Building an LLM-powered writing assistant in Roo Code. A meta exercise - this post is written by an LLM based on analysis of the author’s posts.

Renting a cloud GPU from RunPod, running a large language model via vLLM’s OpenAI compatible endpoint, and load testing it with K6.

Notes on converting a transformer model from PyTorch to Safetensors format and quantising to ExLlamaV2, using a code-based calibration dataset.

A short example of using LLMs to converse with a fictional character from a seminal science fiction universe.