datasette-enrichments-llm

Enrich data by prompting LLMs

This is an early alpha.

Installation

Install this plugin in the same environment as Datasette.

datasette install datasette-enrichments-llm

This plugin depends on datasette-llm for LLM model management, API key handling, and model provider integration. See the datasette-llm README for instructions on installing model providers and configuring API keys.

Configuration

datasette-enrichments-llm registers an enrichments purpose with datasette-llm. You can optionally configure which models are available and set a default model for enrichments using datasette-llm's purpose-specific configuration:

plugins:
  datasette-llm:
    purposes:
      enrichments:
        model: gpt-5.4-mini
        models:
        - gpt-5.4-nano
        - claude-opus-4.6

Usage

Enrichments can be run against any LLM model that has an LLM plugin providing asynchronous support for that model.

Multi-modal models are supported via the media_url parameter.

Development

To set up this plugin locally, first checkout the code. Run the tests like this:

cd datasette-enrichments-llm
uv run pytest

To try the plugin:

echo 'id,name\n1,Bob\n2,Kate' | uvx sqlite-utils insert data.db people - --csv

Then:

uv run datasette --root --secret 1 data.db