Simon Willison reported on 14 November 2026 that datasette-llm version 0.1a11 adds support for OpenAI-compatible endpoints, allowing Datasette to connect to LLMs using that protocol.reported
What was released or claimed
The update enables configuration of LLMs for other Datasette plugins via any host speaking the OpenAI pseudo-standard web protocol, according to the published reportreported
Who it affects
Developers running local models with LM Studio, Ollama, or llama.cpp can now integrate these models directly into Datasette instances.
What is still unknown
The source material does not specify performance benchmarks, latency impacts, or compatibility issues with other non-OpenAI API standards beyond the pseudo-standard.
So what
Users can run a local Datasette Agent instance powered by their own local model, as described in the release notes.
What the paper shows
The release demonstrates that Datasette can interface with local inference engines through a standardized API layer. It shows that configuring a local server with a specific model, such as Qwen 3.8 27B, requires only passing JSON configuration to the `uvx` command.reported The limitation is that this functionality depends on the local model server adhering to the OpenAI protocol structure.
Why it matters
- The update allows Datasette to use LLMs hosted anywhere that talks the OpenAI pseudo-standard web protocol.
- It enables integration with local model servers like LM Studio, Ollama, and llama.cpp.
- Users can run a Datasette Agent instance powered by a local model using the `uvx` command.
- The release supports configuring LLMs for use with other Datasette plugins.
Topics: datasette, llm, local models, openai-compatible, python
Updated
Then / now / next
Datasette-LLM did not support OpenAI-compatible endpoints for local model integration.
Version 0.1a11 allows configuration of local LLMs via the OpenAI pseudo-standard protocol.
Watch for further updates to datasette-llm or Datasette-agent that may expand plugin compatibility or performance features.
Story timeline
- Then Datasette-LLM did not support OpenAI-compatible endpoints for local model integration.
- 14 Nov 2026 datasette-llm 0.1a11
- Now Version 0.1a11 allows configuration of local LLMs via the OpenAI pseudo-standard protocol.
- Next Watch for further updates to datasette-llm or Datasette-agent that may expand plugin compatibility or performance features.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
datasette-llm
An intermediate plugin for Datasette that supports configuring LLMs for use with other plugins.
uvx
A command-line tool used to run Python packages and scripts, here used to launch Datasette.
OpenAI pseudo-standard
A web protocol standard for LLM APIs that many providers and local tools now support.
Questions readers ask
Which local model servers are compatible with this release?
LM Studio, Ollama, and llama.cpp are explicitly mentioned as compatible options for hosting local models.
How do I run Datasette with a local model?
You can use the `uvx` command with the `datasette-agent` plugin and specify the model configuration in JSON format.
Does this require an internet connection?
The source does not state whether an internet connection is required, only that the models are local.
Reading receipt
- Version 0.1a11 of datasette-llm adds support for OpenAI-compatible endpoints.
- Users can configure Datasette to use local models hosted by LM Studio, Ollama, or llama.cpp.
- The release allows running a Datasette Agent instance locally using the `uvx` command.
Still open: How does the latency of local models impact the responsiveness of Datasette Agent interactions?
Still open: Are there security implications for running local LLMs with database access via this plugin?
