Havuz

RikkaHub Agent · Pure — an Android AI agent that survives long runs. Independent hardening pass on the rikkahub on-device agent: usage ledge

Open-source, self-hostable, always-on personal agent with its own computer. Runs on Cloudflare Workers or Node.

Let an AI companion get sleepy, fall asleep, dream, and be woken by you. 让 AI 伴侣会困、会睡、会做梦，也会被你叫醒。

Typed long-term memory for AI agents. One core, an MCP server, and a Hermes plugin.

Mental-R1: Aligning LLM Reasoning for Mental Health Assessment

A living AI civilisation you can watch: tiny creatures discover, build, teach and pass knowledge on, with any LLM in their minds.

Nucleo Agentic Runtime

Daily tracker of AI repositories trending on GitHub, with weekly analysis of rising topics and dominant languages, charts and CSV data, auto

Smart component library for AI apps. With Chai UI, every AI model is controlled from one place, through a unified interface.

Hermes plugin: one /crew message becomes verified work on the kanban board. A coordinator owns every card until its proof passes.

Stanford CS336《从头构建大语言模型》2026 中配版：18 篇中文技术博客 + 19 篇「图-字幕」原始文稿，基于 video2blog-skill 整理。

Voice I/O for LLM chat UIs: dictation, read-along and a voice-reactive aura, for any OpenAI-compatible speech engine.

The last agentic harness you'll ever need. Bring any model; Pennant does the rest on your Mac and iPhone, and asks before anything is sent,

An open-source recreation of AI Co-Scientist: a multi-agent scientific research harness.

Redacts secrets and personal data in LLM API traffic and restores them in responses

EasyRead, akademik PDF’leri yerel bir kütüphanede düzenleyip LLM’lerle Çinceye çevirmeyi ve çeviriyle özgün metni yan yana okuyup not almayı sağlayan bir uygulamadır.

Neden yükseliyor · Repo iki günlükken 577 yıldız almış; PDF üzerinde çeviri, özgün metinle karşılaştırma, AI soru-cevap ve referans yönetimini bir araya getiren somut ürün demosu ilgi çekmiş olabilir. Ancak yükselişin kesin nedeni bilinmiyor; son 28 günlük commit ve farklı yazar verileri ölçülmediğinden geliştirme temposu ve katkı çeşitliliği değerlendirilemiyor.

yomiyasu, Codex, Claude Code ve Cursor gibi agent ortamlarında AI tarafından üretilen Japonca metinleri daha doğal ve okunabilir kılmaya yönelik bir Agent Skill sunuyor.

Neden yükseliyor · Bir günlük repo yaşına karşılık 625 yıldız, dikkat çekici erken ilgiye işaret ediyor. Ancak bu ilginin kullanıcı benimsemesine dönüşüp dönüşmediği bilinmiyor; son 28 günde 2 commit ve 1 farklı yazar, bakım ve topluluk katkısı için henüz güçlü kanıt oluşturmuyor.

Spend fewer tokens on the same work. A local CLI that finds wasted tokens in Claude Code, Codex, opencode and your own API calls.

A structured, versioned archive of adversarial prompt engineering techniques against Large Language Models — for AI safety research, red-tea

Автоотклики на hh.ru: письмо под каждую вакансию, ответы на анкеты и анализ резюме нейросетью, отказы и приглашения из чатов, подъём резюме

A free pit crew for your expensive coding agent: the superpowers development workflow, run by cheap workers.

An AI workforce on your desktop. Describe the work; a team of AI agents does it in the open and hands it back checked. Any model, any harnes

Review code changes by what they do, and learn the concepts behind them.

Official codebase for the paper "Sharpening Tax in Post-Training"

ИИ-ассистент в вашей базе 1С: читает конфигурацию и документы, отвечает по вашим данным, заполняет и создает документы. Бесплатная внешняя о

Self-hosted LLM gateway audit station: model fingerprint attribution, cache/tool/web/thinking evidence, IQ and integrity scores.

AI agents and workflows for Kotlin Multiplatform, as small typed graphs: loops, parallel steps, streaming, save points and human approval. I

Run Qwen3.8-Flash-Next on ONE RTX 3090 (24 GB) + 64 GB RAM: 128K context, up to 2,100 tok/s prefill, 43–51 tok/s decode. vLLM runtime with h

Evidence-grounded academic research agent with claim verification, prior-work analysis, and persistent research workspaces.

Claude, ChatGPT, Grok ve Gemini'nin yerel CLI'larını kullanarak kendi kendine ilerleyen bir web grup sohbeti çalıştırır.

Neden yükseliyor · 35 yıldızın 1 günlük bir repoda görülmesi erken ilgiye işaret ediyor; ancak son 28 günlük commit ve yazar verileri ölçülmediğinden ilginin nedeni veya projenin geliştirme ivmesi doğrulanamıyor. Dört tanınmış modeli tek bir görsel sohbet odasında buluşturması dikkat çekici bir demo sunuyor.

Open decision models from 0.8B to 27B: a calibrated probability for every option in one pass, and a short thought only when unsure. Engine f

AI论文系统 · AI论文生成、AI论文写作，一键论文生成，降AIGC、降重、答辩PPT · OEM白标招商加盟

Autumn-Recruit Algo Notes(秋招算法学习仓库)，面向大模型时代的算法、工程与AI学习共建仓库。从数学基础、数据结构，到机器学习、深度学习、LLM、Agent、AI-Infra、语音与数据科学，通过「理解→推导→手写 → 运行 → 验证 → 复盘构建可持续迭

Mojito — the local, evolving alternative to cloud personal AI like Meta Muse.

STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization

把 DeepSeek Harness 打包成能直接装的安卓 APK：不用 Termux、不用 root、不用配电脑，读屏/点按/虚拟屏/定时任务和整套技能都随包；它还能在手机上自己编译、打包、签名、发布。 Android APK for DeepSeek Harness — no

先进且稳定的企业级智能分析平台；面向企业问数、洞察分析的Data Agent，基于数据分析方法论打造可审计、可治理、可复现workflow

OpenAI- and Anthropic-compatible API for Duck.ai, driven by a real Chrome browser via Playwright. Streaming chat, image generation, session

GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution

Turn messy GitHub Stars into organized Lists automatically with AI agent skills and npm CLI.

Voice-first desktop assistant for Windows — talk to your PC and it works inside real apps, safely. By VORNEX.

Governed AI agents for sensitive data and business-critical apps. Agents never hold credentials: a gateway checks every call against per-res

Instinct: single-forward-pass decision models (choice / true-false / score) returning a probability per candidate

A local-first language-learning app for the desktop. Lessons, stories and conversation practice are generated by an AI model running on your

把 OpenCode 的免费模型（big-pickle、LongCat、MiMo、Nemotron、Space Bunny 等）接进 DSH 的模型选择器，作为 opencode-xdbridge 分组使用，并在设置 → OpenCode-XD 里提供一个可视化的交互入口。

你是几B的模型？How many B are you? A meme quiz where you play the LLM, open-sourced with the full creative trail: every prompt, briefs, and a 3-par

DAEDALUS: Bootstrapping Agent Memory from Self-Generated Tasks. Code, paper and blog post.

Several coding agents, one repo, at the same time: a reproducible lab to measure how they coordinate, and Médula, a kernel that decides who

AIHOT, LLM destekli içerik seçimi ve kümelendirmeyle farklı sektörlere uyarlanabilen, self-hosted bir haber ve gündem sitesi çerçevesidir.

Neden yükseliyor · Repo bir günde 2.842 yıldız almış; bu, dikkat çekici erken ilgiye işaret ediyor. Farklı sektörlere uyarlanabilir self-hosted bir ürün olarak sunulması ve özelleştirme adımlarının belgelenmesi ilgiye katkı sağlamış olabilir; ancak bu ilgiyi hangi etkenin yarattığına dair veri yok. Son 28 günlük commit ve farklı yazar sayısı ölçülmediğinden geliştirme etkinliği değerlendirilemiyor.

Coordinate Claude, Codex and Cursor windows with shared tasks, acknowledgements, prechecks, parallel tests, source-bound proofs and independ