• b5n 8 months ago

    `gptel` supports perplexity, among many others.

    https://github.com/karthink/gptel

    • michaelcampbell 8 months ago

      Recognized the domain; Lars' documentation of the Ding! NNTP reader in emacs from... 30(?) years ago was one of the few pieces of doc that had me audibly laughing as I read it.

      • Scarblac 8 months ago

        Is there really not some API out there that knows about all published books? Paid, if necessary?

        • wredcoll 8 months ago

          There is absolutely not such an api. For one thing, nobody actually knows because nothing stops random people from publishing a book at any time.

          For another, the major book publishers like amazon want to prevent anyone else from having their data.

          Beyond that, book metadata is probably the least standardized thing I have ever seen in my entire life.

          Titles, subtitles, editions, editors vs authors, series...

          • Scarblac 8 months ago

            I thought ISBN numbers might be somewhat centralized. But I don't know anything about the domain.

            • wredcoll 8 months ago

              The short answer is that isbns are sold more or less like ipv4 addrs. Companies buy a block and then some get assigned to books.

          • squeegee_scream 8 months ago

            I think there are, the author says so, but according to author they don’t make clear if a book is a new edition, translation, etc of a previously published book. So getting a list of published works wherein “A Book Title” is a single result and “A Book Title 2nd edition”, “A Book Title 3rd edition”, etc are not listed in addition to “A Book Title”, doesn’t exist. I would think it’s possible to write a layer of logic that takes a list of published books and removes extra editions, translations, etc to get what the author wants but perhaps the problem is more difficult than I realize

            • paulwarren 8 months ago

              we built this "work-level" catalog at Margins [1] -- 12mo+ of work -- book data is the messiest data I've worked with in my 10+ years of building things with data

              [1] iOS app to track and discover books: https://apps.apple.com/us/app/id6737528718

              • wredcoll 8 months ago

                I spent a couple of years working on this as a hobby and, yeah, book metadata is difficult to explain just how irregular it actually is. It might actually be worse than people names.

                That being said, the reason I was working on this was because I wanted a simple and effective way to get alerted to new books published by authors I want to track, can this work for that?

                • paulwarren 8 months ago

                  As a feature within our app, probably Q2/Q3 this year; we want to finish backfilling historical books before turning actual notifications on

                  As an API for you to query yourself for free, not for a while (but on the long-term roadmap!)

              • Y_Y 8 months ago

                When is a book the same as another?

              • precompute 8 months ago

                It'd probably be easier to scrape wikipedia or goodreads.

              • xhevahir 8 months ago

                The author believes that LLMs are toys and sets out to show that you can get some useless results if that's what you're looking for. Tiresome polemical exercise, but it does have the virtue of brevity.

                • dullcrisp 8 months ago

                  We may need to start putting trigger warnings on articles that mention LLM skepticism.

                  • signa11 8 months ago

                    they are not ?