Local LLM setups with LM Studio

LM Studio is my default recommendation for anyone who needs local inference: it encapsulates the complexity of GPU drivers, offers an OpenAI-compatible API and runs on a laptop with 32 GB of RAM. I've set it up for 10+ people by now — the job is always closing the gap between "runs in the GUI" and "runs in daily use".

Typical setups include: choosing the right quantised model for the use case, standing up an agent harness with tools, building a lean RAG layer around local documents and wiring the LM Studio API into existing workflows. The result is always the same — capable AI without any cloud dependency.