At build time, it embeds all your posts into vectors and puts them into a vector database. Then, for each paragraph in a post, it queries the vector database for the 20 most similar paragraphs from all your notes. It stores these in a precomputed index. It then statically generates a blog for you (like Gatsby or any other static site generator), and uses this index so users can see related ideas when they click on a paragraph.
There are pros and cons to this approach. The biggest pro is that it generates a static blog. There's no backend to maintain, and it's quite lightweight and loads fast, while also providing the benefit of semantic hyperlinks. The con is that since the index is precomputed, you can't search for arbitrary things like you can in nomad hypertext.
At build time, it embeds all your posts into vectors and puts them into a vector database. Then, for each paragraph in a post, it queries the vector database for the 20 most similar paragraphs from all your notes. It stores these in a precomputed index. It then statically generates a blog for you (like Gatsby or any other static site generator), and uses this index so users can see related ideas when they click on a paragraph.
There are pros and cons to this approach. The biggest pro is that it generates a static blog. There's no backend to maintain, and it's quite lightweight and loads fast, while also providing the benefit of semantic hyperlinks. The con is that since the index is precomputed, you can't search for arbitrary things like you can in nomad hypertext.