仓库/Skill 库

GitHub 上最新 Agent 框架的实时索引，按发布时间而非 star 数排序。Live index of the newest agent frameworks shipping on GitHub, sorted by recency not stars

在 LIBERO 与 RoboTwin2.0 上，基于状态融合解码与冻结视频骨干网络训练面向机器人操作的轻量级世界动作模型。Train lightweight world action models for robot manipulation using state-fusion decoding and frozen video backbones on LIBERO and RoboTwin2.0.

自托管、多用户转录平台：录制或上传音频。支持说话人标记、带时间戳的转录、跨录音识别说话人、摘要、提取行动项，并可使用自有 OpenAI 兼容 LLM 与转录内容对话。你的音频、你的服务器、你的模型。已在笔记本 RTX4070、台式机 RTX3090 与 RTX5090 上测试。Self-hosted, multi-user transcription platform: record or upload audio. Speaker-labeled, timestamped transcripts, Recognize speakers across recordings, Summarize, extract action items and chat over your transcripts with your own OpenAI-compatible LLM. Your Audio, your Server, your Model. Tested on Laptop RTX4070, Desktop RTX3090 and RTX5090

用于生成基于 Transformer 的 LLM 完整注意力头热力图的一组脚本。A set of scripts to generate full attention-head heatmaps for transformer-based LLMs

自托管、可扩展的 AI agent 平台，具备受治理的 MCP 工具、持久化项目记忆、持久工作流、能力包、自主执行，以及分布式本地模型计算能力。Self-hosted, extensible AI agent platform with governed MCP tools, persistent project memory, durable workflows, capability packs, autonomous execution, and distributed local-model compute.

🛠️ 通过 Local-LLM（一款基于 BERT 模型的轻量级 Python 库）在安全环境中实现离线 NLP 工作流，确保可复现性与可靠性。🛠️ Enable offline NLP workflows with Local-LLM, a lightweight Python library for secure environments using the BERT model, ensuring reproducibility and reliability.

驱动 LLM agent 完成结构化学术文献综述的通用工具包。Topic-agnostic toolkit for driving an LLM agent through a structured academic literature review

金融与生活交易学习 RAG 与 QuantConnect LEAN 回测工作流（FastAPI、Qdrant、pgvector、Docker）。금융·생활거래 학습 RAG와 QuantConnect LEAN 백테스트 워크플로우 (FastAPI, Qdrant, pgvector, Docker)

LM Studio 的 Web 前端——支持浏览器访问、自适应记忆、多用户鉴权与 MCP 工具。Web frontend for LM Studio — browser access, adaptive memory, multi-user auth, MCP tools

从零开始构建开源预训练 LLM，附带已发布的数据、训练代码、消融实验与结果，以高效提升模型质量。Build open pretraining LLMs from scratch with released data, training code, ablations, and results to improve model quality efficiently

🤖 Hermes Agent 2026：可学习、可自适应的开源 AI 助手 🚀。🤖 Hermes Agent 2026: AI Assistant That Learns & Adapts 🚀 | Open Source

AI Agent 集群的自托管指挥中心——创建 Agent、派发任务、见证交付。FastAPI + Next.js + Docker。Self-hosted command center for AI agent fleets — create agents, dispatch tasks, watch them ship. FastAPI + Next.js + Docker.

准确捕获 macOS 听写内容，不改写你的措辞，保留真实输入并规避常见应用问题Capture macOS dictation accurately without rewriting your words, keeping your input true to what you speak and avoiding common app issues.

MCP server，可从 URL 提取任意视频（包括微信视频号）。可选择性地分析并提取场景感知的去重关键帧，并对视频进行转录。完全本地运行MCP server that extracts any video from URL(inc/ WeChat channels). Optionally analyzes and extracts scene-aware deduplicated key frames and transcribes the video. ALL LOCAL

面向商业智能 Agent 的 MCP server——基于真实 KPI 分析提供 6 个 tools / 6 个 resources / 1 个 prompt；兼容 Claude Desktop、LangGraph、CrewAI、DSPy。基于 FastMCP。MCP server for business-intelligence agents — 6 tools / 6 resources / 1 prompt over real KPI analytics; works from Claude Desktop, LangGraph, CrewAI, DSPy. FastMCP.

开源核心的「AI Agent 版 Spotify」——可自托管的可运行 Agent 工件注册中心：skill、bundle、loop 与 personality。docker compose up 即可在 60 秒内启动可用注册中心，零注册。MCP 原生，MPL-2.0 许可。Open-core 'Spotify for AI agents' — a self-hostable registry of runnable agent artifacts: skills, bundles, loops, and personalities. docker compose up → working registry in 60s, zero signup. MCP-native, MPL-2.0.

面向殡葬/保险/遗产服务机构的 To B 多租户 AI 平台。包含案件管理、审计日志、知识库、团队 RBAC、许可证及数据导出。基于 FastAPI + LangGraph 构建。To B multi-tenant AI platform for funeral/insurance/estate service organizations. Case management, audit logs, knowledge base, team RBAC, license & data export. Built on FastAPI + LangGraph.

面向 AI coding agents 的 local-first 分层记忆中枢——自动采集、LLM 整合、hybrid retrieval、实时 Neural Universe 监控、日常自审计。兼容 Claude Code / Codex / 任何 MCP 客户端。Local-first hierarchical memory brain for AI coding agents — auto-capture, LLM consolidation, hybrid retrieval, live Neural Universe monitor, daily self-audit. Claude Code / Codex / any MCP client.

将 Justin Sun 的营销逻辑提炼为心智模型、决策启发式与文案模板，用于注意力驱动的增长。Distill Justin Sun's marketing logic into mental models, decision heuristics, and copywriting templates for attention-based growth.

Run the native 284B-A13B DeepSeek-V4-Flash-0731 LLM locally on a single laptop CPU: pure C, 8 GB RAM minimum, no GPU, best TPOT 0.892 s/token. ｜ 在笔记本单颗 CPU 上本地运行原生 284B-A13B DeepSeek-V4-Flash-0731 大模型：纯 C，最低 8 GB 内存，无需 GPU，最优 TPOT 0.892 秒/token。

Ultimate AI Spellbook Workflows 2026 – 构建多平台 Skill 与自动化Ultimate AI Spellbook Workflows 2026 – Build Multi-Platform Skills & Automation

面向会议的对话智能：将转录文本转化为摘要、决策、行动项、风险与机会，支持多租户自托管。FIAP NEXT Challenge 2026。Conversation intelligence for meetings: turns transcripts into summaries, decisions, action items, risks and opportunities. Multi-tenant, self-hosted. FIAP NEXT Challenge 2026.

基于 Python（Anthropic SDK）的多 Agent 系统，为私人教练生成个性化训练与饮食计划——内置免费规则引擎 + 可选 LLM 层、显式状态编排器、Streamlit 仪表盘与确定性安全校验器。A multi-agent system in Python (Anthropic SDK) that generates personalized workout routines and meal plans for a personal trainer — free rule engine + optional LLM layer, explicit-state orchestrator, Streamlit dashboard, and deterministic safety validator.

面向 Kubernetes 环境的 air-gapped RAG pipeline。在 LibreChat/LiteLLM/Keycloak/vLLM 之上增加文档摄取、强制的分类/可发布性标签、策展人审核与基于声明的访问控制检索，并以自定义 MCP 工具的形式对外暴露。Air-gapped RAG pipeline for Kubernetes environments. Adds document ingestion, mandatory classification/releasability tagging, curator review, and claims-based access-controlled retrieval on top of LibreChat/LiteLLM/Keycloak/vLLM — exposed as a custom MCP tool.

面向银行内部产品审计的深度研究 LLM Agent + RAG 平台，提供带引用报告、PDF 导出与 pgvector 支持。Deep-research LLM agent + RAG platform for internal bank-product audit — cited reports, PDF export, pgvector

🐙 AI Agent Pipeline 基于意图将查询路由至文档、天气或聊天模块，结合 LangGraph、ChromaDB 与 LangSmith，实现跨 CLI 与 UI 的模块化、可观测工作流🐙 AI Agent Pipeline routes queries by intent to docs, weather, or chat, with LangGraph, ChromaDB, and LangSmith for modular, observable workflows across CLI and UI.

使用 ConsensusMind 这一自主 AI Agent 对区块链共识机制进行全面深入的研究，涵盖文献综述、模拟与学术写作。🤖 Conduct thorough research on blockchain consensus mechanisms with ConsensusMind, an autonomous AI agent for literature review, simulations, and academic writing.

面向 Node 与浏览器的 AI 数据库对话：用自然语言提问、审视查询、获取答案。支持 Postgres、MySQL、SQLite、DuckDB、Oracle、MongoDB。具备只读 AST 守卫、React UI、MCP server、VS Code 与 JetBrains 集成。AI database chat for Node and the browser: ask in plain language, review the query, get answers. Postgres, MySQL, SQLite, DuckDB, Oracle, MongoDB. Read-only AST guard, React UI, MCP server, VS Code & JetBrains.

🤖 自主研究论文分析器——下载 arXiv 论文，通过 HuggingFace Transformers 运行本地 LLM 分析，并将 AI 生成的文献综述发布至 GitHub Pages。🤖 Autonomous Research Paper Analyzer — Downloads arXiv papers, runs local LLM analysis via HuggingFace Transformers, and publishes AI-generated literature reviews to GitHub Pages

用知识图谱刻画 AI/ML 模型从创建到部署的完整生命周期。Knowledge Graph to capture AI/ML model lifecycle from creation through deployments.

Bruno 的 MCP server：AI agent 可编写、读取并运行 .bru 与 .yml API 集合。无需 bru CLI，与 Bruno 字节级一致。兼容 Claude、Cursor、Codex。MCP server for Bruno: an AI agent authors, reads and runs .bru and .yml API collections. No bru CLI, byte-parity with Bruno. Claude, Cursor, Codex.

AI Agent 集群的开源任务控制中心 —— 对话即可配置，Agent 负责执行，人类负责审批。Open-source mission control for AI agent fleets — chat to provision, agents to operate, humans to approve.

教学用 LLM 应用teaching LLM applications

AI agent memory 与基础设施全景——912 个系统 × 68 列的对比目录，覆盖记忆层、agent 框架、运行时、vector store、知识图谱、MCP server、benchmark。支持按类型化边、谱系、引用进行检索。AI agent memory & infrastructure landscape — comparative catalog of 912 systems × 68 columns covering memory layers, agent frameworks, runtimes, vector stores, knowledge graphs, MCP servers, benchmarks. Searchable with typed edges, lineages, citations.

韩国加密货币 × AI 垂直媒体与社区——alpha.moss.land：频道立场、AI 简报、RAG、8 个带业绩记录的 AI persona、12 工具 MCP server。Korean crypto × AI vertical media + community — alpha.moss.land. Channel stance, AI briefs, RAG, 8 AI personas with track records, 12-tool MCP server.