datasette-enrichments-ollama

Datasette enrichment for analyzing row data using a locally hosted Ollama instance

I'm not a programmer so feel free to takeover this repo through a fork.

Requirement

you must have a running instance of ollama to use this plugin. At present this is only avialable on Linux or Macos. See more here: https://github.com/ollama/ollama

This is a straight fork from datasette-enrichment-gpt

Installation

Install this plugin in the same environment as Datasette.

datasette install datasette-enrichments-ollama

Configuration

This plugin needs an API url for the ollama instance. Configure that in metadata.yml like so

plugins:
  datasette-enrichments-ollama:
    api_url: "http://localhost:11434/api/generate"

Or to avoid that key being visible on /-/metadata set it as an environment variable and use this:

plugins:
  datasette-enrichments-ollama:
    api_url:
      $env: OLLAMA_API_URL

Usage

Once installed, this plugin will allow users to select rows to enrich and run them through prompts using models that are installed and registered in the ollama service, then saving the result of the prompt in the specified column.

Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment:

cd datasette-enrichments-ollama
python3 -m venv venv
source venv/bin/activate

Now install the dependencies and test dependencies:

pip install -e '.[test]'

To run the tests:

pytest