仓库/Skill 库

The Convergence Gap 的代码与 artifact：指令微调模型何时收敛到下一 token 预测Code and artifacts for The Convergence Gap: when instruction-tuned models settle on next-token predictions.

本地化 MCP 数据平台。自然语言问题按规则分派到向量搜索、NL2SQL、知识图谱三条通道并行查询，通过 RRF 融合后由本地 LLM 基于证据作答。无需调用外部 API。온프렘 MCP 데이터 플랫폼. 자연어 질문을 벡터검색·NL2SQL·지식그래프 3레인으로 규칙 기반 분기해 병렬 조회하고 RRF로 융합, 로컬 LLM이 근거 기반으로 답한다. 외부 API 호출 없음.

基于 Llama 3.2 的本地多模态 RAG AI 助手，可与 PDF、图片和视频对话，完全离线，零数据外泄。Local multimodal RAG AI assistant where you can chat with PDFs, images, and video using Llama 3.2, fully offline, zero data exposure.

prompt工具箱

一道 Python 守门机制：在不分类提示词的前提下，阻止不可信的 agent 数据去构造敏感的工具调用参数。A Python gate that prevents untrusted agent data from authoring sensitive tool-call arguments—without classifying prompts.

开源 Raspberry Pi 5 rover 项目，用于建图、视觉、语音、本地 LLM 及液态神经网络实验。Open-source Raspberry Pi 5 rover project for mapping, vision, voice, local LLMs, and liquid-neural-network experiments.

一次编写 AI Agent 知识，编译输出为所有 Agent 配置格式——CLAUDE.md、AGENTS.md、.cursor/rules 等。Write your AI agent knowledge once. Compile it to every agent config format - CLAUDE.md, AGENTS.md, .cursor/rules and more.

🔍 使用多种 prompt 技巧在多步数学问题上分析 Mistral-7B 模型的数学推理能力。🔍 Analyze the mathematical reasoning abilities of the Mistral-7B model using diverse prompting techniques on multi-step math problems.

AI agents 的身份、记忆与协作平台——通过 CLI、MCP 和 Go 服务器提供事实、记忆、密钥管理及 agent 间消息传递。Identity, memory & collaboration platform for AI agents — facts, memories, secrets, and inter-agent messaging via a CLI, MCP, and a Go server.

数据在被构建成有意义的东西之前只是噪声——将原始数据转化为真正可用的系统，涵盖欺诈检测、RAG pipeline、计算机视觉追踪器以及数据仓库等领域。FAST NUCES 数据科学本科生，在"搞坏东西"和"交付产品"之间反复横跳。持续构建，持续学习，欢迎合作和实习交流Data is noise until someone builds something meaningful out of it, turn raw data into systems that actually work, from fraud detection and rag pipelines to computer vision trackers and data warehouses. DS undergrad at FAST NUCES, somewhere between breaking things and shipping them. always building, always learning, open to collabs and interns

通过实战教程构建有效的 AI 评测，使用真实 agent 提升生产系统技能🎯 Build effective AI evaluations through a hands-on tutorial, using a realistic agent to enhance your production system skills.

基于 FastAPI、LangChain、ChromaDB 和 Google Gemini API 构建的快速、轻量级 RAG 聊天机器人。A fast, lightweight Retrieval-Augmented Generation (RAG) chatbot built with FastAPI, LangChain, ChromaDB, and the Google Gemini API.

AI 增强的 IT 运维工作台——规格驱动、多提供商、本地优先，将事件与服务请求转为带 KB 引用的摘要，配备可审计会话、PII 脱敏与沙箱化操作。AI-augmented IT ops workbench — spec-driven, multi-provider, local-first. Turns incidents and service requests into KB-cited summaries with auditable sessions, PII redaction, and sandboxed actions.

销售收件箱的 AI 后台：读取每条询盘，仅推荐实际有货的商品，以你的口吻起草回复——并在任何内容发出前等待人工点击批准。An AI back office for your sales inbox: it reads every enquiry, recommends only what is actually in stock, drafts the reply in your voice — and waits for a human to click Approve before anything sends.

学术知识图谱系统——符号驱动的研究发现，配合轻量级向量检索。Academic knowledge graph system — symbol-driven research discovery with lightweight vector retrieval

Embrace the choas（拥抱混乱）Embrace the choas

某私有组织的公开主页，包含多个 .git 大学项目，涵盖软件工程、全栈 Web 开发、数据科学与工程（AI/ML）、嵌入式系统、GIS 等先进计算技术，分别在 Uppsala、Canterbury、Auckland 完成，并含多次海外正式交换经历；多数项目涉及广泛协作，且均以 distinction 等级通过Public profile for a private org containing multiple .git university projects in software engineering, full-stack web dev, data science & eng (AI/ML), embedded systems, GIS. & other advanced computing tech completed at Uppsala, Canterbury, Auckland & formal exchanges abroad; many involving extensive collaboration, all averaging with distinction

自动更新的全球最热门 AI 仓库集合，汇聚 LLM、Agent、RAG 等前沿项目。An auto-updating collection of the world's most popular AI repositories, bringing together LLMs, agents, RAG, and other cutting-edge projects.

以挑战为中心的多跳 RAG 研究枢纽：五类挑战分类法、208 篇文献综述快照，以及七篇综述参考库。Research hub for challenge-centered multi-hop RAG: five-challenge taxonomy, 208-work review snapshot, and seven-survey reference library.

贝叶斯统计研究语料库：面向贝叶斯统计（推断、计算、先验、模型选择、层次模型、非参数、深度贝叶斯、应用）的数据驱动、自动化校验文献综述Bayesian Statistics Research Corpus: Data-driven, auto-validated literature review for Bayesian statistics (inference, computation, priors, model selection, hierarchical, nonparametric, deep Bayesian, applications)

基于 AI 的简历优化工具，可根据任意职位描述定制简历。使用 Streamlit、LlamaIndex 与 Nebius AI 构建。AI-powered resume optimizer that tailors your resume to any job description. Built with Streamlit, LlamaIndex and Nebius AI.

在消费级 GPU 上自动化机器学习模型从论文阅读到实验执行与同行评审的完整研究工作流Automate the full research workflow from paper reading to experiment execution and peer-review for machine learning models on consumer GPUs.

将 Claude-English 重写为通俗英语。基于开源模型的 MCP 服务器、REST API 和 Claude Code hookRewrites Claude-English into plain English. MCP server, REST API and Claude Code hook, on an open-source model.

2026 年由 AI 驱动的 Things 3 CLI 自动化工具——macOS 智能任务管理器AI-Powered Things 3 CLI Automation Tool 2026 – Smart Task Manager for macOS

Go 工具链：将 Bourbaki PDF 转换为带标签的 Markdown 语料，并完成翻译与习题求解Go toolchain that turns Bourbaki PDFs into a tagged Markdown corpus, translates it, and solves the exercises

与供应商无关的 coding-agent 框架，具备确定性工作流图、持久化证据与 fail-closed 沙箱执行。Provider-neutral coding-agent harness with deterministic workflow graphs, durable evidence, and fail-closed sandboxed execution

一个为 Model Context Protocol 提供策略执行能力的 gateway。聚合 MCP server，强制工具调用策略，扫描结果中的 prompt injection 与凭证泄露，并审计一切。Rust 实现。A policy-enforcing gateway for the Model Context Protocol. Aggregates MCP servers, enforces tool-call policy, scans results for prompt injection and leaked credentials, and audits everything. Rust.

一套面向 LLM-based RTL 验证开放研究问题的架构：分块多表征记忆（BM25 / HNSW / 结构化事实）、基于本地 7B LLM 的多 Agent 编排器，以及具备漂移保护的反馈驱动优化。An architecture answering open research questions on LLM-based RTL verification: chunked multi-representation memory (BM25 / HNSW / structured facts), a multi-agent orchestrator over a local 7B LLM, and feedback-driven refinement with drift protection.

在小型 LLM 中植入认知美德的激活引导实验——三项发现与一个失败模式数据集（DOI 见 README）Activation-steering experiments on installing epistemic virtues in small LLMs — three findings + a failure-mode dataset (DOIs in README).

CreditFlow AI Agent——小商户贷款预审与自动小额放款 Agent（Google Cloud AI × Solana 实现的下一代 AI Agentic 黑客松项目）CreditFlow AI Agent - 소상공인 대출 사전심사 및 자동 소액대출 집행 에이전트 (Google Cloud AI × Solana로 구현하는 차세대 AI Agentic 해커톤)

一款个人知识引擎，注重综合而非简单转录——每个声明都附带回到原始来源的引用。为 LLM agent 提供写入时综合 + 带引用的检索能力。A personal knowledge engine that synthesizes instead of transcribes — every claim carries a citation back to its source. Write-time synthesis + cited retrieval for LLM agents.

面向开源库的机器可读错误知识库——MCP server、REST API 与静态数据集，将每条错误消息映射到抛出该错误的源代码行。Machine-consumable error knowledge base for open-source libraries — MCP server, REST API, and static dataset mapping every error message to the source line that raised it

面向空间 LLM 与地理空间 AI Agent 的生产工程模式——验证优先架构、地理空间 RAG、prompt-to-spatial-SQL、tool routing 与评估。Production engineering patterns for spatial LLMs and geospatial AI agents — validation-first architecture, geospatial RAG, prompt-to-spatial-SQL, tool routing, and evaluation.

多语言 AI agent 服务 — 基于单一 schema 驱动的 WebSocket 协议，提供知识对话、工具调用、持久检查点、人机协同与多方会话。基于 smooth-operator-core 引擎构建（5 种语言特性一致）。可部署到 Kubernetes、AWS serverless 或本地运行。托管于 lom.smoo.ai。Polyglot AI agent service — knowledge chat, tools, durable checkpoints, human-in-the-loop, and multi-participant conversations over one schema-driven WebSocket protocol. Built on the smooth-operator-core engine (5-language parity). Deploy to Kubernetes, AWS serverless, or run locally. Hosted at lom.smoo.ai.

AI 驱动的学术研究助手，使用 RAG 与 LLM 进行论文分析、生成文献综述、识别研究空白、产出深度解读，并生成带来源引用的演示文稿。AI-powered academic research assistant that uses RAG and LLMs to analyze research papers, generate literature reviews, identify research gaps, create deep explanations, and generate presentations with source citations.