Modifying Nomic Visualizations

[star-nox]

No description provided.

[lintrule-review]

You need to setup a payment method to use Lintrule

You can fix that by putting in a card here.

[railway-app]

This PR is being deployed to Railway 🚅

flask: ◻️ REMOVED

[KastanDay]

Thank you this is wonderful.

Edit: omg I'm impressed, you literally followed every best practice for every line you committed in the PR. Thank you for such high quality Python in the /onResponseCompletion endpoint.

I'll make it my top priority tomorrow to call your new endpoint from the front end so you can test.

[star-nox]

Re-did maps to include model responses for all courses except 2, they had a lot of data and I was hitting rate limit errors on openAI. Will try those again later.

[star-nox]

Re-did maps to include model responses for all courses except 2, they had a lot of data and I was hitting rate limit errors on openAI. Will try those again later.

All the maps are updated to reflect both query and response.

[KastanDay]

Re-did maps to include model responses for all courses except 2, they had a lot of data and I was hitting rate limit errors on openAI. Will try those again later.

All the maps are updated to reflect both query and response.

Is this expected behavior?

[star-nox]

Re-did maps to include model responses for all courses except 2, they had a lot of data and I was hitting rate limit errors on openAI. Will try those again later.

All the maps are updated to reflect both query and response.

Is this expected behavior?

These are the old links. I haven't updated the new map links in getMaps() function because we are still logging only user queries on prod. I was hoping to test your frontend endpoint and push everything together.

[KastanDay]

Oh okay, thank you. I'll really try to finish those endpoints today. Sorry for the delay, I had a lot of meetings in the wake of our first public launch.

[KastanDay]

@star-nox I pushed that endpoint. You will get response on /onResponseCompletion every time the AI finishes sending a message to the user. It will be the same format as conversation logs on supabase.

Here's the code: CAII-NCSA/uiuc-chat-frontend@190c738

Update: I made it so that it'll push to the flask server for this Pull request. So you should see the responses on Railway, for this PR's deployment.

[star-nox]

@KastanDay I am trying to create maps for all the courses right now and I keep getting a read operation timed out error. I got this first when I queried all the data at once and also when I query data course-wise. Have the supabase settings been changed? I was not getting this error when I created the maps before.

[KastanDay]

did adding the timestamps enable filtering by time? thanks!!

[star-nox]

did adding the timestamps enable filtering by time? thanks!!

nope. the filter still shows up blank.

[star-nox]

did adding the timestamps enable filtering by time? thanks!!

nope. the filter still shows up blank.

Just found one of their maps which has a filter by year - https://atlas.nomic.ai/map/neurips. They have uploaded year in their metadata, same as us. I am browsing the docs again to check if there is any attribute i need to set.

[star-nox]

So close, but still facing error! I am uploading datetime objects instead of strings.
Found a release note which mentions datetime validation: https://docs.nomic.ai/release_notes.html#v110

Update: Found this on the Nomic discord. This is not working for us though.

[star-nox]

Another update: added a random number field to check filtering and that works.

[KastanDay]

@star-nox note in adb1b50 you have some setting that's adding a whitespace character ( ) to every empty line. Please figure out how to change your settings so that doesn't happen, because it's causing your commits to be super messy.

I resolved this using git checkout origin/main <filename> to revert individual files to their status on the main branch.

[KastanDay]

Or please just add yapf to your vscode config using the config settings we have in our .github/workflows folder.

[KastanDay]

I think everything is good to go!