dkatz23238

How does dehydrated compare to nginx-proxy with acme-companion from an operational perspective? I've been looking at other options and the best so far seems to be Caddy.

Parade DB are working on this problem: https://github.com/paradedb/paradedb/tree/dev/pg_analytics

I think something like this for Duckb would also be amazing!

I don't understand how a short tutorial on calling sklearn class methods consists of a "deep dive" into ML algorithms.

Seems to me like you have three options: keep on growing bootstrapped 100%, grow to 10k-20k mrr range and find investors to help you build a team to further grow and compete, third is to close the project and find better opportunities (a talented dev like you will not have a problem finding one). All of these options have expected rewards and risks, you must weigh out these alternatives compared to what you can currently sustain.

Option 2 will require you to focus on what are the 20% of tasks you can do that will have the highest short term impact to build traction.

Great choice. Setting up https from scratch using nginx is something you should definitely do ONCE in your life but then use automated tools such as nginx-proxy/acme-companion.

I am very happy with gcp. You can setup a cloud build trigger to build and push to gcr then deploy to gke or cloud run. A bit more involved that pythonanywhere but more flexible and arguably more powerful.

What statistics/metrics are used to evaluate RAG systems? Is there any paper that systematically compares different RAG methods (chunkings, models, ect)? I would assume that such metric would be similar to something used for evaluating summarization or question and answering but I am curious to know if there are specific methods/metrics used to evaluate RAG systems.

When I try to access the file it says "file has been deleted".

What would happen if each word in "tokenized" to an integer and then you generate tokens instead of characters to produce a string of coherent words instead of random strings? Maybe the answer is obvious but not to me without diving into it at a deeper level. Would be interested to hear anyones thoughts on this.

A perfect song to pair with this article is "Jagged Roots", by British electronic trio "Ivy Lab"[1]

1. https://www.youtube.com/watch?v=xJp4MB9PK8A

As a developer who has been building IDP solutions I can assert that although this model is a lot larger (more weights) than a Graph Neural Network on OCR tokens, industry standard before transformers, it outperforms given enough data. Depending on how heterogenous the data is usually 200 documents can reach human levels of accuracy on documents, scoring by levenshtein ratio.

Smaller graph models could get away with using less data. The problem that the "traditional" approach had is the the quality of the OCR was the bottleneck for overall model performance. It amazes me how this problem shifted from a node classification problem to a image to text problem.

Training on CPU was possible with GCN but not with Donut.

A small detail that you can improve UX would be to ensure that the button to execute a task is focused when you hit the tab key. In some of your tools I need to hit the tab key two or three times until I get to the execute/do button. Usually the delete and copy buttons are focused first. I would believe most people want to just enter their payload hit tab and enter and get their result... As mentioned this is a small detail, great work overall!

Get him another laptop, preferably Dell (just because I have great experience with them) so he can install Ubuntu on it and have a linux dev machine. This way he can develop software for linux servers (such as a web app or data science tool) without resorting to Linux subsystem for linux and all the related shenanigans.

Wow this is really cool and well designed. Sorry for this simple question but what did you use to get the nice sliding like animation? Is that intentional or just a result of updating the currently selected article?

Thanks for sharing and creating, I have not been able to find easy no-boilerplate 3d plotting in the browser for a while. grafar seems to fit this niche well.