ATRIUMsearch → argument graph
Article · 2026-07-30 · 4 moments

llm-chat-completions-server 0.1a0

Release: llm-chat-completions-server 0.1a0 A key goal of the new content-addressable logs in LLM 0.32rc1 was being able to support OpenAI Chat Completion style requests where each incoming message extends the previous conversation, like this: curl http://localhost:8002/v1/chat/completions \ -H 'Content-Type: application/json' \ -d '{ "model": "qwen3.5-4b", "messages": [ {"role": "user", "content": "Capital of France?"}, {"role": "assistant", "content": "Paris."} ✦ AI generated

Read original ↗
GUEST AuthorHOST Simon Willison
01
Context◆

Content-addressable logs in LLM 0.32rc1 are designed to support OpenAI Chat Completion style requests where each incoming message extends the previous conversation.

The key design goal of content-addressable logs in LLM 0.32rc1 was enabling OpenAI Chat Completion style requests with growing, client-tracked conversation state.

transcript

Author: A key goal of the new content-addressable logs in LLM 0.32rc1 was being able to support OpenAI Chat Completion style requests where each incoming message extends the previous conversation

provides context · 1

03
Data

The llm-chat-completions-server plugin starts a localhost server on port 9001 that exposes all installed LLM models through a ChatGPT Completions compatible endpoint.

Installing and running the plugin starts a localhost server on port 9001 that makes all LLM models available via an OpenAI Chat Completions compatible API.

transcript

Author: To test that out, I built this plugin: uv tool install llm --pre llm install llm-chat-completions-server llm chat-completions-server -p 9001 Running this starts a localhost server on port 9001 that exposes your full collection of LLM models (from any plugins you have installed) using a ChatGPT Completions compatible endpoint.

04
Mechanism

The new schema design in LLM de-duplicates conversation messages using hashes of individual message parts.

The content-addressable log scheme uses hashes of individual message parts to avoid storing duplicate conversation data as requests grow longer with client-tracked state.

transcript

Author: Here the conversation state is tracked by the client, so each of these requests gets longer and longer. The new schema design in LLM is designed to de-duplicate these using hashes of the individual message parts.

explains mechanism · 1

Highlight slides
Content-Addressable Logs: Design Goal✦ from: Content-addressable logs in LLM 0.32rc1 are designed to support OpenAI Chat Completion style requests where each incoming message extends the previous conversation.GPT-5.6 Sol Wrote the Entire Plugin✦ from: GPT-5.6 Sol wrote the entire llm-chat-completions-server plugin, demonstrating thorough knowledge of the OpenAI Chat Completions API specification.Why Content-Addressable Matters✦ from: Content-addressable logs in LLM 0.32rc1 are designed to support OpenAI Chat Completion style requests where each incoming message extends the previous conversation.Deep Knowledge of the OpenAI API Spec✦ from: GPT-5.6 Sol wrote the entire llm-chat-completions-server plugin, demonstrating thorough knowledge of the OpenAI Chat Completions API specification.
Related episodes
Video · a16z Podcast

Ben Horowitz on the Global Race for Tech, Power, and Influence

→
Video · Lenny's Podcast

The rise of taste, human authenticity and judgment in an AI world | Adam Mosseri (Head of IG)

→
Audio · All-In Podcast

OpenAI Misses Targets, Codex vs Claude, Elon vs Sam Trial, Big Hyperscaler Beats, Peptide Craze

→