Everything here runs on your own computer with Ollama: nothing is uploaded, nothing costs money per use. Each tool publishes how accurate it is on a test set you can run yourself. Our products, Dhoon and Factwiser, are built from the same parts.

The tools

Each one does one job a person usually does by hand. A local AI model does the reading or writing, and code checks the result.

Open sourceMIT licence

receipt-reader

Receipts and invoices to a spreadsheet, read on your own computer.

Photos and PDF invoices go in; a CSV with the date, shop, category, tax and total comes out. An AI model reads each receipt, then code checks that the numbers add up, so a misread total is flagged for you instead of saved silently.

What building it teaches Structured outputs, checking a model's answers with code, measuring accuracy

Install (needs Ollama and Python)

pip install git+https://github.com/knowreva/receipt-reader
Guides

Build them yourself

Every tool comes with a step-by-step guide built from its real code.

Early access

Have a process that eats your week?

We're opening a small number of early-access places for automations built around your own tools. Join the list, or tell us what you'd like to automate.