Jekyll AI Related Posts 🪄

Rubygems: jekyll_ai_related_posts

Jekyll ships with functionality that populates related_posts with the ten most recent posts. If you install classifier_reborn and use the --lsi option, Jekyll will populate related_posts using latent semantic indexing.

Using AI is a much better approach. Latent semantic indexing seems promising, but in practice requires libraries like Numo or GSL that are tricky to install, and still produces mediocre results. In contrast, OpenAI offers an embeddings API that allows us to easily get the embedding vector (in one of OpenAI's models) of some text. We can use these vectors to compute related posts with the accuracy of OpenAI's models (or any other LLM, for that matter).

• MikeKasberg.com

(Feel free to open a PR to add your website if you're using this gem in production!)

Jekyll AI Related Posts is a Jekyll plugin. It can be installed using any Jekyll plugin installation method. For example, in your _config.yml:

plugins:
  - jekyll_ai_related_posts

You should also ignore the cache files that this plugin generates. (This will help avoid a regeneration loop when using jekyll serve.)

exclude:
  - .ai_related_posts_cache.sqlite3
  - .ai_related_posts_cache.sqlite3-journal

All config for this plugin sits under a top-level ai_related_posts key in Jekyll's _config.yml.

The only required config is openai_api_key -- we need to authenticate to the API to fetch embedding vectors.

• openai_api_key Your OpenAI API key, used to fetch embeddings.
• fetch_enabled (optional, default true). If true, fetch embeddings. If false, don't fetch embeddings. If this is a string (like prod), fetch embeddings only when the JEKYLL_ENV environment variable is equal to the string. (This is useful if you want to reduce API costs by only fetching embeddings on production builds.)

ai_related_posts:
  openai_api_key: sk-proj-abc123
  fetch_enabled: prod

When the plugin is installed and configured, it will populate an ai_related_posts key in the post data for all posts. Here's an example of how to use it:

<h2>Related Posts</h2>
<ul>
  {% for post in page.ai_related_posts limit:3 %}
    <li><a href="{{ post.url }}">{{ post.title }}</a></li>
  {% endfor %}
</ul>

The first time the plugin runs, it will fetch embeddings for all your posts. Based on some light testing, this took me 0.5 sec per post, or about 50 sec for a blog with 100 posts. All subsequent runs will be faster since embeddings will be cached.

On an example blog with ~100 posts, this plugin produces more accurate results than classifier-reborn (LSI) in about the same amount of time. See this blog post for details.

The API costs to use this plugin with OpenAI's API are minimal. I ran this plugin for all 84 posts on mikekasberg.com for $0.00 in API fees (1,277 tokens on the text-embedding-3-small model). (Your results may vary, but should remain inexpensive.)

If you're already using Jekyll's built-in site.related_posts and you want to upgrade to AI related posts:

• Install the plugin.
• Replace site.related_posts with page.ai_related_posts in your templates.
• If you were using LSI, stop. It's no longer necessary. Don't pass the --lsi option to the jekyll command. You can remove the classifier-reborn gem and its dependencies (Numo).

This plugin will cache embeddings in .ai_related_posts_cache.sqlite3 in your Jekyll source root (typically the root of your project directory). The file itself is a SQLite database file. For most cases, I'd recommend adding this file to your .gitignore since it's a binary cache file. However, you may choose to check it in to git if, for example, you want to share cached embeddings across many machines (and are willing to check in a binary file on the order of 1-10Mb to do so). If the file is not present, it will be re-created and embeddings will be fetched from the API (which may result in higher API usage fees if done frequently).

Jekyll AI Related Posts is implemented as a Jekyll Generator plugin. During the build process, the plugin will call the OpenAI Embeddings API to fetch the vector embedding for a string containing the title, tags, and categories of your article. It's not necessary to use the full post text, in most cases the title and tags produce very accurate results because the LLM knows when topics are related even if they never use identical words. This is also why the LLM produces better results than LSI. These vector embeddings are cached in a SQLite database. To query for related posts, we query the cached vectors using the sqlite-vss plugin.

After checking out the repo, run bin/setup to install dependencies. Then, run rake spec to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and the created tag, and push the .gem file to rubygems.org.

Bug reports and pull requests are welcome on GitHub at https://github.com/mkasberg/jekyll_ai_related_posts.