datasette-llm 0.1a1

2026-03-25 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, quick

Summary

Datasette-llm 0.1a1, a new base plugin, integrates large language models (LLMs) from the LLM tool into other Datasette plugins, such as datasette-enrichments-llm. This release introduces a `register_llm_purposes()` plugin hook and a `get_purposes()` function, enabling the configuration of specific LLM models for distinct tasks. For instance, users can designate "GPT-5.4-nano" for data enrichment and "Sonnet 4.6" for SQL query assistance. Dependent plugins can now request models based on purpose using `llm.model(purpose="enrichment")`, and register their own purpose strings, facilitating the development of administrative UIs for model assignment.

Key takeaway

For AI Engineers developing Datasette plugins, this release simplifies LLM integration by centralizing model selection based on task purpose. You should update to datasette-llm 0.1a1 to leverage the new `register_llm_purposes()` hook, allowing your plugins to clearly define their LLM usage and enabling more flexible, configurable deployments for end-users. This reduces hardcoding and improves maintainability.

Key insights

Datasette-llm 0.1a1 enables purpose-driven LLM selection and configuration within Datasette plugins.

Principles

• Modular LLM integration
• Purpose-based model assignment

Method

Plugins register purposes via `register_llm_purposes()` and request models using `llm.model(purpose="...")`, allowing centralized model-to-purpose configuration.

In practice

• Assign GPT-5.4-nano for enrichment
• Use Sonnet 4.6 for SQL assistance
• Build admin UIs for model mapping

Topics

• Datasette Plugins
• LLM Integration
• Model Configuration
• Plugin Hooks
• Purpose-Driven AI

Code references

• datasette/datasette-llm
• datasette/datasette-enrichments-llm

Best for: AI Engineer, Software Engineer, Machine Learning Engineer, AI Architect

Related on AIssential

• datasette-llm 0.1a8 — Datasette-llm 0.1a8 fixes a bug in its `llm_prompt_context()` hook, improving LLM response chain collection.
• datasette-llm 0.1a6 — Datasette-llm 0.1a6 simplifies model configuration and improves API documentation for plugin developers.
• Concepts of LLM Serving — LLM serving requires managing significant compute and memory resources, differing from traditional ML deployment.
• The Model You Love Is Probably Just the One You Use — Developer preference for LLMs is often driven by familiarity and integration ease, not objective superiority.
• Prodigy v1.12: OpenAI integration, prompt engineering, task routers, deployment docs and more — Prodigy v1.12 integrates LLMs and advanced task routing to accelerate data annotation and prompt engineering workflows.
• Big lab leaks — Agent harnesses enable LLMs to interact with external systems and execute commands by providing tools and managing context.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.