Extracting geneology data from a rich and complicated source

I have a rather unusual use case. Tiiny allowed me to investigate something in a way I really did not have access to before I had my Tiiny. I should probably also point out that when I made my Tiiny pledge I had no real concept of how I would use Tiiny, I saw it more as a device I would learn things about, and maybe use from time to time, but in that process I would have gained some personal knowledge that I would never have otherwise learned. And all that was fine with me.

But as I realized in the last few days, I had something quite unique to investigate and Tiiny provided the real key to that. I'll tell you about that...even if the exact case probably won't be useful to anyone else. I'm only sharing stuff to share it, but it's a real case Tiiny allowed me complete. Note however, the Tiiny staff may also be interested in some of the lessons learned. I'm no great user of AI (outside of doing a lot of text-to-image renders), though I've used Claude a time or two to help with some small occasional matters.

The task. There is a 304 page physical genealogy book in the family that consists of sections of introductory history, and sections containing such things as point-by-point proofs of genetic descent. Without being absolutely precise, for privacy concerns, the book is essentially a family genealogical history that traces my mother's genetic line back to the first Governor of one the original colonies of the United States. If I mentioned the Colony, U.S. residents would certainly know of it, and it is also the name of one of the States.

The book is complicated. A couple of years back I discovered an exact, searchable text, 304 page PDF copy of the book online at one of the "heritage sites" to my absolute amazement, and I passed copies on to my brothers and sisters. I've glanced at the contents of the book, know my mother's eleven generation line (and naturally, my line) traces back to before the formation of the U.S., but as you might guess it's pretty dense stuff, and I haven't given any real time to actually reading the book.

Here's where my Tiiny and the Qwen3.8-27B model in "task" mode come into play. I decided to have the Tiiny and model read the book, give me a summary of each of the book's twelve sections, then after summarizing the sections, to summarize the entire book for me. I gave Qwen no more instructions than that--I didn't want specifics written that made use of any of my personal understanding. I just pointed out the location of the book in the Tiiny Vault and asked for the set of content summaries. Following that, I asked for the direct narrow lineage of the "First Governor" down to my mother.

I had first uploaded (via the Tiiny icon for the PDF upload) the 304 page PDF (in excess of 454,000 characters) to my Tiiny Vault so Qwen could process it.

After taking an initial look at the document, I could see scrolling by in the TiinyOS terminal that Qwen was unhappy with something about the nature of the Vault-formatted PDF (Tiiny folks, please note that...also, the PDF states when opened in Acrobat that it that it claims compliance with PDF/A standard and has been opened read-only to prevent modification) and Qwen decided to look through my laptop to see if there was a non-vault PDF that Qwen could make use of. It found two copies of the book on my laptop, decided both copies were identical to each other, and picked one copy as the PDF source to work with. That seemed to be much better with Qwen.

At this point there were two problems revealed, perhaps somewhat intermixed.

1. The PDF is quite old, could be a couple of decades I suspect, though I've had no problems reading it via Acrobat.

2. The "read" mechanism Qwen had access to only allowed the reading of 100,000 characters (another alert for the Tiiny folk), but Qwen knew the document length (through PDF header information, if that is part of file header format?) and knew it had to read the entirety of the book's 454K characters for valid summary purposes.

It's not clear to me how Qwen figured out how to read the full number of characters in the family history book, but it figured something out and was able to read everything! However, Qwen was also slightly unhappy with the PDF tools it had readily available (I had the Tiiny WiFi off) and thought maybe it could find something on my laptop to help. On the laptop I have (non-subscription) Adobe Photoshop CS6 and also an excellent Foxit PhantomPDF application, but Qwen found my installation of the Gimp image editor and decided it could make use of Gimp's PDF reading. Gimp PDF elements permitted Gwen to give me very clear and readable summaries of the PDF's twelve sections. The summaries were an absolute success.

Summaries in hand, I asked Qwen to give me the direct lineage from the Governor to my mother, and generate a PDF of the results so I could e-mail copies to my brothers and sisters. This is where Tiiny/Qwen put in some real work. I didn't observe wall-clock time, but call it one hour. The major hiccup was that now Qwen had to write PDF--and once again it turned to the Gimp on my laptop and cobbled something together to write the PDF file. The results were excellent, and probably more understandable than anything anyone had previously seen come out of the dense and lengthy book. The lineage including my mother was clear, complete, and easy to read.

Tiiny, Qwen, very, very good work indeed! I'm sure there will be more in the future.