How it works
From a folder of PDFs to an answer with a page number
There is no data project and no tagging exercise. You point Tinker at the documentation you already have, and technicians start asking questions.
Four steps, in the order you would actually do them
-
Bring in your documentation
Upload PDFs directly, or connect a Google Drive or OneDrive folder and let Tinker keep it in sync. Re-syncing updates the manuals you already have rather than importing them twice.
-
Tinker reads every page
Each manual is split page by page. The text is indexed for both meaning and exact terms — so “why is it locking out” and “code 33” both find the right page — and the figures on each page are extracted so a diagram can travel with the answer.
-
Set up your jobs
Create a job with your own job number, attach the site’s documents or a synced folder, and assign the technicians who are on it. They see that job; nobody else does.
-
Technicians ask, and get the page
A technician opens the job and asks in their own words. The answer names its sources, and each source opens the original PDF at the cited page for anyone who wants to check it.
Closing the loop
What your technicians learn stays with the company
The manual tells you what the manufacturer intended. Your technicians know what actually happens on a fifteen-year-old unit in a mechanical room with no clearance.
-
A technician writes up the fix
The problem, the root cause if they found one, what resolved it, the steps, and the parts. It takes a couple of minutes at the end of the call, while it is still fresh.
-
Management reviews it
Approve it, or reject it with a reason so the author can revise. Nothing reaches another technician without a decision — and a rejection without a reason is refused outright.
-
It answers questions from then on
An approved fix is searched alongside the manuals, cited the same way, and shown with its steps. A fix written on a job is ranked first for that job.
Common questions
How long does it take to get our manuals in?
Uploading is immediate. Processing a manual — extracting its pages, its figures, and building the search index — takes roughly a minute or two for a typical service manual, and runs in the background. You can watch each document’s status on the Documents screen.
What happens if a manual fails to process?
It is marked Failed with the reason: a scanned file with no extractable text, a password-protected PDF, or a file that is not really a PDF. A failed document is never used to answer a question, and it can be reprocessed once the underlying problem is fixed.
Can Tinker answer from more than one manual at once?
Yes, and it is designed to. Retrieval deliberately leaves room for a second source rather than filling the answer with ten passages from the same manual, which is what makes a disagreement between two documents visible instead of hidden.
What does a technician actually see?
A direct answer, ordered steps, the tools and parts named in the source, any safety warnings the document attaches to that procedure, and the sources themselves — each opening the original PDF at the cited page.
See it on your own manuals
We will set up a pilot with your documentation, your technicians, and your equipment — so you are judging Tinker on your own work, not on a demo dataset.