llm-chat-completions-server 0.1a0 Turns Your Local LLM Model Collection Into a ChatGPT-Compatible Endpoint — Written Entirely by GPT-5.6 Sol

Simon Willison·medium signalXBlueskyLinkedInCopy link

Willison released this plugin on 2026-07-30 as the first consumer of LLM 0.32rc1's content-addressable logs: it exposes every model in your LLM installation behind an OpenAI Chat Completions endpoint, so `curl http://localhost:8002/v1/chat/completions` works against local, Anthropic, or any plugin-provided model. Setup is three lines — `uv tool install llm --pre`, `llm install llm-chat-completions-server`, `llm chat-completions-server -p 9001`. Willison notes 'GPT-5.6 Sol wrote the whole thing' and showed strong recall of OpenAI's own API specification, which is a quietly telling data point about where model-written glue code is now competent.