# HuiaHub

> AI tools, open-source projects, agent skills and practical guides, each researched from primary sources, with every fact linked and every page dated.

HuiaHub is an editorial directory of AI tools, open-source AI projects, agent skills, videos and guides. Pages are researched and checked by hand in editing sessions, then published; nothing is written or published by an automated job. Every page lists its sources and shows when it was last checked. A fact the sources do not state is shown as unknown.

Edited by [Hamza Ahmad Aslam](https://hamzaahmadaslam.com). Corrections and suggestions: [contact page](https://huiahub.com/contact).

## Tools

- [Tools](https://huiahub.com/tools): AI tools with pricing, free tier and API facts from official pages
- [ngram](https://huiahub.com/tools/ngram): AI video generator that turns a prompt or screen recording into onboarding, training, demo and explainer videos for teams.
- [Refs](https://huiahub.com/tools/refs): Video reference library that turns launch films and explainers into blueprints and prompt recipes Claude and other agents can build from.
- [PocketWebTools for Mac](https://huiahub.com/tools/pocketwebtools): Mac app that runs open AI models up to 35B parameters offline for chat, transcription, OCR, upscaling and video clipping.
- [AgentDock](https://huiahub.com/tools/agentdock): A Mac app that shows Claude Code, Codex and other coding agents as status dots and lets you answer them from one Inbox.
- [ej](https://huiahub.com/tools/ej): An 11 MB open-weight model from HITL Systems that answers routing, yes/no and priority questions about text on CPU.
- [KernelAI](https://huiahub.com/tools/kernelai): An iOS and Android app that runs 53 open-weight models offline on your phone, with no account or subscription.
- [Lune](https://huiahub.com/tools/lune): MCP server that grounds AI agents in full-text computer science papers from top-tier conferences, for students and researchers.
- [Scrape Anything](https://huiahub.com/tools/context-dev-scrape-anything): Context.dev's scraping API that turns any URL into Markdown, HTML, JSON or screenshots for AI agents and apps.
- [Polylane](https://huiahub.com/tools/polylane): Watches production logs, metrics and traces, investigates issues and opens pull requests with fixes for your team to review.
- [JarvisCore](https://huiahub.com/tools/jarviscore): Open-source Python framework for multi-agent systems with peer-to-peer coordination, layered memory and a credential broker.
- [Lloyal](https://huiahub.com/tools/lloyal): A TypeScript framework for AI apps that run an open-weight model inside the app, offline and without an API key.
- [Agent Activity](https://huiahub.com/tools/agent-activity): Mac app that shows what AI coding agents do in Xcode: live sessions, build timelines, errors and test results in one window.
- [WeftCut](https://huiahub.com/tools/weftcut): A free, open-source desktop video editor that AI agents drive over MCP while you watch each edit and can undo it.
- [Clair](https://huiahub.com/tools/clair): Answer Claude Code permission prompts, questions and plan reviews from a Wear OS watch or an Android phone.
- [Communicate](https://huiahub.com/tools/communicate): AI support agent for website chat that answers from your help content with citations, runs approved tasks and hands off to people.
- [Singularity](https://huiahub.com/tools/singularity): Desktop development environment that runs several AI coding agents in parallel, one ticket and one Git worktree each.
- [Clef](https://huiahub.com/tools/clef): Cloudflare's open-weight decision models that return typed answers with probabilities, hosted on Workers AI.
- [Earlyn](https://huiahub.com/tools/earlyn): Mac app that keeps an encrypted, searchable memory of your screen and meetings, with local MCP access for AI coding agents.
- [Yubi](https://huiahub.com/tools/yubi): Push-to-talk dictation for the Mac that types where your cursor is, plus an AI helper that points to controls on screen.
- [qarunbook](https://huiahub.com/tools/qarunbook): A test runbook for web and mobile apps: an AI assistant drafts the plan, and testers or agents record results over MCP.
- [Muse Gadgets](https://huiahub.com/tools/muse-gadgets): Meta's open-source SDKs and firmware for building ESP32 and Raspberry Pi devices that connect to its Muse assistant.
- [Sapien](https://huiahub.com/tools/sapien): Market research service that builds synthetic populations from consumer data to simulate reactions to products, prices and campaigns.
- [Sellio](https://huiahub.com/tools/sellio): Shared support inbox and website chat widget for small teams, with an optional AI agent trained on your site, docs and FAQs.
- [Langfuse](https://huiahub.com/tools/langfuse): Open-source tracing, evaluation and prompt management for LLM apps and agents, as a cloud service or self-hosted.
- [n8n](https://huiahub.com/tools/n8n): Workflow automation platform for technical teams, with a visual editor, code steps, AI agent nodes, and cloud or self-hosted deployment.
- [Gemini Notebook](https://huiahub.com/tools/notebooklm): Google's source-grounded research notebook, formerly NotebookLM: add your documents and get cited answers, summaries and Audio Overviews.
- [Runway](https://huiahub.com/tools/runway): Web and mobile app plus API for generating and editing AI video, images and audio with Runway's Gen-4.5 and other models.
- [Suno](https://huiahub.com/tools/suno): AI music generator that makes full songs with vocals from a text prompt or your own lyrics, plus editing tools and a browser DAW.
- [Descript](https://huiahub.com/tools/descript): Audio and video editor where you edit a recording by editing its transcript, with AI tools for cleanup, clips and translation.
- [ElevenLabs](https://huiahub.com/tools/elevenlabs): AI audio platform for text to speech, voice cloning, dubbing, music and voice agents, with a web app and a developer API.
- [Replicate](https://huiahub.com/tools/replicate): A cloud API for running, fine-tuning and deploying open-source and proprietary AI models, billed by usage.
- [LM Studio](https://huiahub.com/tools/lm-studio): Desktop apps for running open LLMs locally, with a local API server and Bionic, an agent for documents, code and automation.
- [Ollama](https://huiahub.com/tools/ollama): An open-source tool and API for running open models on your own machine or on Ollama's cloud.
- [Devin Desktop](https://huiahub.com/tools/windsurf): Cognition's AI code editor, formerly Windsurf, for running and reviewing local and cloud coding agents from one app.
- [GitHub Copilot](https://huiahub.com/tools/github-copilot): GitHub's AI coding assistant for code completion, chat, agent mode, code review and cloud agents across editors, the CLI and GitHub.
- [Claude Code](https://huiahub.com/tools/claude-code): Anthropic's coding agent that reads a codebase, edits files and runs commands from the terminal, IDE, desktop app or web.
- [Cursor](https://huiahub.com/tools/cursor): An AI code editor and coding agent that plans, writes and reviews code on your desktop, in the terminal, in Slack and in the cloud.
- [pi pod](https://huiahub.com/tools/pi-pod): Self-hosted server that runs pi coding-agent sessions in isolated remote sandboxes you can reach from a terminal or phone.
- [Offrun](https://huiahub.com/tools/offrun): A macOS workspace for running and reviewing multiple coding agents side by side.

## Projects

- [Projects](https://huiahub.com/projects): open-source AI repositories with momentum, maintenance and licence
- [debpalash/VoiceStudio](https://huiahub.com/projects/voicestudio): VoiceStudio is a desktop app that runs speech AI on your own computer: it clones and designs voices, generates speech from text, dubs videos, transcribes audio and has a dictation widget.
- [QwenAudio/qwen-audio-agent](https://huiahub.com/projects/qwen-audio-agent): qwen-audio-agent is a realtime voice runtime that sits in front of a coding or task agent.
- [rohitg00/ai-engineering-from-scratch](https://huiahub.com/projects/ai-engineering-from-scratch): AI Engineering from Scratch is a free, MIT-licensed course on GitHub that teaches how AI systems work by having you write each piece yourself, starting with linear algebra and ending with…
- [DietrichGebert/ponytail](https://huiahub.com/projects/ponytail): Ponytail is a rule set and skill collection that pushes an AI coding agent to write as little code as a task needs.
- [ifixai-ai/iFixAi](https://huiahub.com/projects/ifixai): iFixAi is an open-source Python tool that audits AI agents and models by running a set of test inspections and returning an A to F grade with a written report. The free engine has 60 inspections.
- [addyosmani/agent-skills](https://huiahub.com/projects/agent-skills): Agent Skills is a repository of 25 Markdown workflows for AI coding agents, covering software work from specification and planning through testing, review, performance, security, deployment and…
- [calesthio/OpenMontage](https://huiahub.com/projects/openmontage): OpenMontage is a Python toolkit that turns an AI coding assistant such as Claude Code, Cursor, Copilot, Windsurf or Codex into a video production setup.
- [stablyai/orca](https://huiahub.com/projects/orca): Orca is a desktop development environment for running multiple AI coding agents side by side, with each task isolated in its own Git worktree.
- [jundot/omlx](https://huiahub.com/projects/omlx): oMLX is a local LLM inference server for Macs with Apple Silicon, built on MLX and managed from a menu bar app.
- [affaan-m/ECC](https://huiahub.com/projects/affaan-m-ecc): ECC is an open-source system that adds reusable engineering workflows, skills, agents, memory, rules, hooks and security checks to coding agents.
- [Panniantong/Agent-Reach](https://huiahub.com/projects/agent-reach): Agent Reach is a Python command-line tool that prepares an AI agent to read and search sites that are awkward to reach from a script, including Twitter/X, Reddit, YouTube, Bilibili, XiaoHongShu and…
- [Graphify-Labs/graphify](https://huiahub.com/projects/graphify): Graphify turns a software project and related material into a queryable knowledge graph. It parses code locally with tree-sitter, connects structures such as calls, imports and inheritance, and…
- [AlexsJones/llmfit](https://huiahub.com/projects/llmfit): llmfit is a terminal tool, written in Rust, that inspects your hardware and tells you which open-source large language models it can run well.
- [JuliusBrussee/caveman](https://huiahub.com/projects/caveman): Caveman is a skill plus a local proxy that reduce the tokens a coding agent spends, both in what it writes back and in what it reads.
- [DeusData/codebase-memory-mcp](https://huiahub.com/projects/codebase-memory-mcp): codebase-memory-mcp is an MCP server, written in C, that parses a code repository into a knowledge graph so an AI coding agent can ask structural questions (what calls this function, which HTTP…
- [nextlevelbuilder/ui-ux-pro-max-skill](https://huiahub.com/projects/ui-ux-pro-max-skill): UI UX Pro Max is an agent skill that hands coding assistants a searchable library of design guidance, so a request such as a landing page for a spa comes back with a matching style, color palette,…
- [headroomlabs-ai/headroom](https://huiahub.com/projects/headroom): Headroom is a Python and TypeScript tool that shrinks the text an AI agent sends to a model (tool output, logs, files, RAG chunks and chat history) so each request uses fewer tokens.
- [ARahim3/DigUp](https://huiahub.com/projects/digup): DigUp is a free macOS menu bar app that searches the files on your Mac by what is in them, using Google DeepMind's EmbeddingGemma 2 model running locally. You pick the folders it may read.
- [rtk-ai/rtk](https://huiahub.com/projects/rtk-ai-rtk): RTK (Rust Token Killer) is a command-line proxy that trims the terminal output of common developer commands before an AI coding agent reads it, so less of the agent's context fills up with noise.
- [MemPalace/mempalace](https://huiahub.com/projects/mempalace): MemPalace is a Python memory layer for AI assistants: it saves your past conversations and project files word for word on your own machine, then lets you or your agent search them by meaning.
- [TencentCloud/TencentDB-Agent-Memory](https://huiahub.com/projects/tencentdb-agent-memory): TencentDB Agent Memory is a self-hosted memory service from Tencent Cloud that stores what AI agents learn and shares it across agents and team members.
- [HKUDS/nanobot](https://huiahub.com/projects/nanobot): nanobot is a self-hosted personal AI agent framework in Python from HKUDS. You run it on your own computer or server and talk to it through a browser WebUI, the terminal, or chat apps such as…
- [Egonex-AI/Understand-Anything](https://huiahub.com/projects/understand-anything): Understand Anything turns a codebase or supported knowledge base into an interactive knowledge graph for exploring files, functions, classes, dependencies, and relationships.
- [StarTrail-org/PixelRAG](https://huiahub.com/projects/pixelrag): PixelRAG is a retrieval system that turns web pages, PDFs and images into screenshot tiles and searches those images instead of extracted text.
- [code-yeongyu/oh-my-openagent](https://huiahub.com/projects/oh-my-openagent): OmO (the repository is oh-my-openagent) is a terminal agent for coding and other work, started with a single omo command.
- [RyanCodrai/turbovec](https://huiahub.com/projects/turbovec): turbovec is an in-process vector index for Python and Rust that stores embeddings at 2 or 4 bits per dimension and searches the compressed vectors directly.
- [deepseek-ai/deepseek-harness](https://huiahub.com/projects/deepseek-harness): DeepSeek Harness is an open-source agent runtime from DeepSeek AI that runs coding agents through a Web UI, headless jobs, SDK connections or ACP automation clients.

## Skills

- [Skills](https://huiahub.com/skills): agent skills with install commands and a security check of what their files do
- [Upstash Vector TypeScript SDK guide](https://huiahub.com/skills/upstash-vector-skills): Quick-start guidance and an entry point for Upstash Vector features, TypeScript SDK usage, integrations and supported frameworks.
- [Dataframely for polars data frames](https://huiahub.com/skills/dataframely-skills): Practices for polars data processing with dataframely: Schema and Collection definitions, .validate() and .filter(), type hints and testing.
- [Python cheat sheets](https://huiahub.com/skills/py-crazyguitar): A Python reference covering syntax, concurrency, networking, databases, ML/LLM work and HPC, for Python questions, interview prep, debugging and development.
- [Slack GIF creator](https://huiahub.com/skills/slack-gif-creator-anthropics): Knowledge and Python utilities for making animated GIFs optimized for Slack, covering size constraints, validation tools and animation concepts.
- [Discernment nudge](https://huiahub.com/skills/discernment-nudge-anthropics): Appends two or three follow-up questions after a substantive answer or draft, helping the user check facts, probe reasoning and spot missing context.
- [Claude Academy guide](https://huiahub.com/skills/academy-guide-anthropics): Recommends a matching course, tutorial or use case from Claude Academy, Anthropic's learning hub, when a user asks how to use Claude or a Claude product.
- [Web artifacts builder](https://huiahub.com/skills/web-artifacts-builder-anthropics): Builds complex, multi-component claude.ai HTML artifacts with React, Tailwind CSS and shadcn/ui, for work needing state, routing or shadcn/ui components rather than single-file artifacts.
- [Theme factory](https://huiahub.com/skills/theme-factory-anthropics): Styles slides, documents, reports, HTML landing pages and other artifacts with one of 10 preset color and font themes, or a new theme generated on request.
- [Internal communications writer](https://huiahub.com/skills/internal-comms-anthropics): Helps Claude write internal communications such as status reports, leadership updates, 3P updates, newsletters, FAQs, incident reports and project updates in a company's preferred formats.
- [Avoid AI writing](https://huiahub.com/skills/avoid-ai-writing): Audits and rewrites text to remove AI writing patterns, with a detect-only mode, an edit-in-place mode for files, optional voice profiles and an iterate-to-convergence pass.
- [Excel spreadsheet tools](https://huiahub.com/skills/xlsx-anthropics): For any task where a spreadsheet file is the main input or output: reading, editing, fixing, creating, cleaning or converting .xlsx, .xlsm, .xltx, .csv or .tsv files.
- [Web app testing with Playwright](https://huiahub.com/skills/webapp-testing-anthropics): A toolkit for testing local web applications with Playwright: checking frontend behavior, debugging the UI, taking browser screenshots and reading browser logs.
- [Skill creator](https://huiahub.com/skills/skill-creator-anthropics): Creates new agent skills, edits and improves existing ones, runs evals and benchmarks on them, and tunes a skill's description so it triggers more reliably.
- [PowerPoint (pptx) files](https://huiahub.com/skills/pptx-anthropics): Used whenever a .pptx or .potx file is involved: creating decks, reading or extracting text, editing presentations, combining or splitting slides, and working with templates, layouts, notes or…
- [PDF processing](https://huiahub.com/skills/pdf-anthropics): Handles PDF tasks: reading and extracting text or tables, merging, splitting, rotating, watermarking, creating, filling forms, encrypting or decrypting, extracting images and OCR on scanned files.
- [MCP server builder](https://huiahub.com/skills/mcp-builder-anthropics): A guide for building MCP servers that let language models use external services and APIs through well-designed tools, in Python with FastMCP or TypeScript with the MCP SDK.
- [Frontend design](https://huiahub.com/skills/frontend-design-anthropics): Guides an agent toward distinctive, deliberate visual design when building or reworking a UI, covering aesthetic direction, typography and choices that avoid templated defaults.
- [Anthropic brand guidelines](https://huiahub.com/skills/brand-guidelines-anthropics): Applies Anthropic's brand colors and typography to artifacts that should carry Anthropic's look, when brand colors, style guides or company design standards matter.
- [Algorithmic art with p5.js](https://huiahub.com/skills/algorithmic-art-anthropics): Helps an agent create original generative art in p5.js with seeded randomness and adjustable parameters, for requests about flow fields, particle systems or code-based art.
- [Word document tools](https://huiahub.com/skills/docx-anthropics): Teaches an agent to create, read, edit, and manipulate Word documents and templates, including formatting, images, comments, tracked changes, and content extraction.
- [Document co-authoring](https://huiahub.com/skills/doc-coauthoring-anthropics): Guides users through a structured process for writing documentation, proposals, technical specs, decision documents, and similar substantial content.
- [Claude API](https://huiahub.com/skills/claude-api-anthropics): Reference for Claude API and Anthropic SDK work, including models, pricing, parameters, streaming, tools, MCP, agents, caching, token counting and migrations.
- [Canvas design](https://huiahub.com/skills/canvas-design-anthropics): Creates original static visual art as PNG or PDF files from a design philosophy, for posters, artwork, designs and similar pieces.

## Videos

- [Videos](https://huiahub.com/videos): learning videos, each with our own write-up and chapters
- [How Inference Keeps Up with Our Insatiable Need for Tokens — Cerebras & Gimlet Labs](https://huiahub.com/videos/how-inference-keeps-up-with-the-need-for-tokens): A 22-minute AI Engineer talk for inference engineers that covers runtime tricks like paged KV cache and chunked prefill, then system-level disaggregation and kernel-writing agents.
- [Sandboxes Aren't Optional: 4 Ways to Isolate Coding Agents — Robert Brennan, OpenHands](https://huiahub.com/videos/sandboxes-4-ways-to-isolate-coding-agents): A 14-minute talk from the OpenHands CEO comparing four ways to sandbox coding agents, for developers who run agents with broad permissions.
- [Stop Renting Intelligence: The Train-to-Deploy Loop for Specialized AI — Fireworks AI](https://huiahub.com/videos/train-to-deploy-loop-for-specialized-ai): A 19-minute Fireworks AI talk on training open models on your own data and redeploying them in a continuous loop, for teams weighing closed-model costs.
- [Three Questions Every Agent Engineer Should Be Able To Answer](https://huiahub.com/videos/three-questions-every-agent-engineer-should-answer): A 14-minute LangChain talk on finding, reproducing and fixing agent failures using production traces, for engineers who run agents in production.
- [n8n Full Course – Architect Scalable AI Automations from Scratch](https://huiahub.com/videos/n8n-full-course-scalable-ai-automations): A 43-minute freeCodeCamp.org course that takes n8n from first workflow to agents, Slack approvals, error handling and cost monitoring, for builders new to the tool.
- [Build & Train a GLM-5.3-Flash Model From Scratch with Python](https://huiahub.com/videos/build-and-train-glm-5-3-flash-from-scratch): A freeCodeCamp.org walkthrough for developers who want to read, train and RL-tune a small GLM-5.3-Flash-style model in Python, with all code on GitHub.
- [How AI Models Scale Beyond a Single GPU Across LLM Workloads](https://huiahub.com/videos/how-ai-models-scale-beyond-a-single-gpu): A nine-minute IBM Technology explainer for engineers who need the basic vocabulary of distributed LLM inference: the four parallelism types and prefill versus decode.
- [Stop Fine-Tuning to Fix Retrieval Problems — Anant Srivastava](https://huiahub.com/videos/stop-fine-tuning-to-fix-retrieval-problems-anant-srivastava-qfll): A 20-minute talk for engineers about to fine-tune a model to fix wrong answers: a clear rule for what belongs in the prompt, in retrieval and in weights.
- [What Makes Open Models Fast in Production — Sujee Maniyam, Nebius](https://huiahub.com/videos/what-makes-open-models-fast-in-production-sujee-maniyam-nebius-tre1): A 20-minute talk for engineers serving open LLMs: a layer-by-layer list of inference optimizations, from NVFP4 to speculative decoding, after a Nebius product intro.
- [An Interaction Is All You Need — Ivan Leo, Google DeepMind](https://huiahub.com/videos/an-interaction-is-all-you-need-ivan-leo-google-deepmind-8avb): Ivan Leo of Google DeepMind explains the Gemini Interactions API and Managed Agents in 17 minutes, useful for developers building agents on Gemini.
- [I Built a Personal AI Agent on a Raspberry Pi — Jeremy Adams, Neo4j](https://huiahub.com/videos/i-built-a-personal-ai-agent-on-a-raspberry-pi-jeremy-adams-neo4j-ouze): Jeremy Adams of Neo4j builds a wearable agent on a Raspberry Pi with NanoClaw, WhatsApp and graph memory, a 20-minute talk for builders of small personal agents.
- [Your LLM App Returned 200 OK. It Was Still Wrong. — Marina Petzel, Datadog](https://huiahub.com/videos/your-llm-app-returned-200-ok-it-was-still-wrong-marina-petzel-datadog-rtoj): An 18-minute talk for engineers who run LLM apps in production: which cost, safety and quality metrics to track beyond latency and error rates.
- [YOLO Mode, Safely: MicroVM Sandboxes for Any Agent — Rowan Christmas, Docker](https://huiahub.com/videos/yolo-mode-safely-microvm-sandboxes-for-any-agent-rowan-christmas-docker-oe-l): A 12-minute Docker talk for developers who let coding agents run unattended: why an agent's built-in guardrails fall short and how microVM sandboxes contain an agent.
- [Why 99% Accurate Browser Agents Still Fail — Derek Meegan, Browserbase](https://huiahub.com/videos/why-99-accurate-browser-agents-still-fail-derek-meegan-browserbase-5xi): A 17-minute talk for anyone moving a browser agent from demo to production: the math of compounding step errors and an architecture that works around it.
- [Your Coding Agent Is 6 Months Out of Date — Jakub Hojsan, Exa](https://huiahub.com/videos/your-coding-agent-is-6-months-out-of-date-jakub-hojsan-exa-ckhp): For developers whose coding or review agents misjudge new code: why a search tool alone falls short, and how Exa trims pages down to what the model needs.
- [Image Decision Models for RPA: Forms, Scans and Screenshots](https://huiahub.com/videos/image-decision-models-for-rpa-forms-scans-and-screenshots-l8yx): Sam Witteveen tries two open decision models, ImaJev 4B and Jev Omni, on forms and screenshots for RPA, useful for automation builders who need a model to make yes/no or multiple-choice calls.
- [In-Context Retrieval with Siddharth Gollapudi - Weaviate Podcast #146!](https://huiahub.com/videos/in-context-retrieval-weaviate-podcast): A 56-minute Weaviate Podcast episode for search and RAG engineers on using an LLM's attention over a million-token context as the retriever, and where that breaks.
- [Gemini 3.8 Flash TTS  with Voice Cloning](https://huiahub.com/videos/gemini-3-8-flash-tts-with-voice-cloning-gdub): Sam Witteveen walks through Gemini 3.8 Flash TTS, from voice design and cloning to pricing, with Colab code, for developers choosing a text-to-speech model.
- [Nemotron 3 Diarization - Who Said That?](https://huiahub.com/videos/nemotron-3-diarization-who-said-that-pzuu): Sam Witteveen explains speaker diarization and runs NVIDIA's Nemotron 3 Diarization on a full podcast, a 13-minute look for anyone building multi-speaker transcription.
- [Transformers.js v4.3: Structured Output in the browser](https://huiahub.com/videos/transformers-js-v4-3-structured-output-in-the-browser-innv): An 11-minute Hugging Face walkthrough of Transformers.js 4.3 for JavaScript developers who need models in the browser to return valid JSON every time.
- [Flow matching for Beginners | @ariG23498](https://huiahub.com/videos/flow-matching-for-beginners-arig23498-hxqz): A 28-minute beginner's explanation of flow matching on the Hugging Face channel that leans on visuals over equations, for people who want the intuition behind modern image generators.
- [Run Local Models in Pi: llama.cpp, GGUF, and the /llama Command](https://huiahub.com/videos/run-local-models-in-pi-llama-cpp-gguf-and-the-llama-command-5dsf): A six-minute Hugging Face quickstart for Pi users who want a private, offline coding agent running on a local GGUF model through llama.cpp.
- [Agent Memory EXPLAINED - Complete Architecture](https://huiahub.com/videos/agent-memory-explained-complete-architecture-ayfz): A 28-minute walk through how Mem0 stores, ingests and retrieves agent memories, for developers who want long-term memory that goes past a single vector search.
- [Prompt Caching Explained: Stop Overpaying for AI Agents](https://huiahub.com/videos/prompt-caching-explained-stop-overpaying-for-ai-agents-skm4): A 17-minute explainer on what prompt caching stores and the agent-code mistakes that break it, for developers paying for long AI agent sessions.
- [But what is cross-entropy? | Compression is Intelligence Part 2](https://huiahub.com/videos/but-what-is-cross-entropy-compression-is-intelligence-part-2-glyg): 3Blue1Brown explains where the loss function for training LLMs comes from, building cross-entropy up from optimal codes. Good for anyone who trains or studies language models.
- [THIS is how you build agentic RAG systems that work](https://huiahub.com/videos/this-is-how-you-build-agentic-rag-systems-that-work-suon): An 11-minute Weaviate explainer for developers whose RAG answers are poor: it splits an agentic RAG pipeline into five systems and where each one breaks.

## Weekly digest

- [Weekly digest](https://huiahub.com/news): a short, sourced round-up of what changed in the tools, projects and skills we track
- [The week on HuiaHub: sandboxed agents, RAG tooling and local models](https://huiahub.com/blog/weekly-digest-2026-10-12): 41 new HuiaHub pages this week: sandboxing coding agents, RAG retrieval, local models on Macs and phones, Claude Haiku 5.5 and six how-to guides.
- [This week on HuiaHub: local models, MCP guides and 18 new videos](https://huiahub.com/blog/weekly-digest-2026-10-04): Local LLMs with Ollama and vLLM, MCP servers in Python, Claude Code hooks, 18 talks and new agent skills: everything HuiaHub published so far.

## Blog

- [Blog](https://huiahub.com/blog): guides, comparisons and explainers with numbered sources
- [AI agents found in US bank accounts: what banks are blocking and why](https://huiahub.com/blog/ai-agents-found-in-us-bank-accounts): See which banks are restricting AI agents in bank accounts, how read-only access differs from credential sharing, and which claims remain unverified.
- [PHP critical security update CVE-2026-109618: Versions and patch steps](https://huiahub.com/blog/php-critical-security-update-cve-2026-109618-cve-2026-109619): Check the PHP critical security update CVE-2026-109618, identify affected servers, install fixed PHP versions, and apply temporary settings if needed.
- [How NVIDIA fine-tuned Nemotron 3 for gold-level IOI and IMO 2026 results](https://huiahub.com/blog/fine-tuning-nemotron-for-ioi-and-imo): NVIDIA details how SFT, RL and test-time loops took Nemotron 3 to gold-level IOI and IMO 2026 results, and which checkpoints and datasets are open.
- [Microsoft rebuilds Agent Lightning to train agents in their real harnesses](https://huiahub.com/blog/agent-lightning-v1-0-agentic-rl-framework): Microsoft rebuilt Agent Lightning as a 3,500-line RL framework that trains agents through the harnesses they already run. What changed for RL teams.
- [GitHub Copilot local sandboxing reaches general availability](https://huiahub.com/blog/local-sandboxing-for-github-copilot-now-generally-available): GitHub made local sandboxing for Copilot generally available, limiting what agent-run commands can touch. Setup, OS needs and gaps for teams.
- [GitHub moves AI secret detection to a fine-tuned model and AI Credits](https://huiahub.com/blog/purpose-built-model-for-leaked-secret-detection): GitHub moved AI secret detection to a fine-tuned model and set AI Credit billing for push protection and Copilot checks. What changes for security teams.
- [Fine tune LLM LoRA adapters: data, training and a first test on one GPU](https://huiahub.com/blog/fine-tune-a-small-model-with-lora): Fine tune LLM LoRA adapters on one GPU: check that fine-tuning fits, prepare data, train with PEFT and TRL or Unsloth, and compare against the base model.
- [How to test MCP server tools with the Inspector, unit tests and a client script](https://huiahub.com/blog/test-an-mcp-server): Test MCP server tools in three layers: explore them in the MCP Inspector, unit test the tool logic, and run a scripted client check in CI.
- [Prompt caching with the OpenAI, Claude and Gemini APIs: what is cached and what breaks it](https://huiahub.com/blog/prompt-caching-openai-claude-gemini): Compare prompt caching API rules for OpenAI, Claude and Gemini, order prompts so the cache hits, avoid silent misses and measure savings from usage fields.
- [How to evaluate a RAG pipeline with Ragas: metrics, a test set and a first run](https://huiahub.com/blog/evaluate-a-rag-pipeline-with-ragas): Set up RAG evaluation with Ragas: build a small test set, match each metric to retrieval or generation, run a first score and compare two versions.
- [Best embedding model for RAG: how to choose by size, dimensions, languages and cost](https://huiahub.com/blog/choose-an-embedding-model-for-rag): Pick the best embedding model for RAG by language, context length, vector size, hosting and price, then test it on your own documents first.
- [Sandbox AI coding agent setup: permissions, containers and network limits](https://huiahub.com/blog/sandbox-ai-coding-agents): Set up a sandbox AI coding agent in Claude Code, Codex, Copilot or Docker, see what each boundary blocks, and spot the gaps it leaves open.
- [Claude Haiku 5.5 brings lower prices and an effort setting to Haiku](https://huiahub.com/blog/claude-haiku-5-5): Anthropic's Claude Haiku 5.5 cuts Haiku input pricing to $0.10 per million tokens and adds effort control. What API teams must change to migrate.
- [Intel details the GPU D270, an air-cooled inference card with up to 480 GB](https://huiahub.com/blog/intel-d270-gpu-ai-accelerator-up-to-480-gb-lpddr5x): Intel detailed the GPU D270 inference card: 480 GB LPDDR5X on partner cards, 160 GB on its own, 350 W PCIe. No price or ship date yet for buyers.
- [Character.ai enters Chapter 11 and puts the company up for a court-supervised sale](https://huiahub.com/blog/character-ai-has-initiated-voluntary-chapter-11-bankruptcy): Character.ai began a Chapter 11 sale process on October 12, 2026. What stays the same for users for now, and how to export chats before the sale ends.
- [QGIS bans LLM, AI and agent-assisted code in its main repository](https://huiahub.com/blog/qgis-enacts-moratorium-on-ai-contributions): QGIS approved a policy that bans LLM, AI and agent-assisted code in qgis/qgis. What changed for contributors, how the vote went and what is still open.
- [Gemini 4 Argon reaches the Gemini API, AI Studio and the Gemini app](https://huiahub.com/blog/gemini-4-argon-built-to-solve-your-hardest-problems): Gemini 4 Argon is now on the paid Gemini API at $2 input and $10 output per 1M tokens until November 30, with a 1M-token window and long decode.
- [NVIDIA releases Nemotron 3.5 Super VL and a 24GB Bolt model for local agents](https://huiahub.com/blog/open-from-data-center-to-desktop-nvidia-expands-nemotron-family-with-h): NVIDIA released Nemotron 3.5 Super VL, a 120B open vision model, and Bolt 74B-A9B for 24GB local use. What developers can run now and what is missing.
- [GitHub AI Scan now checks PHP, Terraform and Dockerfiles in pull requests](https://huiahub.com/blog/ai-scan-for-pull-requests-generally-available): GitHub's AI Scan for pull requests is now GA. It flags vulnerabilities in PHP, Bash, Terraform and other code CodeQL skips, billed in AI credits.
- [Google DeepMind releases SynthID Bio code, weights and Nature paper](https://huiahub.com/blog/introducing-synthid-bio): Google DeepMind published SynthID Bio, with Apache 2.0 code, a Nature paper and model weights for marking AI-generated protein data. What shipped.
- [Google announces Gemini 4 Argon, starting with vetted cyber defenders](https://huiahub.com/blog/gemini-4-argon-our-next-era-of-frontier-intelligence): Google announced Gemini 4 Argon with a 1M-token output limit and set API prices. Only vetted cyber defenders get it for now; developers must wait.
- [GitHub Copilot can now operate desktop apps on macOS and Windows](https://huiahub.com/blog/github-copilot-can-now-interact-with-desktop-apps-with-computer-use): GitHub Copilot CLI and the Copilot app can now operate desktop apps on macOS and Windows in public preview. How to turn it on, approvals and risks.
- [Cloudflare releases Clef decision models and a managed RL fine-tuning service](https://huiahub.com/blog/clef-open-weight-decision-models-and-new-rl-fine-tuning-platform): Cloudflare released Clef and Clef-flash, decision models that return typed answers with probabilities, on Workers AI and Hugging Face. Prices, specs and gaps.
- [Aleph Alpha releases Kolibri, an open-weight German and English MoE model](https://huiahub.com/blog/kolibri-a-sovereign-open-weight-model): Aleph Alpha released Kolibri, a 78B MoE model with 3.46B active parameters and Apache 2.0 weights. What changed, what it needs to run, and what is unknown.
- [NVIDIA adds a 64GB DGX Spark for local AI, with two-unit clustering](https://huiahub.com/blog/nvidia-dgx-spark-64gb-gives-developers-more-ways-to-build-and-scale-lo): NVIDIA's DGX Spark gets a 64GB configuration at $4,999 from six partners, plus two-unit clustering. What changes for local AI developers.
- [Ai2 opens AstaBrief 8B, the model that writes Asta's fast cited reports](https://huiahub.com/blog/open-sourcing-astabrief-the-fast-report-generation-model-in-asta): Ai2 released AstaBrief 8B, the open model behind Asta's Fast mode, with its SFT and DPO data. What it does, how it was trained and how to run it yourself.
- [GitHub Copilot deprecates four models: what replaces each one and where](https://huiahub.com/blog/selected-models-in-github-copilot-deprecated): GitHub removed Gemini 3.5 and 3.6 Flash, Kimi K2.7 Code and Claude Opus 4.7 from Copilot on October 2, 2026. Replacements, plan limits and admin steps.
- [Copilot code review gets API requests and a Balanced default effort level](https://huiahub.com/blog/copilot-code-review-api-support-and-new-default-effort-level): GitHub Copilot code review can now be requested through the REST and GraphQL APIs with a per-request effort level, and Balanced is the default effort.
- [Ultralytics YOLO object detection: predict, train and export a custom model](https://huiahub.com/blog/object-detection-with-ultralytics-yolo): Run Ultralytics YOLO object detection from Python or the CLI, train on your own labeled images and export to ONNX, after checking the AGPL-3.0 licence terms.
- [ComfyUI local image generation: install, model folders and custom node safety](https://huiahub.com/blog/run-image-models-locally-with-comfyui): Set up ComfyUI local image generation: pick an install, put model files in the right folders, run a first workflow and vet custom nodes before installing.
- [LLM structured outputs with JSON schema in the OpenAI and Claude APIs](https://huiahub.com/blog/structured-outputs-openai-and-claude): Get schema-valid JSON from OpenAI and Claude with LLM structured outputs, JSON schema and strict tool use, and know what each API guarantees.
- [Claude Code hooks: run formatters, tests and guards automatically](https://huiahub.com/blog/claude-code-hooks-guide): Set up Claude Code hooks that format files after edits, run tests and block risky commands, using today's event names, JSON input and exit codes.
- [How to write a SKILL.md for Claude Code: structure, examples, pitfalls](https://huiahub.com/blog/write-a-skill-md-for-claude-code): Write a SKILL.md for Claude Code that loads at the right time: frontmatter fields, trigger-ready descriptions, scripts, file locations and testing.
- [Tracing an LLM app with Langfuse: setup, traces, costs and redaction](https://huiahub.com/blog/trace-llm-apps-with-langfuse): Add Langfuse tracing to a Python LLM app, choose cloud or self-hosted, read traces and token costs, and mask personal data before it leaves your app.
- [How to test prompts and LLM outputs with promptfoo, with a CI step](https://huiahub.com/blog/evaluate-llm-outputs-with-promptfoo): Set up LLM evaluation with promptfoo: write a YAML test suite with assertions, compare two models side by side and fail a CI build on regressions.
- [Serve an open model with vLLM behind an OpenAI-compatible API](https://huiahub.com/blog/serve-an-open-model-with-vllm): Install vLLM, start a vLLM OpenAI compatible server with vllm serve, tune GPU memory and context length, and point the OpenAI SDK at your own GPU.
- [Open source text to speech you can run yourself: Piper, Kokoro and more](https://huiahub.com/blog/open-source-text-to-speech-models): Compare open source text to speech engines Piper, Kokoro, Coqui TTS and Chatterbox by licence, languages and hardware, then pick one you can ship.
- [Transcribe audio locally with Whisper: openai-whisper, whisper.cpp or faster-whisper](https://huiahub.com/blog/transcribe-audio-locally-with-whisper): Transcribe audio locally with Whisper using openai-whisper, whisper.cpp or faster-whisper, export SRT or VTT subtitles, and match the tool to your hardware.
- [pgvector, Qdrant, Chroma or Milvus: how to choose a vector database](https://huiahub.com/blog/choose-a-vector-database): Choose a vector database for RAG or search by comparing pgvector, Qdrant, Chroma and Milvus on deployment, filtering, index types and licence.
- [Build a RAG pipeline with Postgres and pgvector, step by step](https://huiahub.com/blog/rag-with-postgres-and-pgvector): Set up RAG with pgvector in your existing Postgres: a chunks table, an HNSW or IVFFlat index, the right distance operator and a filtered retrieval query.
- [How to connect Claude Code MCP servers and keep them safe](https://huiahub.com/blog/connect-mcp-servers-to-claude-code): Add Claude Code MCP servers with the claude mcp command, pick local, project or user scope, share them in .mcp.json and vet each server first.
- [How to build an MCP server in Python with the official SDK](https://huiahub.com/blog/build-an-mcp-server-in-python): Build an MCP server in Python with the official mcp SDK: add a tool and a resource, run it over stdio or HTTP, and test it in the MCP Inspector.
- [llama.cpp vs Ollama vs LM Studio: which local model runner fits your setup](https://huiahub.com/blog/llama-cpp-vs-ollama-vs-lm-studio): Compare llama.cpp vs Ollama vs LM Studio on APIs, GPU and OS support, model formats and licences, then pick a local model runner with a decision table.
- [How to run a local LLM with Ollama and call it from your code](https://huiahub.com/blog/run-a-local-llm-with-ollama): Run a local LLM with Ollama on macOS, Windows or Linux, call it over its REST and OpenAI-compatible APIs, size models to your RAM and manage disk space.

## Topics

- [Topics](https://huiahub.com/topics): one hub per subject, linking the tools, projects, skills and guides that belong together
- [AI agents](https://huiahub.com/topics/ai-agents): Autonomous and semi-autonomous agents that plan, call tools and finish multi-step tasks, with the frameworks, runtimes and guides for building them.
- [AI automation](https://huiahub.com/topics/ai-automation): Workflow automation powered by AI: no-code and low-code builders, agent pipelines and the guides for connecting models to everyday tools.
- [AI developer tooling](https://huiahub.com/topics/ai-dev-tooling): SDKs, frameworks and libraries for building AI apps, with the open projects, developer tools and guides that go with them.
- [AI observability & evaluation](https://huiahub.com/topics/ai-observability): Tracing, evals, guardrails and monitoring for AI apps: the platforms, open-source libraries and guides for measuring what a model does.
- [Agent skills](https://huiahub.com/topics/agent-skills): Reusable SKILL.md instructions for Claude Code and other coding agents, each checked file by file for what it runs and what it needs.
- [Coding agents](https://huiahub.com/topics/coding-agents): AI that writes, reviews and ships code: coding agents, their permission and sandbox modes, and the skills and guides for working with them.
- [Image generation](https://huiahub.com/topics/image-generation): Diffusion models, image editing and generation tools, from hosted services to models you run locally, with the guides that show how.
- [Inference & model serving](https://huiahub.com/topics/inference-serving): Fast inference engines and model servers, with throughput, batching and deployment guides for serving models on one machine or a cluster.
- [Large language models](https://huiahub.com/topics/llms): Open-weight and commercial large language models, with the APIs, fine-tuning methods, prompting techniques and guides for putting them to work.
- [Local AI](https://huiahub.com/topics/local-ai): Running models on your own hardware, privately: local runners, quantised models, hardware limits and the guides for setting them up.
- [Model Context Protocol (MCP)](https://huiahub.com/topics/mcp): Model Context Protocol servers and clients that connect models to tools and data, with guides for building, testing and securing them.
- [Multimodal AI](https://huiahub.com/topics/multimodal): Models that combine text, image, audio and video, with the tools, open projects and guides for using them beyond plain text.
- [RAG](https://huiahub.com/topics/rag): Retrieval-augmented generation: grounding a model in your own documents with embeddings, chunking, vector search and evaluation, step by step.
- [Speech, TTS & STT](https://huiahub.com/topics/speech): Speech recognition, text-to-speech and voice cloning: transcription tools, open models, diarization and the guides for running them.
- [Vector databases](https://huiahub.com/topics/vector-databases): Storage and search for embeddings: vector databases, Postgres extensions and libraries, compared on hosting, filtering and cost.
- [Video generation](https://huiahub.com/topics/video-generation): Text-to-video models, video editing tools and avatar generators, with what each one costs and the open projects behind them.

## Trust pages

- [About HuiaHub](https://huiahub.com/about): HuiaHub tracks AI tools, open-source projects, agent skills and videos, checks each one against its own sources, and dates every page.
- [Contact](https://huiahub.com/contact): How to reach HuiaHub with corrections, listing requests, security reports or privacy questions.
- [Methodology](https://huiahub.com/methodology): How HuiaHub researches, checks, scores and updates AI tools, open-source projects, agent skills, videos and articles.
- [Editorial policy](https://huiahub.com/editorial-policy): HuiaHub's standards for sources, independence, AI assistance, updates and corrections.
- [AI disclosure](https://huiahub.com/ai-disclosure): Where HuiaHub uses AI for research and drafting, and where a person decides.

## Optional

- [llms-full.txt](https://huiahub.com/llms-full.txt): a short plain-text summary of every page listed here
- [Sitemap](https://huiahub.com/sitemap.xml)