Offline Document Request

guobiao_cn

How to use the documentation, references to download, I have searched for a long time, still can not find, whether there is a PDF document or API retrieval (all the examples);

At present let me have a dilemma is that some development environments do not allow access to the Internet, resulting in a very much time-consuming in datatables. I wish there was a way to have offline documentation and case code for developers to try.

I look forward to your reply, thanks!

allan

Hi,

I've wondered about this a few times as requests have come in for such docs. Its been a few years since the last one though I think, so it isn't something I've spent any time one.

There are a few options:

1. The docs are available in github and you could XSLT them into local files
2. An offline documentation tool such as Dash could be populated by those docs.
3. Download a copy of the reference docs from this site.

It is something I'd like to do, but unless I receive more feedback from people wanting this, it isn't something that I want to spend time on at the moment I'm afraid.

Allan

guobiao_cn

Not bad, solves part of the problem so far, but xml is obviously for advanced engineers, and also this catalog isn't as convenient as the html examples archived by datatables.net;
Tried reading datatables.net on a cell phone or tablet with ready access to the internet, and there was an incompatible display;

Maybe the document generation tool (Sphinx or MkDocs) has been very mature, such as product architecture encountered multiple language file formats, maybe Jekyll, Hugo or Hexo can be combined to do;

In fact, I would like to go further and throw offline documents to AI learning training, because from many AI models (limited knowledge base time) recognize datatables version is outdated, and the example is incomplete; I think that even rummage through the efficiency of datatables.net is far less than the AI search, generation, interpretation, correlation processing fast.

Speaking of which, I would like to make a suggestion, even if the vendors of artificial intelligence models want to learn datatables, if there is no good document format, it almost depends on the crawler to access to capture the information, which almost no one vendors original intention to do multiple times to capture the information of multiple pages (also afraid that it will be considered a malicious crawler).

guobiao_cn

I hope I can be understood if I have not expressed myself well enough, as this proposal will inevitably add to your working hours.

allan

Thanks for the feedback. The XML documentation of the API, events and options means that the data is structured and can readily be transformed into another format if someone wishes to do so. It should also help train an LLM if it understands the XML format.

It is something I'll look into when I have some time available, but time is often not my friend unfortunately. If someone else wants to look into this, I would certainly support that effort.

Allan

guobiao_cn

I'm currently working on AGI as well, and it does allow for progress, but to be open there is still the issue of deployment correlation of models;

The good news is that it is true that a lot of data can be supported, including XML as you say, and that this type of structured data can improve the qualitative reasoning, and this is true because I have tested structured data, and this is a very happy thing;

Then I run into two problems:
1, if the vector database is not combined with data cutting options, the quality of training is not good enough, especially over 10,000 files;
2, the full volume of training, that generated GUFF file must be dependent on the GPU machine, good news speed ideal, the understanding is more comprehensive; but it is very time-consuming;

Here please allow me to declare one point:
1, AGI is not a training class, just retrieve enhancement, which is only a temporary solution, a but the increase in the amount of data, it is not ideal;

2, full-volume training is to have almost its own model, they have to mention arch comprehensive data set to the choice of model training;
The good news is that the market is currently arching the model device operation, so as to save GPU machines, but the cost is still not ideal (for long-term operation, or more than 100 users of the GPU time-consuming high), for this problem, I am still considering smaller models to train, because from loading to computing, which utilizes the GPU with less time for the CPU, and perhaps this is the market GPU arithmetic determines the direction of this. Or maybe this is the ideal and cost-saving aspect at the moment;

To summarize: AI models added to the product is a good thing,
but the bad thing is that it will make the users lazier and lazier, because the FAQ fast belt to find the answer leads to less users to submit ISSUE or discussion, the users rely more on AI to communicate to deal with the problem, which in turn makes me pay more attention to the training data quality technology;

guobiao_cn

I'm glad to have a good experience to tell you that the choice of model is generally CODE type, and then the product's own architecture and components are to train the file data, such as datatables depend on the package pdfmake, native JS, associated with the VUE, pdfmake are to be increased, if the data quality is required to be too good, including the popular version, the special definition version, the development version Generally need to be updated once every few months; for the user to choose first, in the public interface in the role of the definition of the user to decide the version of this form of data and interaction is very good;

I share my experience above, there may be some words or side of the customary statements will make people misunderstand.
If my way is not correct, you can slowly discuss;

allan

I'm afraid I don't understand how offline documentation for DataTables could possibly help with AGI, which was the original question in this thread, but doesn't seem relevant to your latest post.

Allan

guobiao_cn

**Quote **“In fact, I would like to go further and throw offline documents to AI learning training, because from many AI models (limited knowledge base time) recognize datatables version is outdated, and the example is incomplete; I think that even rummage through the efficiency of datatables.net is far less than the AI search, generation, interpretation, correlation processing fast.”

**Maybe I'm just greedy, sorry. **

allan

I'd have thought the XML documents about be easier to train an LLM on than a rendered HTML or PDF document, which is what the "offline documentation" would be.

Allan