Release Milestone And Plugin Role In The Datasette Ecosystem

Issue 84 Edition 2026-03-25 4 min read

General

Sources: 1 • Confidence: High • Updated: 2026-03-26 03:27

Key takeaways

datasette-llm version 0.1a1 has been released.
Dependent plugins can request a model by purpose (e.g., await llm.model(purpose="enrichment")) and can register their purpose strings via register_llm_purposes() for centralized discovery (e.g., for an admin UI).
datasette-llm supports centrally configuring which models are used for different purposes (e.g., data enrichment versus SQL query assistance).
datasette-llm 0.1a1 adds a register_llm_purposes() plugin hook and a get_purposes() function for retrieving registered purpose strings.
datasette-llm is a base plugin that exposes models from the LLM project for use by other Datasette plugins, including datasette-enrichments-llm.

datasette-llm version 0.1a1 has been released.
datasette-llm is a base plugin that exposes models from the LLM project for use by other Datasette plugins, including datasette-enrichments-llm.

Dependent plugins can request a model by purpose (e.g., await llm.model(purpose="enrichment")) and can register their purpose strings via register_llm_purposes() for centralized discovery (e.g., for an admin UI).
datasette-llm 0.1a1 adds a register_llm_purposes() plugin hook and a get_purposes() function for retrieving registered purpose strings.

Dependent plugins can request a model by purpose (e.g., await llm.model(purpose="enrichment")) and can register their purpose strings via register_llm_purposes() for centralized discovery (e.g., for an admin UI).
datasette-llm supports centrally configuring which models are used for different purposes (e.g., data enrichment versus SQL query assistance).

Which Datasette plugins (if any) have adopted register_llm_purposes() and purpose-based model selection since datasette-llm 0.1a1?
Is there an implemented admin UI (or equivalent management interface) that consumes get_purposes(), and what capabilities does it expose?
What is the configuration format and precedence model for mapping purposes to models (e.g., per-instance, per-database, per-plugin overrides), and how are conflicts handled?
What model backends are supported through the LLM project in this integration context, and are there constraints on authentication, key management, or sandboxing?
Are there any performance, latency, or resource-usage bottlenecks introduced or mitigated by using a shared base plugin for model access?

datasette-llm becoming shared infrastructure could lower integration work for other Datasette plugins, increasing ecosystem cohesion and encouraging more LLM enabled extensions.
Purpose based model selection could shift model choice and cost control to centralized configuration, making deployments easier to standardize across tasks like enrichment and query help.
Registering purposes and enumerating them could enable management tooling such as an admin UI that discovers available LLM tasks and configures model routing consistently.

Multiple third party Datasette plugins adopt register_llm_purposes and request models by purpose rather than hard coding specific models.
A released admin or management interface consumes get_purposes and allows configuring purpose to model mappings in one place.
Documentation clarifies configuration format and precedence for purpose to model routing and shows stable behavior across instances and plugins.

Purpose registration remains unused by other plugins, and most integrations continue selecting models directly without llm.model purpose routing.
Centralized routing proves hard to configure or conflict prone, leading plugin authors to bypass datasette-llm for model selection.
Operational drawbacks such as latency or resource contention emerge from shared model access, reducing willingness to standardize on the base plugin.