Breaking01 / AI
datasette-llm 0.1a9 now uses the llm library's default model
Calls with no model ID raise an error if the fallback model lacks a required API key or is excluded by configuration.

Version 0.1a9 of datasette-llm, released on 31 Oct 2026, applies the llm library's default model for users who have not set a default, according to its GitHub release notes. The notes mark the version as a pre-release.
The project's README describes datasette-llm as a tool that lets plugins for Datasette, a tool for exploring and publishing data, use LLM models through the llm library. The new behaviour applies when neither default_model nor a purpose-specific model setting is configured.
The README identifies the fallback as the model set with the llm models default command, or llm's built-in default if none has been set. The plugin resolves a model in a fixed order: the purpose-specific model, the global default_model, then the llm default.
Conditions of use. According to the README, the fallback model needs an API key where require_keys is on. Models excluded by the models or blocked_models settings, or by an llm_filter_models hook, a function through which other plugins can filter the model list, are not eligible. A call with no model ID raises a ModelNotFoundError when the fallback is unavailable.
Simon Willison described the change on a blog as a small usability improvement to the plugin that underlies Datasette Agent. Users previously named a model separately through the -s run option, for example gpt-6-astra.
Neither the release notes nor the README gives a date for a stable release.