潜力新星（90天）

The most capable Calibre MCP server — full read/write tools plus multilingual semantic search over your whole ebook library or a single book. For Claude & any MCP client.

Reproducible kit to deploy DeepSeek-V4-Flash-DSpark on a 2× NVIDIA DGX Spark (GB10) cluster: vLLM TP=2 over QSFP 200GbE, NVFP4 KV, DSpark speculative decoding, 1M context, systemd self-heal. Apache-2.0.

Omega-Forge: unofficial code to finetune or train VGGT-Omega (VGGT-Ω) on your own dataset, for use while official training code is not yet available. Additionally includes a standalone zero-GPU 3D reconstruction visualizer.

SkyPortal SDK & CLI - Open Source

AI 面试训练系统：跨会话能力建模闭环 — LangGraph 编排 + 三层上下文融合出题 + Mem0 式画像更新 + RAG 知识增强 + SM-2 复习调度。FastAPI · React 19 · LlamaIndex · Qdrant

TokenFlow is an enterprise-grade, token-aware load balancer and AI gateway for vLLM. It prevents GPU OOM crashes using Redis-backed admission control, semantic caching, and dynamic token routing. Built in TypeScript for extreme bare-metal performance.

Agent skill that prepares a Linux server with NVIDIA GPUs for LLM workloads - driver and CUDA, Docker with NVIDIA Container Toolkit, staged verification gates, compose presets for vLLM, Infinity, OpenWebUI and Ollama

Production-style Retrieval-Augmented Generation document Q&A: LangChain ingestion, Chroma/pgvector store, hybrid retrieval, citation-grounded answers and a streaming chat API.