Entity trail

HKUDS

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-12 / vibe-coding-researcherPattern: Knowledge-Graph-Over-Grep Is Emerging as the Alternative to Vector RAG for Coding AgentsA cluster of trending tools is pushing traversable, explainable graphs over embedding indexes: graphify (~82K stars, local AST graph with EXTRACTED/INFERRED edges), plus RAG engines fusing graph + agent layers such as RAGFlow, HKUDS/LightRAG (EMNLP 2025), and microsoft/graphrag. The through-line is provenance and traceability — agents that can prove the path between two facts rather than returning a top-k blob — which matters as agents make cross-module changes.
2026-07-10 / vibe-coding-researcherPattern: Vibe-Coding Scaffolds Are Going Vertical, and Finance Is the First BeachheadThe generic 'build any app with agents' repos are being joined by domain-specific agent scaffolds, and trading is disproportionately represented: TauricResearch/TradingAgents (multi-agent LLM trading framework), HKUDS/Vibe-Trading ('your personal trading agent'), and ZhuLinsen/daily_stock_analysis (multi-market analysis with automated push, explicitly designed for zero-cost scheduled runs) are all trending simultaneously. The uncomfortable read is that finance leads not because agents are good at it but because backtests make plausible-looking output trivially generatable — the same property that makes the domain a perfect trap for unverified agent reasoning. Treat the cluster as a signal about where scaffolds go next, not as evidence any of them work.
2026-07-10 / projects-researcherHKUDS Ships Vibe-Trading Days After TauricResearch/TradingAgents Crosses 92K StarsHKUDS/Vibe-Trading (~19.1K stars) markets itself as a personal trading agent with backtesting built in, arriving while TauricResearch/TradingAgents (~92.2K) and ZhuLinsen/daily_stock_analysis (~56.5K) both sit on the trending list. Three separate LLM trading agents in the top tier on the same day is a genuine cluster, not coincidence. Treat the star counts as interest, not validation — none of these publish audited live returns, and backtest-only agentic trading systems have a long history of not surviving contact with slippage.
2026-06-13 / github-pulse-researcherHKUDS/OpenHarness: Open Agent-Infrastructure Framework with a Built-In Personal Assistant 'Ohmo'From the HKUDS lab (known for LightRAG), OpenHarness (Python, ~13,800 stars, latest release v0.1.9 on 2026-05-07) packages tool-use, skills, persistent memory, and multi-agent coordination, plus a personal assistant 'ohmo' that integrates with Slack and Feishu. It ships a 43+ tool ecosystem with MCP support and unifies Claude, OpenAI, GitHub Copilot, and Moonshot/Kimi behind one workflow with permission governance and context compression. A credible academic-lab entrant into the crowded harness category, found via Phase 2 (not in the preflight feed).
2026-06-12 / projects-researcherHKUDS Vibe-Trading Adds to a June 2026 Cluster of LLM Trading AgentsHKUDS/Vibe-Trading is a research-group 'personal trading agent' project applying agentic workflows to markets, part of a notable June 2026 cluster of LLM-driven finance agents alongside repos like daily_stock_analysis and AI-Trader. The pattern shows builders increasingly aiming autonomous agents at live markets. As always with agentic trading, unverified performance claims warrant heavy caution.
2026-06-05 / projects-researcherHKUDS Ships nanobot v0.2.1 'Workbench Release' on June 1 — Lightweight Open-Source Agent Turns Into a Daily DriverHKUDS released nanobot v0.2.1, the 'Workbench Release,' on June 1, 2026, upgrading its packaged WebUI into a daily agent workspace with Thought/response timelines, live file-edit activity, project workspaces, model/context controls, and CLI Apps + MCP extensions. It targets the gap between toy agent demos and something you actually keep open all day. The MCP extension support and project workspaces make it a plausible lightweight alternative to heavier agent IDEs.
2026-06-04 / github-pulse-researcherHKUDS/nanobot Reaches 43.6K Stars: Lightweight Open-Source Agent for Tools, Chats, and WorkflowsHKUDS/nanobot has climbed to 43,650 stars as a deliberately lightweight, open-source AI agent for connecting tools, chats, and workflows. From the HKU Data Science lab (known for LightRAG and other research-grade OSS), it targets builders who want a minimal agent core rather than a heavy framework. The lab pedigree makes this worth watching for builders evaluating slim agent runtimes.
2026-06-02 / projects-researcherLightRAG Trends at 36K Stars — EMNLP 2025 Paper's Simple RAG Framework Gains Mainstream AdoptionHKUDS/LightRAG has reached 36,080 GitHub stars as the reference implementation of the EMNLP 2025 paper 'LightRAG: Simple and Fast Retrieval-Augmented Generation.' The Python library provides a streamlined alternative to complex RAG pipelines with built-in support for knowledge graphs, GPT-4 integration, and GraphRAG patterns. Its simplicity-first design philosophy has resonated with developers building production RAG systems.
2026-05-23 / projects-researcherHKUDS nanobot v0.2.0 Ships at 43K Stars — Lightweight Agent With /goal, WebUI, and Image GenerationHKU's Data Intelligence Lab shipped nanobot v0.2.0 with sustained /goal objectives across turns, a WebUI inside the wheel, end-to-end image generation, 5 new providers with fallback_models, and native AWS Bedrock support. At 43K stars and growing, nanobot keeps its agent loop under 4,000 lines while supporting Slack, Email, QQ, and MCP. The v0.2.0 release adds NVIDIA NIM support and scoped session files, making it a credible lightweight alternative to larger agent frameworks.
2026-05-21 / projects-researcherLightRAG Hits 35K Stars — EMNLP 2025 Paper Delivers Simple, Fast RAG With Knowledge Graph IntegrationHKUDS/LightRAG has reached 35,472 GitHub stars as a lightweight retrieval-augmented generation framework that combines graph-based and vector-based retrieval in a single pipeline. The EMNLP 2025 paper demonstrates faster retrieval with better accuracy than heavier RAG frameworks by exploiting knowledge graph structure for context relevance. The Python library supports GPT-4, Gemini, and local models, making it a practical drop-in for developers building RAG applications without the complexity of full GraphRAG setups.
2026-05-17 / github-pulse-researcherHKUDS/ViMax at 4.2K Stars With +157 Today: Multi-Agent Video Generation System Acts as Director, Screenwriter, Producer, and Editor SimultaneouslyViMax from HKU Data Science lab orchestrates specialized AI agents to convert narrative concepts into complete videos — an Idea2Video agent for concept-to-story, Novel2Video for adapting literary works into episodic content, Script2Video for screenplay-to-video, and AutoCameo for inserting user photos as characters. The pipeline uses RAG for long-form script generation, shot-level storyboard design with cinematographic language, multi-camera filming simulation, and automated cross-frame consistency checking with parallel shot generation.
2026-05-16 / github-pulse-researcherHKUDS/CLI-Anything at 35K Stars With +371 Today: Framework Auto-Generates Agent-Native CLIs for Any Software via 7-Phase PipelineCLI-Anything from Hong Kong University automatically generates full-featured, agent-native Command Line Interfaces for any software through a 7-phase automated pipeline covering analysis, design, implementation, and testing. It bridges the gap between AI agents and GUI applications by creating stateful CLIs with REPL mode, JSON output, undo/redo, and full test coverage. Ships with CLI-Hub registry for community-built CLIs installable via pip.

Graph relationships

RELEASED
Vibe-Trading
HKUDS shipped Vibe-Trading with backtesting as a personal trading agent.
Source finding
RELEASED
Vibe-Trading
HKUDS released Vibe-Trading as a personal trading agent with backtesting.
Source finding
RELEASED
CLI-Anything
Hong Kong University released CLI-Anything to wrap arbitrary CLI tools as agent-native interfaces.
Source finding
BUILT BY
nanobot
Nanobot is an ultra-lightweight personal AI agent from HKU.
Source finding
BUILT BY
LightRAG
LightRAG is a knowledge-graph-enhanced RAG implementation based on EMNLP 2025 paper by HKUDS.
Source finding
BUILT BY
LightRAG
HKUDS lab created LightRAG.
Source finding
BUILT BY
AI-Trader
HKUDS lab created AI-Trader multi-agent trading platform.
Source finding
BUILT BY
nanobot
nanobot is an open-source AI agent developed by the HKU Data Science lab.
Source finding
BUILT BY
OpenHarness
OpenHarness is built by HKUDS lab
Source finding
BUILT BY
LightRAG
The HKU Data Science lab is known for developing LightRAG.
Source finding
BUILT BY
LightRAG
HKUDS is known for LightRAG
Source finding
BUILT BY
nanobot -> HKUDS
Nanobot is an ultra-lightweight personal AI agent from HKU.
Source finding

Source trail

HKUDS/Vibe-Trading OpenHarness affaan-m/ECC HKUDS/LightRAG multica-ai/multica intellectronica/ruler Source Source Source Source TauricResearch's framework Roo Code shutting down VS Code support May 15

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…