Release And Configuration Centralization

Issue 91 Edition 2026-04-01 3 min read

Not accepted General

Sources: 1 • Confidence: Medium • Updated: 2026-04-13 03:32

Key takeaways

datasette-extract version 0.3a0 has been released.
Available models for datasette-extract extraction tasks can be controlled using the "extract" purpose in LLM model configuration.
datasette-extract now uses datasette-llm to manage model configuration for extraction tasks.

datasette-extract version 0.3a0 has been released.
datasette-extract now uses datasette-llm to manage model configuration for extraction tasks.

Available models for datasette-extract extraction tasks can be controlled using the "extract" purpose in LLM model configuration.

What specific changes (features, breaking changes, migrations) are included in datasette-extract 0.3a0 compared to prior versions?
What was the prior mechanism for configuring models in datasette-extract, and is it deprecated or still supported in 0.3a0?
How exactly is the "extract" purpose expressed in configuration (schema, defaults, examples), and what are the default allowed models if it is not set?
Where and how is model gating enforced (UI, CLI, API), and what happens when a disallowed model is requested (error vs. fallback)?
Are there any new constraints introduced by the datasette-llm dependency (version pinning, plugin compatibility, configuration precedence)?

Centralizing extraction model configuration into datasette-llm may reduce fragmentation across the Datasette ecosystem, potentially improving maintainability and easing multi-plugin deployments, which could support broader adoption of LLM-enabled workflows in Datasette.
Purpose-based model access control using an extract label suggests growing emphasis on governance and risk control for LLM usage, which could increase enterprise suitability by enabling task-scoped allowlists and limiting model exposure.
A pre-release that changes configuration surfaces implies near-term churn for users and plugin authors, which could temporarily slow upgrades while the ecosystem aligns on the new configuration approach.

Release notes for datasette-extract 0.3a0 show clear migration guidance and minimal breaking changes while confirming datasette-llm is the single source of truth for extraction model configuration.
Documentation or examples demonstrate extract purpose configuration schema, defaults, and consistent enforcement behavior when a disallowed model is requested, indicating the control surface is practical and predictable.
Subsequent releases move from pre-release to stable with community uptake, fewer configuration-related issues, and other plugins adopting the same purpose-based gating pattern.

Backlash or rapid follow-on changes indicate the datasette-llm integration introduces incompatibilities, unclear precedence rules, or frequent breaking changes that increase operational burden.
Model gating proves inconsistent across UI, CLI, and API, or disallowed model requests lead to confusing failures or silent fallbacks, undermining trust in purpose-based controls.
The project reverts or de-emphasizes the extract purpose approach or restores multiple configuration mechanisms, signaling the centralization strategy did not stick.