Github Top Repositories

🔍 Deep-diving into unslothai/unsloth — fresh off the trending list. 🔗 https://github.com/unslothai/unsloth 📝 Local UI to run and train LLMs and diffusion models, including Qwen3.8, Kimi K3, MiniMax-H3, Gemma 4, DeepSeek-V4, FLUX and more. ────────────────────────────── Unsloth is a cutting-edge desktop app that allows users to run and train AI models locally, with support for various types of models. It features a user-friendly interface, native support for CPU, NVIDIA, AMD, and Intel hardware, and multi-GPU setups. Users can download the app for Windows, macOS, or Linux, and start using it right away. The app offers a range of features, including model training, fine-tuning, and deployment, as well as private and unlimited web search, deep research, and RAG. It also supports image and video diffusion, audio models, and reinforcement learning. Unsloth is designed for developers, researchers, and AI enthusiasts who want to work with AI models locally, without relying on cloud services. The app is free to use and offers a range of community resources, including documentation, tutorials, and forums. To get started with Unsloth, users can simply download the app and follow the installation instructions. The app also offers a web-based interface and a command-line interface for more advanced users. One of the key benefits of Unsloth is its ability to run AI models locally, which provides a high level of security and control for users. The app also offers remote access capabilities, allowing users to access their models from anywhere. Overall, Unsloth is a powerful and flexible tool for working with AI models locally, and it offers a range of features and benefits that make it an attractive choice for developers, researchers, and AI enthusiasts. Takeaway: Unsloth is the ultimate tool for running and training AI models locally, offering a unique combination of power, flexibility, and security. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🔥 altic-dev/FluidVoice is trending — and it deserves your attention. 🔗 https://github.com/altic-dev/FluidVoice 📝 Fastest and only macOS Dictation app with on-device STT and custom trained AI enhancement model. A local Wispr Flow alternative. ⭐ helps a ton :) Windows & iOS waitlist open. Linux soon. ────────────────────────────── FluidVoice is an open-source, on-device AI-enhanced voice-to-text dictation app for macOS. It features Fluid Intelligence, a local AI model for smart formatting and context-aware capitalization, and supports various speech models like Nemotron, Parakeet, and Whisper. To get started, brew install --cask fluidvoice or download the latest release. Grant microphone and accessibility permissions, set your hotkey, and go through onboarding to choose your voice model. Key Features: - On-device AI enhancement with Fluid Intelligence - Multiple speech models for different languages and latency needs - Command Mode for voice control and Write Mode for dictation in any text field - Live preview with notch support and adaptive theming Technical Highlights: - Built with Swift and Swift Package Manager - Supports Apple Silicon and Intel Macs (via Whisper models) - Requires macOS 15.0 or later and ~1 GB disk space for a voice model Audience: - Individuals looking for a free, open-source dictation app with on-device AI enhancement - Developers interested in contributing to the project or building from source Get started with FluidVoice today and experience fast, accurate, and private dictation on your Mac - your voice, your words, your way. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🔥 cactus-compute/needle is trending — and it deserves your attention. 🔗 https://github.com/cactus-compute/needle 📝 14MB foundation model for tiny devices; phones, wearables, smart home, and robots. ────────────────────────────── The Needle repository on GitHub is home to an open 45M-parameter model designed for tool calling, device use, and structured extraction. This model is unique in that it's a single 14MB binary that requires approximately 28MB of RAM to run a full session. Key features of the Needle model include its self-contained nature, with weights baked into the engine, a simple contract for tool calls, confidence-gated responses, and bounded memory usage. To use the Needle model, you can install it via pip with pip install cactus-needle and then describe your tools to interact with the model. The model supports tool retrieval, where it can select the most relevant tools based on the input query, and confidence-gated responses, where the model returns a confidence score with each response. Technical highlights of the Needle model include its use of a Simple Attention Network architecture, which is a dense small-model recipe that incorporates a Hadamard MLP, GQA attention, and engram key-value memory. The model is also compressed using CQ2-bit with Cactus Quants, which reduces its size while maintaining its performance. The Needle model is suitable for a wide range of audiences, including developers, researchers, and anyone interested in natural language processing and tool calling. With its unique combination of features and technical advancements, the Needle model is a valuable resource for anyone looking to explore the possibilities of AI-powered tool calling and extraction. In summary, the Needle model is a powerful and efficient tool for natural language processing and tool calling, with a unique combination of features and technical advancements that make it an exciting and valuable resource - and with Needle, you can have the power of a large language model in the palm of your hand. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🔥 anthropics/skills is trending — and it deserves your attention. 🔗 https://github.com/anthropics/skills 📝 Public repository for Agent Skills ────────────────────────────── The anthropics/skills GitHub repository is a treasure trove of skills for Claude, a platform that leverages these skills to improve performance on specialized tasks. These skills are essentially folders containing instructions, scripts, and resources that Claude loads dynamically. The repository includes a wide range of skills, from creative applications like art and music to technical tasks such as testing web apps and enterprise workflows. To get started, you can browse through the various skills in the repository, which are all self-contained in their own folders with a SKILL.md file. Many of these skills are open source, while others, like the document creation and editing skills, are source-available. You can use these skills as inspiration for your own or to understand different patterns and approaches. The repository also includes a template-skill that you can use as a starting point to create your own custom skills. Creating a basic skill is straightforward, requiring only a folder with a SKILL.md file containing YAML frontmatter and instructions. The skills can be used in various ways, including through Claude Code, Claude.ai, and the Claude API. Whether you're a developer looking to teach Claude how to use specific software or an enterprise seeking to improve workflows, the anthropics/skills repository has something to offer. One-liner takeaway: With the anthropics/skills repository, you can unlock Claude's full potential and teach it new tricks to streamline your workflow and boost productivity! ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🔍 Deep-diving into semantica-agi/semantica — fresh off the trending list. 🔗 https://github.com/semantica-agi/semantica 📝 Graph-Native Infrastructure for Context and Accountable AI Systems ────────────────────────────── Semantica is an open-source, graph-native infrastructure for building context and accountable AI systems. It's designed to ingest enterprise data, extract relevant information, and construct a knowledge graph with full decision provenance. Key features include graph analytics, deterministic reasoning, ontology management, and end-to-end traceability. Technical highlights of Semantica include polyglot graph storage, RDF and LPG support, and compliance with W3C standards. The platform is self-hostable, auditable, and governed, with zero vendor lock-in. Audience for Semantica includes AI/ML platform teams, data platform teams, compliance and risk teams, and regulated enterprises. It's ideal for high-stakes, regulated domains where explainable and trustworthy AI decision-making is crucial. To get started with Semantica, simply run pip install semantica and explore the ContextGraph API. With Semantica, you can build a transparent and accountable AI system that provides real, structured answers to "why did this happen?" In short, Semantica is the missing infrastructure layer that brings trust and transparency to your AI decisions. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

💡 cathrynlavery/diagram-design just hit the trending charts — here's why it matters. 🔗 https://github.com/cathrynlavery/diagram-design 📝 29 editorial diagram types for Claude Code. Self-contained HTML + SVG. No shadows, no Mermaid-slop. ────────────────────────────── The diagram-design GitHub repository provides a Claude Code skill for creating high-quality, editorial diagrams that match your brand. With 27 visual types and three static variants, you can easily generate diagrams such as architecture sketches, flowcharts, and pyramids. The skill uses a shared-memory hub and semantic patterns to describe behavior separately from layout, allowing for flexible and accessible diagrams. To get started, you can install the skill using /plugin marketplace add cathrynlavery/diagram-design for Claude Code or codex plugin marketplace add cathrynlavery/diagram-design for Codex. The skill will then pull your brand's colors and typography from your website and apply them to the diagrams. The repository also includes an online gallery where you can browse the available diagrams. Overall, the diagram-design skill is a powerful tool for creating high-quality, accessible diagrams that match your brand's style. Take your diagrams to the next level with this skill - no more generic rounded boxes! ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🚀 21-Day CCNA & CCNP Sprint – Aug 17 to Sep 6 🤝 Peer Group – Share Insights | Exchange Knowledge | Support Each Other No more studying alone. Join a community of CCNA/CCNP candidates, learn together, and win prizes. How it works: ① DM admin: "I'M IN + cert name" ② Join the group ③ Check in 18/21 days → win 🎁 Prizes (first come, first served): $50 Amazon card ×1 | SD-Access Training ×1 | SD-WAN Training ×1 | CCNA Pro Package ×10 | Free Learning Pack (all finishers) Daily check-in: 1️⃣ What you learned 2️⃣ Explain it in your own words 3️⃣ (Optional) Ask the group Join now: https://chat.whatsapp.com/KZrAj2HZ3Y5K9UhhNhrApf DM to register: https://wa.me/8619559123054

⚡ cactus-compute/needle is making waves. Here's the full picture. 🔗 https://github.com/cactus-compute/needle 📝 14MB foundation model for tiny devices; phones, wearables, smart home, and robots. ────────────────────────────── Needle 2 is a compact, open-source model for tool calling, device use, and structured extraction. It's a single 14MB binary that runs in approximately 28MB of RAM. Key features include a self-contained engine, simple contract for tool calls, and confidence-gated responses. Needle 2 is built on the Simple Attention Network architecture, which includes a Hadamard MLP, GQA attention, and engram key-value memory. The model is compressed to CQ2-bit with Cactus Quants and can be used for a variety of tasks, including tool calling, data extraction, and more. To use Needle 2, you can install the cactus-needle Python package and describe your tools using a simple contract. The model can then be used to call tools, extract structured data, and more. Audience: This model is suitable for developers and researchers looking for a compact, efficient, and easy-to-use model for tool calling and data extraction tasks. One-liner takeaway: Needle 2 is a powerful, compact model that makes it easy to build tool calling and data extraction applications with confidence-gated responses. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🔥 embabel/embabel-agent is trending — and it deserves your attention. 🔗 https://github.com/embabel/embabel-agent 📝 Agent framework for the JVM. Pronounced Em-BAY-bel /ɛmˈbeɪbəl/ ────────────────────────────── Embabel Agent Framework is a Java-based framework for building intelligent agents that can seamlessly mix large language model (LLM) interactions with code and domain models. It's designed to support sophisticated planning, extensibility, and reuse, making it an ideal choice for developers who want to create complex, dynamic systems. The framework is built on top of the JVM, leveraging the strengths of Spring and Kotlin, and provides a natural usage model for Java developers. It's highly extensible, allowing developers to add new domain objects, actions, goals, and conditions without modifying existing code. Key features include: - Actions: Steps an agent takes - Goals: What an agent is trying to achieve - Conditions: Conditions to assess before executing an action or determining that a goal has been achieved - Domain model: Objects underpinning the flow and informing actions, goals, and conditions Technical highlights include: - Sophisticated planning using Goal Oriented Action Planning (GOAP) or Utility AI - Strong typing and object-oriented benefits - Platform abstraction for clean separation between programming model and platform internals - Designed for LLM mixing and cost-effective solution Audience: Embabel Agent Framework is ideal for developers who want to create complex, dynamic systems that leverage the power of LLMs and agentic programming. Get started with Embabel in under 5 minutes using the Java or Kotlin template, and explore the Embabel Agent Examples Repository for tutorials and examples. Embabel Agent Framework: where intelligence meets adaptability. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🔍 Deep-diving into Lightricks/LTX-2 — fresh off the trending list. 🔗 https://github.com/Lightricks/LTX-2 📝 Official Python inference and LoRA trainer package for the LTX-2 audio–video generative model. ────────────────────────────── The Lightricks/LTX-2 GitHub repository presents a groundbreaking DiT-based audio-video foundation model, LTX-2, which integrates all core capabilities of modern video generation into one model. This includes synchronized audio and video, high fidelity, multiple performance modes, production-ready outputs, API access, and open access. To get started, users can clone the repository, install dependencies, and download the required models using the Hugging Face CLI. The repository provides a Quick Start guide, which demonstrates how to generate video using the distilled model and pipeline. LTX-2 features various models, including the transformer, text encoder, video VAE, audio VAE, and spatial upscaler. Each model has different versions, allowing users to choose the best fit for their specific needs. The repository also offers multiple pipelines, such as DistilledPipeline, DFRPipeline, and TI2VidTwoStagesPipeline, each with its unique features and use cases. LTX-2 is suitable for a wide range of users, from researchers and developers to content creators and artists. In a nutshell, LTX-2 revolutionizes video generation, and its capabilities are a game-changer: unleash your creativity with LTX-2, where AI meets art. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

⚡ localsend/localsend is making waves. Here's the full picture. 🔗 https://github.com/localsend/localsend 📝 An open-source cross-platform alternative to AirDrop ────────────────────────────── LocalSend is a free, open-source app that lets you securely share files and messages with nearby devices over your local network without needing an internet connection. The app uses a REST API and HTTPS encryption for secure communication. Key features include cross-platform compatibility, secure communication protocol, and no requirement for an internet connection or third-party servers. To use LocalSend, simply download and install the app on your device, and follow the setup instructions to configure your firewall and router settings. From a technical perspective, LocalSend uses a secure communication protocol that generates a TLS/SSL certificate on the fly on each device, ensuring maximum security. The app is built using Flutter and Rust, and the code is available on GitHub for contributors to review and modify. The app is suitable for anyone who wants to securely share files and messages with nearby devices, including individuals, businesses, and organizations. In short, LocalSend is a fast, reliable, and secure solution for local file and message sharing - share files and messages with ease, without the need for internet. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🌟 smicallef/spiderfoot caught my eye on GitHub Trending today. 🔗 https://github.com/smicallef/spiderfoot 📝 SpiderFoot automates OSINT for threat intelligence and mapping your attack surface. ────────────────────────────── SpiderFoot is an open-source intelligence (OSINT) automation tool that integrates with various data sources and utilizes multiple methods for data analysis. It offers a web-based interface and command-line functionality, making it easy to navigate and use. With over 200 modules, SpiderFoot can perform tasks such as host enumeration, email address extraction, and threat intelligence queries. The tool is written in Python 3 and is MIT-licensed. Key Features: - Web-based UI or CLI - Over 200 modules - Python 3.7+ support - YAML-configurable correlation engine - CSV/JSON/GEXF export - API key export/import Technical Highlights: - SQLite back-end for custom querying - TOR integration for dark web searching - Dockerfile for Docker-based deployments Audience: - Security professionals - Researchers - Penetration testers Takeaway: With its extensive module library and customizable correlation engine, SpiderFoot is the ultimate OSINT tool for anyone looking to streamline their intelligence gathering process. ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

🔍 Deep-diving into ZuodaoTech/everyone-can-use-english — fresh off the trending list. 🔗 https://github.com/ZuodaoTech/everyone-can-use-english 📝 人人都能用英语 ────────────────────────────── Everyone Can Use English is an innovative project that leverages AI to help users learn and practice English. The project includes a web version, browser extension, and upcoming desktop application. The web version, available at https://enjoy.bot, offers features such as video and ebook learning, flashcards, and courses. The browser extension supports YouTube and Netflix, allowing users to practice their English skills while watching their favorite shows. Technical highlights of the project include automated testing and deployment workflows, as well as a latest version badge that displays the current app version. The project also provides a range of learning resources, including a book with chapters on speaking, pronunciation, reading, and grammar. The project is suitable for anyone looking to improve their English skills, from beginners to advanced learners. With its comprehensive resources and interactive tools, Everyone Can Use English is an excellent choice for those who want to take their English skills to the next level. Learn English with AI - it's a game-changer! ────────────────────────────── 🧠 Channel: https://t.me/GithubRe

💡 NVIDIA-NeMo/Switchyard just hit the trending charts — here's why it matters. 🔗 https://github.com/NVIDIA-NeMo/Switchyard 📝 No description. ────────────────────────────── The NVIDIA-NeMo/Switchyard GitHub repository offers a fascinating Rust-based proxy and library designed to streamline Large Language Model (LLM) traffic. Its primary purpose is to route requests across different providers, translate between OpenAI and Anthropic APIs, record operational metrics, and provide typed, composable routing algorithms. The key features of Switchyard include protocol translation, multi-backend routing, and operational metrics. It supports various routing strategies, such as LLM classifier, stage router, escalation router, and random routing. To use Switchyard, you can choose from three paths: the launcher path for launching coding agents like Claude Code or Codex, the server path for running Switchyard as a standalone proxy, or the library path for embedding routing algorithms in your own Rust application. From a technical standpoint, Switchyard is built using Rust and utilizes a modular architecture, making it easy to extend and customize. It also provides a range of documentation, including getting started guides, core concepts, and API references. The target audience for Switchyard includes developers and researchers working with LLMs, particularly those interested in routing and translation between different APIs. In summary, Switchyard is a powerful tool for managing LLM traffic, and its flexibility, customizability, and ease of use make it an attractive solution for those looking to streamline their LLM workflows: Switchyard helps you route LLM traffic with ease, so you can focus on building innovative AI applications! ────────────────────────────── 🧠 Channel: https://t.me/GithubRe