Talk about putting books in Scrapbox.

Shall we do it? Or not?

A feeling of (what?)

hmm

Oh, I happen to see a careless mistake that splits them into 190 and 832 pieces.

190 sides Ready

So let's pretend it's a revelation and do this.

Uploaded an estimated 40,000 images to Gyazo

You have fired too many requests.

/villagepump/nishio.icon

API has a usage limit of 12,500 times/day.

If the usage limit is exceeded, a 429 is returned in the Status Code of the Response.

Well, you can't have unlimited uploads and have them used as OCR, so there's a limit.

With 12500, well, it's not a problem for normal use cases.

Now that we know we can't "just upload 40,000 images for now," we need to think about Next Action.

Upload as well as Get were subject to the API limit.

In other words, the text has been used up in Upload, and Gyazo has OCR'd text, but it cannot be retrieved.

For a routine process, "25 additional 250-page books per day" is more than enough upper limit, but the question is what to do during the transitional period

Well, for now, we can't do the latter stage of the experiment until about 24 hours have passed, so we'll wait.

I wonder if there will be an increase in API quota if I pay extra for Gyazo.

If I signed up for about 10 Gyazo accounts, I could use 10x API.

You can't merge them into one account later and say, "Huh? No?" because you can't merge them into one account afterwards.

I'm going to have to migrate these 10, then I'm going to have to cancel nine.

Gyazo Team, the minimum monthly fee was 5,000 yen.

I thought it would be better to finally consolidate to Gyazo Pro so I would have less to think about, but I estimate it will take about 50 days due to the API quota during the transition period...

page.lines must be less than 10000 lines

(~_~)

Either they just can't import it or they can't get beyond it.

Or divide it into about 9,000 lines.

14274 items, 2.45 GB, 50 volumes

It took three days.

Changed implementation to split and import long pages.

It's done.

Talk about pinching books to Gyazo.

I finished fixing all the incomplete data due to Gyazo's lack of Quota.

It took 2.5 days.

Three groups of 20 books each were created for each new book to be processed.

If all goes as expected, we should be able to handle one group per day.

Well, there are 1,000 books, so even if we stay on schedule, that's a 50-day course...

It's a hassle to think "I have to repair the corrupted data..." or "I've already used up all my money today..." and so on.

I'm not sure, but there is a possibility that some of the 20 books are very long, and that they may exceed Quota after all...

I've implemented a recovery mode, so in that case, I'll recover the next day and then run a new one.

2023-09-19

I have a few images that I uploaded 9 hours ago that have not been OCR'd.

It sounds like a phenomenon where the OCR API call fails with a probability of failure and is not retried properly...

In the end, only four pages in one of the last books were symptomatic of that.

Filled with "OCR not available", what should I do?

---