notes/vision

this project is intended to modernize the ritzmenardi.com website.
primarily by leaving it mostly alone, but adding a new section inside of
words.pdf called - words.pdf-sorted.

the goal of words.pdf-sorted is to enable users to better find the words that
best resonate with them. To that end, we are going to create a MASSIVE cache of
all of the poems that are in words.pdf. They will then be SORTED, and each one
will have a similarity score calculated between it and EACH OTHER POEM in words.

then, when a user clicks on a poem, it will show them ALL of the other poems
ranked in order of similarity. We will do this by using a local LLM via Ollama
and test the similarity using an embedding model. Then, we will pre-process the
words.pdf file (or output.txt) and extract the poems one by one, and then
generate (cached) an HTML document which has all the poems in the right order.

Once we've generated a sorted list for each poem, we can then add a link to each
sorted list (in an HTML document of course) inside of each poem, essentially
making it so that if the user clicks on a poem, it shows them all the other
related poems... and all the other non-related poems below that. With the
related ones prioritized near the top, of course.

you can find the format for the poems in words.pdf. It's generated in this
directory: /home/ritz/programming/ai-stuff/words-pdf

you can find best practices for front-end development in the docs directory:
/home/ritz/programming/ai-stuff/neocities-modernization/docs

the website should be created using only HTML, no javascript or CSS allowed.

you can find backups of the neocities website at /home/ritz/neocities
please copy these files to the project directory if you want to use them.

for the cached file generation, Ollama should be used. You can find the Ollama
installation at /home/ritz/programs/ollama. Ollama should not run on localhost,
but instead on the local IP address 192.168.0.115 with the port number 10265 -
make sure the ollama serve command is actually listening on this port and IP.

we can use the embeddinggemma:latest model for generating similarity scores.

The goal of the project in general is to create a flat HTML page with an easy
URL (simply linking to the hierarchical file location of the particular poem)
which shows all the related poems without removing any poems from the page.

Thus, if a player finds a poem that really resonates with them, by clicking on
it then they will be enabled to see all the poems that are meant to be. for them
and for thee, a poem or a note to see. vay, what beauty! lao, mon duty! perfect
it for me, computee. my faith is in your design, my hope is in your confine.