---
type: "post"
id: "release-0-1a11-datasette-datasette-llm-4df40e"
slug: "release-0-1a11-datasette-datasette-llm-4df40e"
title: "Release 0.1a11: datasette/datasette-llm"
description: "The datasette-llm 0.1a11 prerelease lets Datasette projects configure OpenAI-compatible Chat Completions and Responses endpoints directly, without an extra LLM plugin."
retrieval_nugget: "The datasette-llm 0.1a11 prerelease lets Datasette projects configure OpenAI-compatible Chat Completions and Responses endpoints directly, without an extra LLM plugin. Its endpoint settings also clarify model naming, credentials, and access checks, giving builders a more explicit way to connect local or hosted model servers to Datasette workflows."
published_at: "2026-11-17"
updated_at: "2026-11-17"
record_date: "2026-11-17"
date_kind: "published_at"
topics: ["developer-tools","inference"]
entities: ["Datasette","datasette-llm","LLM","Simon Willison","LM Studio"]
related_patterns: []
novelty_label: "new_fact"
what_changed: null
prior_refs: []
source_url: "https://github.com/datasette/datasette-llm/releases/tag/0.1a11"
source_urls: ["https://github.com/datasette/datasette-llm/releases/tag/0.1a11"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:release-0-1a11-datasette-datasette-llm-4df40e"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/release-0-1a11-datasette-datasette-llm-4df40e/","markdown":"https://newruntime.com/posts/release-0-1a11-datasette-datasette-llm-4df40e.md","json":"https://newruntime.com/posts/release-0-1a11-datasette-datasette-llm-4df40e.json"}
markdown_url: "https://newruntime.com/posts/release-0-1a11-datasette-datasette-llm-4df40e.md"
---

# Release 0.1a11: datasette/datasette-llm

## Retrieval answer

The datasette-llm 0.1a11 prerelease lets Datasette projects configure OpenAI-compatible Chat Completions and Responses endpoints directly, without an extra LLM plugin. Its endpoint settings also clarify model naming, credentials, and access checks, giving builders a more explicit way to connect local or hosted model servers to Datasette workflows.

## Summary
The datasette-llm 0.1a11 prerelease lets Datasette projects configure OpenAI-compatible Chat Completions and Responses endpoints directly, without an extra LLM plugin. Its endpoint settings also clarify model naming, credentials, and access checks, giving builders a more explicit way to connect local or hosted model servers to Datasette workflows.

## What happened
Simon Willison released datasette-llm 0.1a11 as a prerelease on November 14. The release adds direct configuration for OpenAI-compatible endpoints. A model reference can now be a dictionary containing an `api_base`, which can point to any OpenAI-compatible Chat Completions or Responses endpoint. The release says endpoint models support tools and streaming by default.

A new `endpoints` setting registers endpoint models globally or for a particular purpose without turning on the models allowlist. Endpoint dictionaries can also appear in `default_model`, a purpose's `model`, and `models` lists. The release gives a YAML example with a local LM Studio server at `http://localhost:1234/v1`, a model named `qwen/qwen3.8-27b`, and `responses: true`. It also shows a hosted endpoint at `https://models.example.com/v1` with the remote model name `organization/assistant` and the credential name `openrouter`.

The configuration separates the endpoint's local model ID from the ID expected by the remote server. The endpoint's model is its local ID and must not match an installed model ID or alias; `model_name` supplies the ID the remote server expects. Credentials are specified with `key`, not extra headers. A named key is looked up first in Datasette secrets and then among LLM's stored keys. The example `key: openrouter` corresponds to a key saved with `llm keys set openrouter`. An endpoint without a key is treated as unauthenticated and does not receive a provider's default key.

The release also changes model lookup and configuration validation. Allowlists, blocklists, and `llm_filter_models` hooks now apply to `model()` and `group()`, including explicit model IDs and configured defaults; selecting an excluded model raises `ModelNotFoundError`. If `require_keys` is enabled, a missing key raises `llm.NeedsKeyException` when a model is selected rather than when its first prompt runs. An explicit model key no longer falls back to the provider's default credential. Model-reference dictionaries reject unknown fields, as well as `key: null` and `options: null`; optional fields should instead be omitted. The release requires LLM 0.37.1 or newer.

## What the archive adds
New Runtime's earlier coverage on [a datasette-llm plugin compatibility break](https://newruntime.com/posts/simon-willison-released-llm-tools-datasette-0-1-3c9d33/) described how changes at the host's dependency boundary could disrupt a plugin. That was a separate event in the same Datasette and LLM ecosystem; 0.1a11 is a new release and a distinct integration capability, not a restatement of the compatibility fix.

## Trend signal
This does not yet fit a pattern New Runtime tracks. Direct endpoint configuration may make model backends easier to swap, but this release alone does not show a broader interoperability layer or a durable protocol boundary across agent systems.

## Why it matters
For a team already using Datasette and LLM, the release reduces the amount of glue needed to try a compatible local server or hosted endpoint. The same plugin configuration can identify the endpoint, remote model name, and credential, while exposing tools and streaming by default. That can make experimentation easier without treating every backend as a separately installed LLM plugin.

The security-relevant details are equally practical. A named key is resolved through Datasette secrets and LLM's stored keys, while an endpoint with no key is unauthenticated and does not inherit a provider credential. Explicit keys no longer silently fall back to a provider default. Teams can therefore make credential intent visible in configuration and catch a missing required key at model selection time. The new allowlist and blocklist enforcement for `model()` and `group()` also closes a gap between listing models and actually selecting them: a model excluded by policy should not become usable merely because code names it directly.

These changes are useful for builders who need to keep model choice configurable while preserving access controls. They do not establish that any particular endpoint is reliable, secure, or compatible in every operational detail; teams still need to test the backend they intend to use.

## Limits and uncertainty
This is a prerelease, and the release notes do not report adoption, production reliability, latency, cost, or comparative performance for local and hosted endpoints. OpenAI compatibility is the stated interface target, but the announcement does not document a conformance test suite or guarantee that every server's implementation of Chat Completions, Responses, tools, and streaming behaves identically. Builders should validate the exact operations their workflows depend on.

The notes specify how credentials are resolved and caution against placing credentials in headers, which are visible in Datasette's `/-/config.json`. They do not provide a broader security review of endpoint servers or deployments. The exact upgrade impact for an existing project will depend on its configuration and installed LLM version; the release requires LLM 0.37.1 or newer.

## Reader takeaway
The useful change is direct, explicit endpoint configuration: set `api_base`, use `model_name` for the remote server's identifier, and put credentials in `key` rather than headers. Before adopting the prerelease, test the chosen endpoint's tools and streaming behavior, confirm the required LLM version, and verify that your allowlists and credentials behave as intended.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Developer Tools](https://newruntime.com/topics/developer-tools.md): Explore the developer-tools topic hub.
- topic hub: [Inference](https://newruntime.com/topics/inference.md): Explore the inference topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)
