In my previous article I wrote about how we implemented vectors-based semantic search in SQLite. Afterwards I asked myself what the best use case for this could be. My conclusion was to use the smallest database on the planet (SQLite) to automatically generate code for the largest database on the planet (CouchBase) 😁
Basically, we created a ChatGPT based AI chatbot that perfectly understands CouchBase, and all its associated code. Though perfectly might be an exaggeration, since we only scraped 3,000 pages from CouchBase documentation, and their documentation website has roughly 7,500 pages in total. However, it's good enough to illustrate the point, which is ...
We created an AI code chatbot using the smallest database on the planet that generates code automatically using ChatGPT for the largest database on the planet
Documenting software systems using ChatGPT
The point being that we can now basically take almost any software documentation website, regardless of how large it is, and generate a machine learning model, that automatically generates code for your software system, with nothing but natural language as its input. The consequences for small and medium sized software vendors I assume is obvious.
To bring the reasoning into perspective, realise that CouchBase is probably the largest database on the planet, powering stuff such as HBO and LinkedIn. Its capacity for storing data is "beyond" human understanding. It's the type of stuff you could build Facebook and Twitter with. And we created an AI chatbot based upon SQLite that understands everything about its code, and can automatically generate code for it, using nothing but natural language as its input. This implies anyone can use our tool chain to create their own personal "CoPilot", that's under the hoods using ChatGPT to generate code.
I think that's pretty cool 😁
Try out our AI code generator for CouchBase below.
- Our CouchBase 'GitHub CoPilot' - Psst, click the "Accept cookies" button, and then click the "Try our AI chatbot" in the bottom right corner of the page to try it out.