AI builder, open-source maintainer & creator of AI超元域.
I build agent tools, test AI in real projects, and share the process.
AI 技术 · 开源项目 · 实践笔记
Projects / Watch & build / Exploring / Writing ↗
Give agents memory that carries across sessions. I created and lead-maintain memory-lancedb-pro: hybrid vector + BM25 retrieval, cross-encoder reranking, and memory isolation across agents, users and projects. Watch the walkthrough ↗
|
MULTI-MODEL WORKSPACE
Ask once. Compare answers from multiple AI models in one desktop workspace. 
|
MULTI-AGENT COORDINATION
Bring CLI coding agents together with OpenClaw as the default agent. 
OpenClaw adaptation of HKUDS/ClawTeam. |
|
CLAUDE CODE FIELD NOTES
Explore notes on context, memory, model switching and multi-agent architecture. 
|
KNOWLEDGE & RETRIEVAL
Connect Microsoft GraphRAG to Open WebUI with local, global and web search. 
|
|
AGENT INTEROPERABILITY
Let OpenClaw agents discover and communicate with each other across servers. 
|
REUSABLE AGENT SKILL
Install, configure, operate and troubleshoot OpenClaw with an AI coding assistant. 
|
Contributing upstream: CJK token estimation and CJK trigram search from lossless-claw-enhanced were merged into lossless-claw. The original fork is retired; use upstream for new installs.
Hands-on AI tests, agent workflows and tutorials — with the projects and implementation details to explore for yourself.
From installation to hybrid retrieval.
|
Explore ClawTeam and OpenClaw in practice.
|
More from the lab: Unlimited-OCR-2 deployment & document tests · YouTube videos · Bilibili videos · Technical articles
- Agent systems — long-term memory, multi-agent coordination and interoperability.
- AI coding — Claude Code, Codex, context engineering and reusable workflows.
- Applied AI — RAG, model fine-tuning, multimodal tools and local deployment.
Build something useful. Share what you learn.
Find me on YouTube, Bilibili, X or the blog · Away from the keyboard: archery.