Release: datasette-llm 0.1a11

Why it matters

The datasette-llm plugin for Datasette has added support for OpenAI-compatible endpoints, letting users point the tool at language models hosted anywhere that speak the OpenAI protocol. The plugin acts as a configuration layer, defining which LLM other Datasette plugins draw on. With this release, someone running local models through LM Studio, Ollama, or llama.cpp can launch Datasette via uvx, supply a small JSON configuration naming a model and a local API base, and get a Datasette Agent instance served locally, powered entirely by their own machine's model. The article walks through the exact commands involved, including downloading a sample database and starting the server with the plugin settings inline. The context a reader needs is that Datasette is a data publishing tool, and this change removes the dependency on a hosted provider: any OpenAI-protocol endpoint, including one on localhost, becomes a valid backend. The author also notes llama.cpp recently reached its first stable release, underlining how mature the local-model tooling has become.