Fetching model details sometimes returns 429 and it's not daily download limit

Hi,

So I’ve run quite a few times into an issue where model details fail to fetch, the code is 429 and we are not hitting our daily download limit.

It seems to be an intermittent error, and it’s not related to a specific account, we can change the account being used and still run into the same issue right away. (Unless you are using our IP rather than account).

This is what we do:

  • From a search, we get the model details, but the license is missing, so when we try to retrieve it via:
    /v3/models/ we get 429. So at this point we can no longer tell our user what the license of the model they are downloading is.

  • If they choose to download it, we can still do that, and we do. But we have to specifically ignore the 429 error and unfortunately we can no longer store the license in our data for future use/restrictions.

  • 429 when getting the download links does work properly when we hit our daily download limit, but like i said, they are two different things and unrelated.

Hope that helps.

A 429 will also occur if you hit any of our API’s from your IP too quickly. The actual rates depend on your plan:

I see,
But why is it a write request?, I’m just trying to obtain the license data that didn’t come in the search results.

It is a bit of a problem that a search can go through, but the request to get info about an item in that search can’t

I’m not sure I follow, what do you mean with ‘write request’? With rate limiting, it can happen of course that your first request passes, but your second doesn’t.

If you could DM me your IP address we can check our logs to see if we can make improvements here.

My “write request” is from the quote you posted: " Hello, Write requests are limited depending on your [account plan]…".

So here’s the thing:

A user makes a search and gets a bunch of results, they inspect them individually by clicking on the icons.

When they click on the icons, i display the info but also immediately ask for model details so i can obtain the license.

So i suppose that those 200 requests an hour will go by quickly when just doing that. and that it will be a common problem.

Ideally the license info would come back with the search result, or obtaining that info wouldn’t be throttled.

In any case, I just wanted to bring it up. Thanks!

Model read requests are also limited:

    'anonymous': '200/hour',
    'basic': '1000/hour',
    'plus': '1000/hour',
    'pro': '1500/hour',
    'prem': '2000/hour',
    'biz': '2500/hour',
    'ent': '3000/hour',

Rather than make an additional request to each model to get its license, you could just filter your initial search by license.