xory

a ton of backend changes to make it play ball with cloudflare workers. switched to drizzle from prisma and debloated next.js a lot (had to explicitly tell it to remove some vercel open graph stuff). it’s pretty clear that i don’t have a “solid stack” yet, but the RR7+hono+workers combo i used for anum seems to have been my best by far. but hey, trying stuff and making conclusions is the scientific method, so here’s hoping i find a good webdev stack that isn’t on the edge of imploding.

all that aside, we’re live on xorydev.xyz (my main domain!). i’ll get back to lana soon enough and wire that up to use hackclub’s api rather than asking for google api keys.

polishing, mobile optimisation and all that fun stuff. tailwind breakpoints got in the way initially by acting all weird, but i eventually got it to work. i can’t really write much because i don’t remember what i did with more detail than the umbrella term that is “polishing stuff”

added a simple blog system based on marked (obviously sanitised) even though i’m the only one making posts. it’s based on marked so i don’t have to handwrite pages and uses the same ol’ cloudflare d1 i’m used to from anum.
also had to do some restructuring so that i could have a different layout from the rest of the website

polished the landing page and navbar a bit. added an “about me” page containing how i got into this whole thing and a basic projects base that i have not filled out yet.

alright, so, i’m taking a quick break from all the lana nonsense because the gemini api is kind of tiring to work with. rest assured, i will get back to that in time and adjust that project accordingly.
in the meantime, i’ve started work on a newer version of my personal website (current one is https://xorydev.xyz). right now i’ve just refined the landing page.
for this project i decided to go with next.js and cloudflare workers, since i’ve had good experiences with that from anum (past non-hackclub project), it’ll be fast and it won’t be dependent on my own hosting which is… volatile, to say the least.

idek why this is devlog-worthy but i made some very slight improvements to make installation & configuration easier to understand

welp, we’re officially on pypi!
mostly worked on making this shipworthy and i think i’ve reached that point.

been a while since last devlog, mostly because i was going back and forth with feature creep.
for a moment i wanted to rebuild openclaw before realising that would take months to get to a ship-able state, although that would be a fun long-term project for after this event ends.

anyway, we now have an actually good cli using the rich library! alongside a ton of other cool ui/ux related gizmos, including but not limited to:

• persistent chats! you can now load and save chats
• loading and saving configuration files
• editing configuration directly from the cli

i have lost track of time so i hope flavourtown keeps it properly…
anyway, after hours effort and dealing with google’s nonsensical documentation and SDK, multimodal tool calls are WORKING!!! LET’S GOOOOOOOOOOOOOO!!!! :D
i can not possibly describe how long this took, seriously. this was such a massive pain and i’m so happy to have it be done.

logging progress @ 2 hours (~3 hours total)
i got a LOT of stuff working:

• everything is async now
• has full access to selenium :D
• some find/replace stuff
• python code and shell command exec
  tbd:
• find some kind of way to run code within a container so we don’t re-enact the antigravity incident
• discord integration because why not
• uhh idk whatever else i can think of c:

to summarise: making good (and fast) progress :3
…also, how do i add a pfp on the flavourtown website ( ._.) is it bound to slack or something?

okay, so, logging @ 1 hour… and 12 minutes.
i have a basic client working, with some functioning tool calls to my searxng instance.
seems to be going fine.
continuous chat is still a wip.

started work, basic nix+python env. got an api key and we’re ready to get going :3