Datasette-LLM 0.1a1: A Revolution in LLM Model Integration

The latest update of the datasette-llm plugin, version 0.1a1, marks a significant advancement in the use of large language models (LLM) within the Datasette ecosystem. This plugin now allows for a smoother integration of LLM models for other plugins, such as datasette-enrichments-llm.

Among the new features, the plugin hook register_llm_purposes() has been introduced. This hook, along with the get_purposes() function, provides the ability to retrieve registered purpose strings. This enables centralized management of the objectives for which different models are used.

A key feature of this update is the ability to configure which models are used for specific objectives. For example, it is possible to specify that data enrichment should use the GPT-5.4-nano model, while SQL query assistance is handled by Sonnet 4.6.

Plugins relying on this infrastructure can use the syntax model = await llm.model(purpose="enrichment") to specify the purpose of the prompts they wish to execute. Furthermore, thanks to the new hook register_llm_purposes(), these plugins can register purpose strings, thereby facilitating the creation of administrative interfaces to assign models to specific objectives.