Worklog for task "Add PDF readers"

To begin with, I added a parser via OpenRouter with preliminary pagination. The speed and cost were surprising. The analysis time for a single-page document is about 8 seconds.

And as for the cost – tenths of a cent.

A bulkier document – 3.2 megabytes, 32 pages.

It took significantly longer – almost 5 minutes.

But this is simply a flaw in the queuing – each page was parsed sequentially, whereas it should be done in batches. I just need to refine the logic so that one page doesn't block another, while not sending everything at once either – otherwise, you might hit rate limit errors. Although for now in the current project, I'll make it send all at once to test whether a rate limit error occurs or not. ChatGPT says that OpenRouter doesn't impose its own limits on paid models, but limits can come from the ultimate provider.