
Building a Knowledge Base in WorkBuddy: Docs Go In but Search Finds Nothing—Am I Doing Something Wrong?
It handles daily tasks pretty smoothly, but the moment I touch the knowledge base I'm lost.
I've been using WorkBuddy for a while — document categorization, spreadsheet merging, email organization — basically no issues. A couple days ago I came across a hands-on post about local deployment and knowledge base integration, saying you can feed your pile of materials in and then just ask questions in natural language. I immediately thought of my own pile — over a decade of client correspondence, product spec sheets, and a bunch of quote PDFs, all scattered across my hard drive, taking forever to find one document.
Yesterday afternoon I gave it a try. Created a folder, dumped about twenty files in — mainly product materials and correspondence with a few old clients. It ran for a while and the interface showed processing complete. I thought it worked, so I asked: "What's roughly the quote for model A to the European client?"
The result was wrong. It answered about a different model, and a price from last year that was way off from what I remembered. I tried rephrasing, and this time it just said it couldn't find relevant materials.
I suspect it's a file format issue. Some of those PDFs are scanned images — probably no text was extracted. But it could also be that I split things too finely — one spec sheet broken into a dozen segments, and retrieval can't match them up.
Another thing I'm not sure about: after processing, there was no notification at all — didn't tell me how many succeeded or failed. That makes it impossible to tell whether it didn't read the files or read them but couldn't find the right answer.
Has anyone stepped on this landmine? Do scanned files need to be converted to text first before putting them in, or is there something about file naming? I'll try again with only one category of materials at a time — will come back with updates if I get results.
Physix Frontier