最有价值项目

The in-browser vector store that remembers — a Rust→WASM HNSW engine that persists to OPFS and stays consistent across tabs. Private, offline semantic search in 3 lines.

Rules you can enforce, state you can derive, context you can query: plain files in your repo, under whichever coding agent you use.

🔒 Privacy-first MCP server for Claude using PostgreSQL + Ollama. Local alternative to cloud-based code context with full data sovereignty. No API keys, no external calls, 100% local.

A clean, modular implementation of an Agentic RAG (Retrieval-Augmented Generation) system built with a production-ready architecture.

A low-token visual evidence compiler for text-only coding agents. Convert images into compact Visual Evidence Packets (VEP) for DeepSeek, Codex, Claude Code, and other text-only models.

It is a voice bot based on LLM.

AI-powered multimodal chat app with real-time responses, file support, token tracking, and dark mode. Built with Next.js. Open source under MIT.

Official implementation of JAMIE, Joint variational Autoencoders for Multimodal Imputation and Embedding

Finalist at Brainhack TIL 2024: Team 12000SGDPLUSHIE

Heterogeneous Edge Realtime Measurement and Execution System: A Unified Open-Source Framework for Realtime Multimodal Physiological Sensing, Edge AI, and Intervention in Closed-Loop Smart Healthcare Applications

A self-hosted omni-modal agentic retrieval platform that understands, searches, and explores documents, images, audio, and video through semantic routing, hybrid retrieval, adaptive agents, and traceable citations.

A research-oriented training and evaluation framework for ECG-Language Models (ELMs)

Disentangled Graph Variational Auto-Encoder for Multimodal Recommendation with Interpretability, IEEE TMM

Codex Skill：让纯文本模型（DeepSeek）借助 StepFun step-3.7-flash 获得看图能力 | Give text-only Codex models (DeepSeek) image understanding via StepFun step-3.7-flash

Collects a multimodal dataset of Wikipedia articles and their images

Learning Adaptive Fusion Bank for Multi-modal Salient Object Detection

The code and weight for LoVA. LoVA is a novel model for Long-form Video-to-Audio generation. Based on the Diffusion Transformer (DiT) architecture, LoVA proves to be more effective at generating long-form audio compared to existing autoregressive models and UNet-based diffusion models.

[ACL 2023] VSTAR is a multimodal dialogue dataset with scene and topic transition information

The official implementation of NeurlPS 2025 D&B paper: IndustryEQA: Pushing the frontiers of Embodied Question Answering in Industrial Scenarios.

Diagnose vLLM inference servers