A few weeks ago, I embarked on a journey to explore how OpenAI could change the way users interact with documentation. Imagine effortlessly accessing the information you need without reading through multiple pages. Instead, you just ask a question about configuring or using a tool, and voilร โ the answer appears before you.
You might be thinking, “That sounds a lot like ChatGPT.” You’re right! But what sets this concept apart is that it would be tailored to your specific dataset.
The Journey Begins: Exploring Large Language Models
To kick off my exploration, I started to read more about Large Language Models and stumbled upon an incredible tool called LangChain. This powerful tool allows you to create applications with LLMs.
My first test involved using two Markdown files and asking a question related to their content.
A few days later, I tuned into the Syntax Podcast, where Fred Schott discussed Astro 2.0 and the AI bot they built for it. My immediate thought was, “That’s precisely what I need!”
infoDive into Houston AI, Astro’s AI assistant.
Astro’s team also used LangChain to integrate AI into their comprehensive documentation.
Mendable: A Game-Changer in AI-Powered Documentation
As Astro’s team showed me I was on the right path, I wanted to learn more about LangChain. That’s when I discovered Mendable, a chat-based search service. Mendable integrates natural language search into documentation, and it’s even free for open-source projects!
The Mendable team was incredibly receptive and helped implement the service on the Front Matter CMS website.
Bring AI to Your Product
With access to the Mendable API, I set out to demonstrate its effectiveness by integrating it into Front Matter CMS. Users can now ask documentation-related questions directly within Visual Studio Code, eliminating context-switching and enhancing productivity.
Introducing GitHub Copilot for Docs
Mendable isn’t the only player in the AI-driven documentation space. On March 22, 2023, the GitHub team unveiled GitHub Copilot for Docs. I had the privilege of testing it before launch and providing my feedback.
infoIf you go to the GitHub Copilot for Docs website, you can add yourself to the waitlist.
Although it can’t be link to specific projects yet, I tried GitHub Copilot for Docs with React, Azure, and GitHub datasets.
The results were impressive, reaffirming my belief that AI is the future of documentation. AI is transforming countless processes, and documentation is no exception. While we’ll always need well-crafted documentation to train AI, the way we consume it is undoubtedly evolving.
The GitHub Copilot team even featured my quote on their website!
What are your thoughts on utilizing AI for documentation? Are you ready to embrace this game-changing shift?