hangs in multithreaded environments that produced the error "Other threads are currently calling into gRPC, skipping fork() handlers," addressing issue #37710.

The problematic Python package was yanked from PyPI in February due to these regressions. The release also modernizes the Python API documentation site.

For Ruby users, version 1.78.1 adds build and test coverage for Ruby 4.0 along with native gem support. These refinements maintain the framework's reputation for reliable performance in production systems.

Now over a decade old, gRPC's C++ core continues powering transparent client-server communication across languages. The high-performance RPC framework simplifies connected systems in cloud-native environments where low latency and strong typing remain essential. Developers can install updated packages through standard managers including pip install grpcio for Python and the updated Ruby gems.

This maintenance release demonstrates the project's focus on long-term stability rather than new features.

Use Cases
  • Professional microservices engineers connecting applications across multiple programming languages
  • Cloud infrastructure teams building high-throughput distributed computing systems
  • Backend developers implementing low-latency APIs between polyglot services
Similar Projects
  • Apache Thrift - offers comparable multi-language RPC with its own IDL
  • Cap'n Proto - provides faster serialization and RPC with less overhead
  • Connect - modern HTTP-first alternative compatible with gRPC protocols

Quick Hits

TDengine TDengine delivers high-performance time-series data storage and queries optimized for Industrial IoT, powering efficient sensor analytics at scale. 24.8k
openwrt OpenWrt supplies customizable Linux firmware for routers and embedded devices, letting builders create advanced networking solutions on any hardware. 26k
codex Codex runs a lightweight AI coding agent in your terminal, generating and debugging code instantly without leaving your workflow. 67k
ollama Ollama lets you run models like DeepSeek, Qwen, and Gemma locally in minutes, enabling fast private AI development on your machine. 165.9k
awesome-rust Awesome Rust curates battle-tested libraries, tools, and resources, helping builders quickly discover production-grade crates for any project. 56.3k
Pake 🤱🏻 Turn any webpage into a desktop app with one command. 46.9k

PiKVM V4 Platform Advances DIY IP-KVM for Remote Infrastructure 🔗

Updated hardware variants deliver industrial-grade reliability while preserving low-latency video capture and virtual media capabilities on Raspberry Pi

pikvm/pikvm · Unknown · 9.9k stars Est. 2019

PiKVM continues to solve one of infrastructure's most persistent problems: reliable remote access to servers when the operating system is unresponsive or absent. Six years after its initial release, the project has evolved with its V4 and V3 platforms, offering fully assembled, industrial-grade devices alongside the original DIY approach.

The system transforms Raspberry Pi boards into complete IP-KVM solutions.

PiKVM continues to solve one of infrastructure's most persistent problems: reliable remote access to servers when the operating system is unresponsive or absent. Six years after its initial release, the project has evolved with its V4 and V3 platforms, offering fully assembled, industrial-grade devices alongside the original DIY approach.

The system transforms Raspberry Pi boards into complete IP-KVM solutions. It captures FullHD video through an advanced HDMI-to-CSI bridge or USB dongle, delivering H.264 streams with 35-50ms latency. This performance enables responsive WebRTC, H.264-over-HTTP, and MJPEG video feeds through its web interface or VNC.

Supported platforms include Raspberry Pi 2, 3, 4, and Zero2W. The project explicitly avoids the Raspberry Pi 5, which lacks the necessary GPU video encoders to improve performance for KVM workloads. This deliberate hardware selection reflects the project's focus on practical engineering rather than chasing the latest silicon.

Key capabilities extend beyond basic video and input. The system provides bootable virtual CD/DVD and flash drive functionality, with image storage supported on NFS shares. USB keyboard and mouse emulation includes LED status, scroll wheel support, Bluetooth HID, mouse jiggling, and complete PS/2 compatibility. ATX functions allow direct power control of target systems.

For enterprise environments, PiKVM integrates with existing infrastructure through IPMI BMC, IPMI SoL, Redfish, and Wake-on-LAN. The included OS features a read-only filesystem for stability, extensible authorization, HTTPS by default, Pi health monitoring, GPIO control, and USB relay support.

The V4 and V3 platforms represent the project's shift toward accessibility. These plug-and-play variants maintain the same open-source software stack while providing robust, pre-assembled hardware suitable for production deployment. This addresses a common barrier for teams that value open source but require reliable, supported hardware.

For builders and administrators, PiKVM eliminates the need for expensive proprietary KVM switches. It delivers hardware-level access at a fraction of the cost, enabling everything from BIOS configuration to complete OS reinstallation without physical presence. The project's documentation and active community resources on Discord and its support forum further reduce the friction of implementation.

As remote and edge infrastructure grows more complex, PiKVM's combination of low-latency video, comprehensive input emulation, and virtual media capabilities becomes increasingly relevant. Its continued development demonstrates that open-source hardware projects can deliver enterprise-grade functionality while remaining accessible to individual builders.

(378 words)

Use Cases
  • Sysadmins troubleshooting servers without OS access
  • Engineers configuring BIOS on remote infrastructure
  • Technicians reinstalling OS via virtual media
Similar Projects
  • TinyPilot - Commercial Raspberry Pi KVM with simpler setup but higher cost and fewer hardware options
  • OpenBMC - Focuses on firmware-level server management rather than external HDMI capture and virtual drives
  • Looking Glass - Provides low-latency display sharing for VMs but lacks physical server power control and ATX features

More Stories

OpenIPC Expands Processor Support in Camera Firmware 🔗

Buildroot-based project now covers chips from 11 manufacturers with active updates

OpenIPC/firmware · C · 2k stars Est. 2021

OpenIPC continues to broaden hardware compatibility for IP cameras, moving well beyond its original HiSilicon focus. The firmware now supports SoCs from Ambarella, Anyka, Fullhan, Goke, GrainMedia, Ingenic, MStar, Novatek, SigmaStar and XiongMai, allowing repurposing of cameras from multiple vendors.

Built on Buildroot, the system generates complete Linux images with u-boot bootloader and C-language components optimized for embedded constraints.

OpenIPC continues to broaden hardware compatibility for IP cameras, moving well beyond its original HiSilicon focus. The firmware now supports SoCs from Ambarella, Anyka, Fullhan, Goke, GrainMedia, Ingenic, MStar, Novatek, SigmaStar and XiongMai, allowing repurposing of cameras from multiple vendors.

Built on Buildroot, the system generates complete Linux images with u-boot bootloader and C-language components optimized for embedded constraints. Users replace proprietary firmware to remove cloud dependencies, eliminate telemetry, and add custom features while maintaining local operation.

The project sustains development through a two-tier support model. Free assistance flows through community channels, while paid commercial subscriptions give businesses direct developer access, prioritized bug fixes, and faster feature implementation. This structure has kept the codebase active nearly five years after the initial release.

With supply-chain security under scrutiny, OpenIPC provides auditable code for organizations that need to verify and control the software running on surveillance hardware. Recent work includes improved FPV support for low-latency video applications alongside traditional security uses.

The community accepts patches, board additions, documentation fixes and donations via Open Collective to maintain long-term viability of the open framework.

(178 words)

Use Cases
  • Hobbyists flashing custom Linux on Xiongmai security cameras
  • Engineers integrating FPV cameras into open source drone builds
  • Companies deploying secure local surveillance without vendor lock-in
Similar Projects
  • OpenWrt - applies community-driven open firmware model to routers
  • Armbian - maintains Debian-based images for embedded ARM hardware
  • Buildroot - supplies the core build system used by OpenIPC

SmartSpin2k Adds BLE Log Streaming to Smart Trainers 🔗

Version 26.1.31 introduces runtime toggle for wireless diagnostics on DIY bike conversions

doudar/SmartSpin2k · C++ · 261 stars Est. 2020

SmartSpin2k has released version 26.1.31, adding BLE log streaming with a runtime toggle and read capability.

SmartSpin2k has released version 26.1.31, adding BLE log streaming with a runtime toggle and read capability. The update, delivered through pull request #714, allows users to access diagnostic data wirelessly without interrupting operation or requiring serial connections.

The project transforms ordinary spin bikes into smart trainers using an ESP32, stepper motor, and 3D-printed parts. Written in C++ with PlatformIO, it connects via Bluetooth Low Energy to Zwift, TrainerRoad and other platforms. The device automatically turns the resistance knob to match app commands while broadcasting power and cadence data.

ERG mode maintains target wattage by continuously adjusting resistance, enabling structured training without manual intervention. The hardware build uses a spin bike with a standard control knob, basic soldering, and printed components, typically completed in under an hour. Pre-assembled kits are available through SmartSpin2k.com.

The new logging feature can be enabled or disabled during runtime through the companion SS2kConfigApp on iOS and Android. A new Sole SB1200.AD_PRT part file was also added in this release. Five and a half years after its initial release, SmartSpin2k continues to refine the bridge between budget fitness equipment and modern training software.

Use Cases
  • Indoor cyclists automate resistance during Zwift structured workouts
  • DIY builders create ERG mode trainers with ESP32 hardware
  • Home fitness users connect spin bikes to TrainerRoad apps
Similar Projects
  • `esp32-zwift-bridge` - simpler BLE broadcast without ERG control
  • `open-trainer` - software-only virtual power solution
  • `ftms-arduino` - basic fitness machine protocol lacking stepper automation

Tinymovr 2.6.0 Fixes JSON Configuration Export 🔗

Update corrects missing parameters and requires simultaneous firmware and Studio upgrade

motionlayer/Tinymovr · C · 302 stars Est. 2020

Tinymovr 2.6.0 has shipped with targeted fixes to its configuration system.

Tinymovr 2.6.0 has shipped with targeted fixes to its configuration system. The release resolves a silent failure in JSON import and export that omitted six settable fields because they lacked meta: {export: True} flags in the YAML specification.

The missing parameters were controller.current.Iq_limit, controller.current.max_Ibus_regen, comms.can.heartbeat, sensors.user_frame.offset, sensors.user_frame.multiplier and watchdog.enabled. A new specification file, tinymovr_2_6_x.yaml, adds the flags, changing the protocol hash. Both firmware and Tinymovr Studio must therefore be updated together.

Earlier changes from the 2.5 series also appear in this release. Non-volatile memory configuration now survives DFU firmware updates, preventing the CAN node ID from resetting to 1. Homing offset calculations have been corrected for repeated runs, and homing planner settings are now persisted to NVM after save_config().

The project supplies open-source firmware for the PAC5527 MCU, a desktop client library, hardware design files, and documentation for a compact brushless motor controller. It implements field-oriented control with an integrated absolute encoder and CAN bus for precise PMSM motor applications.

Developers should consult SAFETY.md before modifying firmware and follow the Quick Start Guide for setup.

Use Cases
  • Robotics engineers tuning precise joint actuators via CAN
  • Automation teams integrating FOC drives in custom machinery
  • Researchers prototyping sensor-equipped brushless motor systems
Similar Projects
  • ODrive - higher-power FOC controller with different host protocol
  • VESC - sensorless-focused ESC popular in personal transport
  • SimpleFOC - Arduino-centric library lacking integrated hardware

Quick Hits

Useful-Youtube-Channels Curated list of top YouTube channels delivering practical electronics and mechanical engineering knowledge for builders. 349
iSMC Go CLI tool that decodes Apple SMC data to display real-time temperatures, fans, battery, voltage and power metrics. 175
librealsense C++ SDK for Intel RealSense depth cameras that delivers precise 3D scanning and spatial awareness capabilities. 8.6k
AxxSolder STM32-based open-source soldering station for JBC cartridges that runs on 9-24V DC or USB-PD power. 1k
TuyaOpen Open-source AI+IoT framework for Tuya chips and ESP32 that speeds up smart hardware and agent integration. 1.5k

Magpie 0.12.1 Refines Window Upscaling Stability and Cursor Control 🔗

Latest release adds automatic cursor hiding and resolves layering, cropping, and termination issues for Windows developers and gamers

Blinue/Magpie · HLSL · 13.5k stars Est. 2021 · Latest: v0.12.1

Magpie has long served as a general-purpose window upscaler for Windows 10 and 11, capturing application output and applying high-quality scaling algorithms in real time. Version 0.12.

Magpie has long served as a general-purpose window upscaler for Windows 10 and 11, capturing application output and applying high-quality scaling algorithms in real time. Version 0.12.1 focuses on reliability and usability, addressing pain points that have frustrated users working with legacy software and modern high-resolution displays.

The update introduces support for automatically hiding the cursor when idle, with a customizable delay. This removes visual clutter during fullscreen scaling sessions, particularly useful for games and emulators. The auto-scaling mechanism has also been improved so that pop-up dialogs no longer interrupt the scaling process, ensuring uninterrupted operation.

Several compatibility problems were fixed. Scaling no longer stops unexpectedly in edge cases. The source window can no longer appear above the scaled output; the scaled window is now permanently topmost, and the previous “Keep scaled window on top” option has been removed. Title bar cropping failures on certain applications have been corrected, and a bug involving monochrome cursors that could freeze the scaled frame has been resolved. A toolbar screenshot menu ID conflict was also eliminated.

At its core, Magpie uses HLSL shaders running on DirectX feature level 11 to perform the upscaling pass. It ships with multiple built-in effects, including Anime4K for line art preservation, AMD FSR for temporal upscaling, and various CRT shaders for retro aesthetics. The tool supports both windowed and fullscreen scaling, works across multiple monitors, and presents a WinUI interface built with cppwinrt and XAML Islands that follows Fluent Design with light and dark themes.

For builders, Magpie solves a persistent Windows problem: older applications and games often render poorly when DPI scaling is forced by the OS. The recommended workaround remains setting “High DPI scaling override” to “Application” in compatibility properties. With this release, the tool becomes more dependable for automated workflows, remote desktop sessions, and testing scenarios that require consistent high-resolution output.

The project’s continued maintenance after five years demonstrates the lasting need for system-level scaling solutions that operate independently of individual applications. Developers working on games, emulators, or cross-resolution UI testing now have fewer interruptions and more predictable behavior when deploying Magpie in their pipelines.

System requirements remain Windows 10 v1903 or Windows 11 with DirectX 11 support. The GPLv3 license and active contributor base continue to make the project accessible for those who need to inspect or extend the shader pipeline.

Use Cases
  • Gamers upscaling older titles lacking native high DPI support
  • Developers testing UI rendering across multiple display resolutions
  • Retro enthusiasts applying CRT and Anime4K shaders to emulators
Similar Projects
  • Lossless Scaling - commercial Steam tool with different upscaling algorithms but no open shader access
  • ReShade - focuses on per-application shader injection rather than system-wide window capture
  • Windows Magnifier - provides basic enlargement without advanced HLSL algorithms or multi-monitor scaling

More Stories

Bevy 0.18.1 Refines Rust Game Engine Stability 🔗

Latest patch delivers targeted fixes to data-driven ECS architecture

bevyengine/bevy · Rust · 45.2k stars Est. 2020

Bevy has released version 0.18.1, the latest incremental update to its Rust game engine.

Bevy has released version 0.18.1, the latest incremental update to its Rust game engine. The patch addresses issues carried over from 0.18.0, with the full set of changes visible in the GitHub diff between the two versions.

The engine continues to centre on an Entity Component System that organises game logic as data, enabling parallel execution where possible. This design supports the project's goals of speed and modularity: developers import only the crates they need and replace components that do not suit their requirements.

Version 0.18.1 maintains the Minimum Supported Rust Version close to the latest stable compiler release, ensuring immediate access to language improvements in memory safety and performance. The release cycle remains disciplined, with breaking changes arriving roughly every three months accompanied by migration guides.

Documentation for the update includes revised API references and official examples that demonstrate rendering, input handling and physics integration. Community resources on Discord and Reddit continue to serve both newcomers and experienced contributors working on 2D and 3D titles.

The 0.18.1 release underscores Bevy's long-term focus on balancing rapid evolution with practical usability for Rust-based game development.

Use Cases
  • Indie developers building 2D platformers with ECS architecture
  • Teams prototyping 3D simulations using modular Rust components
  • Contributors extending engine features through plugin development
Similar Projects
  • Godot - offers visual scripting versus Bevy's code-first ECS
  • Unity - uses C# components with larger commercial asset ecosystem
  • Fyrox - provides Rust engine with built-in editor contrasting Bevy's minimalism

raylib 5.5 Adds One-Click Web Builds for C Games 🔗

Latest version simplifies WebAssembly targeting in new Windows package for novice developers

raysan5/raylib · C · 31.7k stars Est. 2013

raylib 5.5 arrived this week, one year after version 5.0 and eleven years after the project's first release.

raylib 5.5 arrived this week, one year after version 5.0 and eleven years after the project's first release. The update brings 800 commits, 270 closed issues, 30 new API functions for a total of 580, and improvements to 110 existing ones.

The most practical change is the revised pre-configured Windows package. New users can now edit C files in Notepad++ and compile the same codebase to WebAssembly with a single mouse click. The addition removes previous barriers for developers moving from desktop executables to browser targets.

The library remains true to its original design: written in plain C99 with no external dependencies. It ships with an OpenGL abstraction layer (rlgl) that supports versions 1.1 through 4.3 plus ES 2.0 and 3.0. Full 3D capabilities include skeletal animation for IQM, M3D and glTF models, PBR materials, postprocessing shaders, and a complete raymath module for vector, matrix and quaternion operations.

Audio streaming handles WAV, OGG, MP3, FLAC and tracker formats. The project continues to target prototyping, tooling, embedded systems and education rather than providing visual editors or high-level frameworks.

**

Use Cases
  • Beginner programmers creating first videogames in pure C
  • Educators teaching graphics programming through concise code examples
  • Embedded engineers adding visuals to Raspberry Pi and IoT projects
Similar Projects
  • SDL - lower-level library requiring more boilerplate for graphics
  • SFML - C++ alternative with object-oriented design approach
  • GLFW - window and input management without built-in drawing tools

GDevelop Fixes Object Variable Errors in Editor 🔗

Version 5.6.261 corrects broken references after using object editor dialog

4ian/GDevelop · JavaScript · 21.5k stars Est. 2014

GDevelop version 5.6.261 addresses a bug that corrupted object variables in the Events Sheet.

GDevelop version 5.6.261 addresses a bug that corrupted object variables in the Events Sheet. The issue surfaced specifically when developers edited objects through the dedicated object editor dialog rather than the properties panel, causing incorrect variable replacement and breaking event logic. The maintainers describe the update as highly recommended for all users.

The open-source engine enables creation of 2D, 3D and multiplayer games without traditional coding. Its event-based system combined with modular behaviors lets users define game mechanics visually while targeting iOS, Android, desktop and web platforms. The JavaScript-powered runtime ensures consistent performance across exports.

GDevelop consists of a visual editor, dedicated game engine, extension ecosystem and supporting online services. An asset store accepts both free and commercial templates, asset packs and extensions. Many titles built with the tool have shipped to Steam, itch.io, mobile app stores and web portals including CrazyGames and Poki.

The project maintains an active roadmap with good first issues for new contributors. Recent additions include AI assistance that can generate or refine events alongside the user. This maintenance release reinforces stability for a maturing tool that continues supporting both hobbyists and commercial developers.

(178 words)

Use Cases
  • Independent developers creating cross-platform 2D games without coding
  • Educators teaching game design through visual event systems
  • Teams building HTML5 multiplayer titles for web browsers
Similar Projects
  • Godot - open-source engine offering both visual scripting and GDScript
  • GameMaker - commercial tool with similar drag-and-drop event system
  • Unity - full-featured engine requiring C# for complex 3D projects

Quick Hits

flame Build cross-platform 2D games in Dart with Flame, a Flutter-based engine delivering fast rendering, physics, and animation tools. 10.5k
Slug Implement the Slug Algorithm in HLSL shaders with this reference code for advanced real-time vector graphics rendering. 1k
Pixelorama Design sprites, tiles, and animations with Pixelorama, a versatile open-source pixel art multitool for desktop and web. 9.3k
GodSVG Create structured SVG graphics with GodSVG, a precision vector editor available on all major desktop platforms and web. 2.4k
godot_dialogue_manager Build branching conversations in Godot with this nonlinear dialogue system for powerful story and NPC interactions. 3.4k