Fine-tuned 230M LLM runs as browser agent, no server needed

A developer fine-tuned LiquidAI's LFM2.5 models (230M and 350M parameters) into a front-end agent that runs entirely in the browser using WebGPU or CPU/WASM. The agent calls tools to browse catalogs, answer questions, and manage a shopping cart without any server or API key. It learns interaction patterns rather than domain facts, so the same weights work across different storefronts without retraining.