Matthew B. Abbott

I appreciate your holding me in your thoughts

Featured

Terra - LLM Webchat

Chat with Terra, my local LLM running on an NVIDIA DGX Spark. A gated webchat interface that connects to my 'terrarium-agent' server, which serves LLM responses from a lone to multiple different harnesses I've built.

Tech Stack: Express + WebSocket relay, React frontend, Python worker connecting to GLM-4.5-Air-AWQ-4bit via vLLM

Try Terra →

The opposite of MCP

I built a little architecture that's the converse of MCP. Rather than serving standard tools to LLMs, I serve a single LLM to a bunch of different tools.

I explain (read: try to justify) my architecture for serving multiple AI harnesses (IRC bot, webchat, annotation tools) with just one lonely local LLM.

Read full post →

Contact & Links

Twitter

GitHub

Resume