Technology of the Year: AI Integrations for Drupal

2023 was that year that Artificial Intelligence emerged from the futuristic shadows and into the spotlight, sparking transformative new levels of innovation, efficiency, and productivity. 

Here at Promet Source, we’ve leaned into the tipping point for AI, and the reality that in 2024, AI is serving as the driver of transformative leaps forward. 

That’s what’s inspired this riff on Time magazine’s annual Person of the Year. AI as an influencer of modern life is truly in a class of technology unto itself. 

AI has huge possibilities within Drupal, and the Drupal artificial intelligence (AI) community initiative is pooling powerful innovative energies into multiple projects. At this point, there are essentially three systems being built. 

1. Open AI / ChatGPT Integration,
2. Augmentor AI, and 
3. AI Interpolator.

Here’s an overview of each of the three projects. 

1. Open AI / ChatGPT Integrations

The OpenAI module is amazingly powerful and is designed enable a suite of modules and an API foundation for OpenAI integration in Drupal for generating text content, images, and content analysis.

OpenAI is the company behind artificially generated intelligence products that powers applications such as:

• ChatGPT
• GPT-4
• GPT-3, DALL-E
• GitHub CoPilot

The goal of this initiative is to find ways of augmenting and adding assistive AI tech leveraging OpenAI API services in Drupal.

While the newest GPT-4 models support up to 128,000 token context window, it’s important to note that the number of tokens that can be sent and received is dependent on the limits of an OpenAI account. This module cannot override tier limitations with OpenAI.

There are a vast range of submodules included in the Open AI / ChatGPT Integration. Among them:

• openai_audio: Adds capability to interact with the OpenAI audio (speech to text) endpoints.
• openai_chatgpt: Enables interaction with the Chat endpoint via ChatGPT API
• openai_ckeditor: Provides a button for CKEditor 5 to send a prompt to OpenAI and get generated text back.
• openai_content: Adds assistive tools for different areas of the content editing process. It adds functionality to adjust the tone of the content, summarize body text, suggest taxonomy terms for nodes, and checks content for Moderation violations.
• openai_dalle: Adds capability to interact with the OpenAI DALL·E (image generation) endpoint, using either the new DALL·E 3 model or DALL·E 2 model.
• openai_eca: Adds capability to build your own custom workflows with the ECA module. With this, you can create and combine your own custom functionality with ChatGPT.
• openai_devel: Adds GPT content generate capability to Devel Generate. This provides Devel a way of generating realistic content (not lorem ipsum) using GPT and ChatGPT models. Users can generate sample content from the Drupal UI or via Drush. This is useful if you want to fill out your site with realistic sounding content for client demonstration, layout, theming or QA.
• openai_dblog: This module demonstrating log analysis using OpenAI to find potential solutions/explanations for error logs. Responses from OpenAI are saved and will persist for common error messages so you can review them.
• openai_prompt: Adds an area in the admin where you can explore OpenAI text generation capability and ask it (prompt) whatever you'd like.
• openai_embeddings: This module that analyzes nodes and generates vectors and text embeddings of your nodes, taxonomy, media, and paragraph entities from OpenAI. Responses from OpenAI are saved and could augment search, ranking, automatically suggest taxonomy terms for content, and improve search relevancy without expensive search backends. Content personalization and recommendation may also be possible with this approach.
• openai_tts: Adds capability to interact with the OpenAI TTS (text to speech) endpoints.

2. Augmentor AI

Augmentor is an AI integration module which allows content to be augmented in Drupal via connections with external services. The module is designed as an AI framework which allows for the easy integration of disparate AI systems. It provides a plugable ecosystem for managing a variety of AI services such as GPT3, ChatGPT, NLP Cloud, and Google Cloud Vision.

Important features include CKEditor 4/5 integration and multiple field widgets such as: 

• Augmentor Default Widget: Designed for a simple augmentor field in which only a single value response needs to be managed, which amounts to the best choice according to the connection API from the augmentor execution. This widget can be compared to the Google "I'm feeling lucky" button.
• Augmentor Select Widget: Intended for the receipt of multiple result options in a select box, allowing for the choice of preferred options. 
• Augmentor Tags Widget: Helps when planning to target a reference field (tag style).

3. AI Interpolator

The idea of the AI Interpolator module is this: take one field and interpolate a result using AI or other modifiers into another type of field. One simple example is to summarize a long text or create an image from a text area with a description.

This modules exposes an API that allows other modules to do the following:

• Attach to a field config and allow any AI Interpolator field plugin to be choosable with custom configs.
• Add a processors plugin that allows other modules to decide how to process the fields. Everything is chainable, meaning that nearly complete entity generations can be completed using very little input.

A few examples of Interpolator AI possibilities:

• Generate podcasts from URLs,
• Research and summarize Google Search words, and
• Set taxonomies.

The website Workflows of AI offers more examples of possible workflows.

Promet's AI-Powered Initiatives

Here at Promet, we’re channeling resources and sharpening AI-related expertise and focus from multiple angles that include:

• The development of AI powered Drupal modules,
• AI as a web development tool, and 
• AI serving as an actual developer.

Metatag AI

Developed and maintained by Promet Source, Metatag AI is an an open-source Drupal module that leverages OpenAI to enhance SEO.

Automatically generating meta descriptions for blogs and other content based on headers and descriptions, Metatag AI provides content creators with new levels of efficiency and expertise by drawing upon SEO best practices and streamlining SEO-related tasks.  

Metatag AI is dependent on the Metatag module, and requires a paid Open AI account.  

AI Modules in Development

We’re currently working on transformative, AI-powered enhancements for search and chat functionality. 

To review: traditional search matches key words in a query and returns a long list of links on a search engine results page (SERP).

While answers to specific search queries are sometimes pulled from what appears to be the most relevant site, and highlighted on the SERP, AI-powered search takes it to the next level, leveraging large language models to construct specific answers to specific questions. 

Utilizing AWS technology we are working through questions pertaining to security with the expectation of releasing an AI-powered search capabilities in the relative near term.

Once intelligent search becomes available, traditional chat and search functionality will begin to appear as cumbersome and tedious. We’re excited about the profound advantages that AI-powered search will have in store for our clients and their users.

Interested in exploring AI-powered possibilities to take your website to the next level? It’s one of our favorite topics these days! Let’s talk.