My idea is to create document embeddings from OpenAI’s API, then upload these embeddings to Pinecone. The purpose of this little app is to search for suttas. I’m not sure what the exactness of this search is, I invite people to test it. I created vectors for the suttas manually.
Yes, it’s indeed returning too many results. Actually, this is a list of all the sutras I’ve indexed. You should only care about the results with the highest score. In the future, I may limit the number of results. Anyway, this doesn’t affect the search.
Yes, as someone unfamiliar with the technologies you are using it’s not clear really what makes this something different than a regular search. And I really don’t understand the “create your own” bit.
I admire the complexity of your code. Admittedly, I understand very little apart from (if / else) basic rules of logic.
I know how difficult it can be. I essentially taught myself some basic C# to pursue a gaming project on Unity. After getting the physics of the player the way I wanted it. I struggled with the game logic.
Thankfully I met an extremely talented Unity game dev whom I’ve been collaborating with for the last six months.
I look at some of his work and it reminds me of yours - extremely complex and beautiful in that it asks the computer to perform unique tasks it likely has never encountered … and it’s all YOUR brain child!
Incidentally, I couldn’t access the search engine. Have you taken it down?