ChatGPT has dramatically reshaped our interactions with technology. What was initially perceived as a platform for generating amusing memes has evolved into a robust tool due to a series of incremental yet revolutionary advancements. The advent of versions like ChatGPT 3.5 and 4 have unveiled their potential to generate effective code, including YAML files, and serve as digital librarians managing our resource libraries with proficiency.
Given Drupal's inherent structure, we posit that we may be able to delegate a significant chunk of site builder functionality to ChatGPT, leveraging its natural language processing capabilities.
In this talk, we will demonstrate two modules we have developed. The first exploits ChatGPT to generate views entirely from scratch while giving ChatGPT the necessary field and entity structure to create more complex views. The second integrates with Search API, enabling queries within the resources hosted on Drupal, including PDFs, nodes, and more.
We will initiate the discussion with a high-level demonstration of the module functionality and then delve deeper into their intricacies. This will help our audience comprehend the potentially advanced applications of ChatGPT within the Drupal ecosystem. Lastly, we will explore the future trajectory of this technology and its potential impact.
The tools we will be exploring are as follows:
- ChatGPT and Large Language Models (LLMs)
- Vector search through Pinecone
- Embedding techniques
- The concept of Langchain
- Drupal's Search API AI project (https://www.drupal.org/project/search_api_ai)
- Freelygive's Views ChatGPT project on Gitlab (https://gitlab.com/freelygive/views_chatgpt)
- OpenAI's Drupal project (https://www.drupal.org/project/openai)
Note: Summary has been written with Chatgpt because I feel that is necessary!