ChatGH

lyogavin / airllm

AirLLM enables running 70B+ LLMs on a single 4GB GPU via layer-wise inference—loading one layer at a time without quantization or pruning. Supports Llama, QWen, Mistral, ChatGLM, Baichuan and more, with optional 4/8-bit compression for 3x speedup, KV cache, and macOS/Apple Silicon support.

zhaoxuya520 / reverse-skill

A cybersecurity skill router pack that gives AI coding agents (Claude Code, Cursor, etc.) the ability to handle reverse engineering, penetration testing, and security research tasks. It auto-routes 38+ scenarios — from APK/ELF analysis to CTF capture — boots the right toolchain on demand, and evolves through a built-in field journal.

firecrawl / pdf-inspector

A fast Rust library for PDF classification and text extraction. Intelligently detects scanned vs. text-based PDFs in ~10-50ms and extracts clean Markdown with tables, headings, and reading-order awareness — enabling smart OCR routing without external services or ML models.

esengine / DeepSeek-Reasonix

Reasonix is a DeepSeek-native AI coding agent built for the terminal — a single static Go binary with config-driven providers, MCP-compatible plugins, and cache-aware context maintenance tuned for low token costs across long sessions. Use it from CLI, desktop, or VS Code.

TencentCloud / TencentDB-Agent-Memory

A fully local long-term memory system for AI agents with a 4-tier progressive pipeline (L0→L3). Combines symbolic short-term memory via Mermaid graphs to slash token usage by up to 61%, and layered long-term memory to distill conversations into structured personas and scenes — zero external API dependencies, white-box debuggable, plug-and-play for OpenClaw and Hermes.

microsoft / AI-For-Beginners

Microsoft's 12-week, 24-lesson AI beginner curriculum covering neural networks, deep learning, computer vision, NLP, and more, with hands-on Jupyter notebooks for both PyTorch and TensorFlow, plus quizzes, labs, and AI ethics—available in 50+ languages.

microsoft / generative-ai-for-beginners

A comprehensive 21-lesson course by Microsoft Cloud Advocates teaching you to build Generative AI applications from scratch. Covers LLMs, prompt engineering, text/image/chat/search apps, RAG with vector databases, AI agents, fine-tuning, and open-source models — with hands-on Python & TypeScript code samples for Azure OpenAI, Microsoft Foundry Models, and local deployment.

donnemartin / system-design-primer

A comprehensive, open-source guide to mastering large-scale system design. Covers scalability, CAP theorem, databases, caching, load balancing, and more, with real-world architectures, interview questions with solutions, and Anki flashcards for effective spaced-repetition learning.

antirez / ds4

DwarfStar is a compact, high-performance inference engine purpose-built for DeepSeek V4 Flash and PRO, with GLM 5.2 support. It runs on Metal, CUDA, and ROCm backends and is optimized for consumer hardware such as MacBooks, DGX Spark, and Strix Halo. Features include SSD streaming for out-of-RAM models, multi-machine pipeline parallelism, tensor parallelism over Thunderbolt, a native coding agent, and an OpenAI/Anthropic-compatible server. Created by Redis author antirez.