#python

Mini-SGLang is a compact, easy-to-read inference framework (~5,000 Python lines) that runs and serves large language models with high speed using optimizations like radix cache, chunked prefill, overlap scheduling, tensor parallelism, and FlashAttention/FlashInfer kernels. It’s CUDA-dependent, quick to install from source, and can launch an OpenAI-compatible API or interactive shell for single- or multi‑GPU serving, letting you test or deploy models (e.g., Qwen, Llama) with low latency and scalable throughput. Benefit: you get a transparent, modifiable engine to deploy fast, efficient LLM inference for development, benchmarking, or production use.

https://github.com/sgl-project/mini-sglang

#typescript #documentation_generator #nuxt #nuxt_theme

Docus is a CLI tool that quickly scaffolds a complete, modern documentation site using Markdown and Vue (Nuxt 4), with responsive design, dark mode, i18n, full-text search, enhanced Markdown components, TypeScript support, and built-in AI/LLM integration via llms.txt and a native MCP server for editor/IDE tools like Cursor and VS Code, letting you start a docs site with npx create-docus and npm run dev so it runs locally instantly. Benefit: you get a production-ready, customizable docs site fast—saving setup time and giving built-in search, localization, performance, and AI tooling to improve authoring and user experience.

https://github.com/nuxt-content/docus

#go #gemma3 #go #gpt_oss #granite4 #llama #llama3 #llm #on_device_ai #phi3 #qwen3 #qwen3vl #sdk #stable_diffusion #vlm

NexaSDK runs AI models locally on CPUs, GPUs, and NPUs with a single command, supports GGUF/MLX/.nexa formats, and offers NPU-first Android and macOS support for fast, multimodal (text, image, audio) inference, plus an OpenAI‑compatible API for easy integration. This gives you low-latency, private on-device AI across laptops, phones, and embedded systems, reduces cloud costs and data exposure, and lets you deploy and test new models immediately on target hardware for faster development and better user experience.

https://github.com/NexaAI/nexa-sdk

#python #ai #bug_detection #code_audit #code_quality #code_review #developer_tools #devsecops #google_gemini #llm #react #sast #security_scanner #supabase #typescript #vite #vulnerability_scanner #xai

**DeepAudit** is an AI-powered code audit tool using multi-agent collaboration to deeply scan projects for vulnerabilities like SQL injection, XSS, and path traversal. Import code from GitHub/GitLab or paste snippets; agents plan, analyze with RAG knowledge, and verify issues via secure Docker sandbox PoCs, generating PDF reports with fix suggestions. Deploy easily with one Docker command, supports local Ollama models for privacy, and cuts traditional tools' high false positives. **You benefit** by automating secure audits like a pro hacker—saving time, reducing errors, ensuring real exploits are caught, and speeding safe releases without manual hassle.

https://github.com/lintsinghua/DeepAudit

#rust #ai #change_data_capture #context_engineering #data #data_engineering #data_indexing #data_infrastructure #data_processing #etl #hacktoberfest #help_wanted #indexing #knowledge_graph #llm #pipeline #python #rag #real_time #rust #semantic_search

**CocoIndex** is a fast, open-source Python tool (Rust core) for transforming data into AI formats like vector indexes or knowledge graphs. Define simple data flows in ~100 lines of code using plug-and-play blocks for sources, embeddings, and targets—install via `pip install cocoindex`, add Postgres, and run. It auto-syncs fresh data with minimal recompute on changes, tracking lineage. **You save time building scalable RAG/semantic search pipelines effortlessly, avoiding complex ETL and stale data issues for production-ready AI apps.**

https://github.com/cocoindex-io/cocoindex

#python

**Reachy Mini** is an open-source desktop robot, 11 inches tall and 3.3 lbs, with a 6-DoF expressive head, 360° body rotation, animated antennas, wide-angle camera, microphones, speaker, and Hugging Face AI integration for 1.7M+ models. Assemble in 2-3 hours as a kit; choose Lite (USB-powered) or Wireless (Raspberry Pi, battery). Use simple Python SDK for quick control, apps like conversation or hand-tracking, and simulation. **You benefit** by easily building, testing, and sharing AI robots at home or work, speeding up embodied AI experiments affordably.

https://github.com/pollen-robotics/reachy_mini

#python #cv #cv_builder #cv_generator #cv_template #python #resume #resume_builder #resume_generator #resume_template #typst

RenderCV lets you write your CV as simple YAML text, then run one command like rendercv render yourfile.yaml to get a perfect PDF with pro typography—no layout fights or formatting hassle. Version-control it easily, focus purely on content, customize themes/colors/fonts, validate strictly, and use any language. You save time, get pixel-perfect resumes every time, and maintain versions without chaos.

https://github.com/rendercv/rendercv

#python #gemini #gemini_ai #gemini_api #gemini_flash #gemini_pro #information_extration #large_language_models #llm #nlp #python #structured_data

**LangExtract** is a free Python library that uses AI models like Gemini to pull structured data—like names, emotions, or meds—from messy text such as reports or books. It links every fact to its exact spot in the original, creates interactive visuals for easy checks, handles huge files fast with chunking and parallel runs, and works with cloud or local models without fine-tuning. You benefit by quickly turning unstructured docs into reliable, organized data for analysis, saving time and boosting accuracy in fields like healthcare or research.

https://github.com/google/langextract

#powershell

EntraGoat creates a safe, vulnerable Microsoft Entra ID setup in your test tenant using PowerShell and a web interface for easy deployment of scenarios like privilege escalation attacks. Clone the GitHub repo, install tools, run scripts for challenges with flags, solutions, and cleanups—no extra costs. You benefit by safely practicing real-world identity hacks, spotting misconfigurations, and boosting your skills to secure production systems without risks.

https://github.com/Semperis/EntraGoat

#python

This repository offers Anthropic's Claude Skills—folders with instructions, scripts, and resources that dynamically teach Claude specialized tasks like branded documents, data analysis, or workflows. Examples cover creative, technical, and enterprise uses; install via Claude Code, .ai, or API, or create your own with a simple SKILL.md template. You benefit by automating repetitive work, boosting productivity, ensuring consistent results, and capturing your team's knowledge for reliable, scalable AI performance.

https://github.com/anthropics/skills

#javascript #aicoding #free

AIClient-2-API is a free Node.js proxy that turns client-only AI models like Gemini 3 Pro, Claude 4.5 Opus, Qwen3 Coder Plus, and Kiro into one easy OpenAI-compatible API you run locally with Docker or a script. Access a web console at localhost:3000 to add keys, switch models, monitor health, and log chats—no code changes needed for tools like Cherry-Studio. You benefit by using top free/cheap advanced AIs seamlessly, bypassing limits with smart account pooling for 99.9% uptime, saving costs, and building private datasets from logs.

https://github.com/justlovemaki/AIClient-2-API

#rust

Miri is a tool that detects bugs in unsafe Rust code by finding undefined behavior—situations where your program violates safety rules and can behave unpredictably. When you write unsafe code, you bypass Rust's normal safety checks, so you must manually ensure your code follows strict requirements like proper memory alignment, no data races, and correct pointer usage. Miri catches violations of these requirements by running your code in a special interpreter that monitors every operation. It detects problems like out-of-bounds memory access, use-after-free errors, uninitialized data, and misaligned pointers. You can easily use Miri by installing it with Rust's nightly toolchain and running `cargo miri test` on your project. The benefit is that Miri finds subtle bugs that would otherwise cause crashes or security vulnerabilities in production, making it an essential tool for anyone writing unsafe Rust code.

https://github.com/rust-lang/miri

#cmake

CMake-based MinGW-w64 Cross Toolchain

https://github.com/shinchiro/mpv-winbuild-cmake

#go #git #go #golang #hacktoberfest #hooks #lefthook #manager

Lefthook is a fast Git hooks manager built in Go for Node.js, Ruby, Python, and other projects. Install it easily via Go, NPM, gem, or pipx, then configure hooks in a simple lefthook.yml file and run `lefthook install`. It runs commands in parallel, filters files with globs/regex, supports scripts, tags, Docker, and local overrides for speed and control. This saves you time on commits/pushes by automating linting and checks quickly without dependencies, keeping code clean effortlessly.

https://github.com/evilmartians/lefthook

#python #docker #fastapi #kbqa #kgqa #llms #neo4j #rag #vue

Yuxi-Know (语析) is a free, open-source platform built with LangGraph, Vue.js, FastAPI, and LightRAG to create smart agents using RAG knowledge bases and knowledge graphs. The latest v0.4.0-beta (Dec 2025) adds file uploads, multimodal image support, mind maps from files, evaluation tools, dark mode, and better graph visuals. It helps you quickly build and deploy custom AI agents for Q&A, analysis, and searches without starting from scratch, saving time and effort on development.

https://github.com/xerrors/Yuxi-Know

#typescript #devops #infra #infrastructure

Make beautiful isometric infrastructure diagrams

https://github.com/stan-smith/FossFLOW

#python #audio_generation #diffusion #image_generation #inference #model_serving #multimodal #pytorch #transformer #video_generation

vLLM-Omni is a free, open-source tool that makes serving AI models for text, images, videos, and audio fast, easy, and cheap. It builds on vLLM for top speed using smart memory tricks, overlapping tasks, and flexible resource sharing across GPUs. You get 2x higher throughput, 35% less delay, and simple setup with Hugging Face models via OpenAI API—perfect for building quick multi-modal apps like chatbots or media generators without high costs.

https://github.com/vllm-project/vllm-omni

#python

Bloom is a free, open-source tool that automates testing AI models for bad behaviors like bias or sycophancy. You define the behavior in a simple config file, add example chats if you want, and it runs four steps: understanding it, creating varied test scenarios, simulating talks with your target model (like Claude or GPT via APIs), and scoring results with metrics like how often the issue appears. View interactive transcripts easily. This saves you hours of manual work, lets you quickly compare models on fresh tests to avoid overfitting, and gives reliable, reproducible insights into AI safety—perfect for researchers building trustworthy systems.

https://github.com/safety-research/bloom

#python #ai_tool #darkweb #darkweb_osint #investigation_tool #llm_powered #osint #osint_tool

Robin is an AI tool that searches and scrapes the dark web, refines queries with large language models, filters results, and produces a concise investigation summary you can save or export, with Docker and CLI options and support for multiple LLMs (OpenAI, Anthropic, Gemini, local models) to fit your workflow. This helps you save hours of manual searching by automating multi-engine dark-web searches, scraping Onion sites via Tor, filtering noise with AI, and producing ready-to-use reports for faster, more focused OSINT investigations.

https://github.com/apurvsinghgautam/robin

#jupyter_notebook

DINOv3 offers powerful self-supervised vision models from Meta AI, like ViT up to 7B parameters and ConvNeXt, pretrained on 1.7B web or satellite images. Load them easily via PyTorch Hub, Hugging Face Transformers (v4.56+), or timm (v1.0.20+), with code examples for features, depth, detection, and segmentation. You benefit by using these top-performing, dense features without fine-tuning or labels—saving time and compute for tasks like classification, object detection, and zero-shot analysis on your images.

https://github.com/facebookresearch/dinov3