27 ms·
Ask HN: Do you have problem with keeping db docs up-to-date? Can AI solve it?
Hey, my name is Igor and I am trying to make a startup.
First of all I started doing something called [DataQuestion](https://dataquestion.io), which is a text-to-SQL tool. Found out no one needs it :)
Now I am switching to another problem. Many of the people I talked to say that their data storage is a mess. No one updates docs, it's hard to find relevevant dataset and onboarding of new analysts takes a ton of time.
Do you have some of this problems? How painful are they? Do you think it is possible to automate updating of db docs using AI? (so using schema and field values algorithm will compose a text description on every change).
Btw, here is a [link](https://datafresh.unicornplatform.page/) to a page that a bit illustrates the idea.
Thanks for your comments!
- gus_massa 5y agoClicky: Old project: https://datafresh.unicornplatform.page/ https://datafresh.unicornplatform.page/ New project: https://datafresh.unicornplatform.page/ https://datafresh.unicornplatform.page/ Is this the part written by AI? > Text field describing in which department employee works. Three unique values IT, Sales and Operations Can the user edit that text and use the AI version as a draft? What happens if the user edit's it and then a now department t is added?
- garrrikkotua 5y agoNo, this particular text on the page is not written by AI. It's just an illustration. Of course, AI version will be a draft. > What happens if the user edit's it and then a new department t is added? I think to separate the description in several parts. So general text description will be editable, but more objective things like unique values will be handled automatically.
- gus_massa 5y agoIt's interesting, but detecting which part to change looks very difficult. Or your plan is do something like wikipedia where you can edit a single section?
- garrrikkotua 5y agoAnd sorry about links, here they are again: Old stuff: https://dataquestion.io/ https://dataquestion.io/ New stuff: https://datafresh.unicornplatform.page/ https://datafresh.unicornplatform.page/
- gus_massa 5y agoHere links in the main text are not clicky, so sometimes I copy them for the lazy. (In this case, I copy the same link twice.)