Questions and answers
The useful questions.
Especially the awkward ones about “local,” limits, and what is not built yet.
Does Papmil run locally on my computer?
No. Papmil is hosted. Your PDF is uploaded to AWS S3 and processed by Papmil's worker. “Local AI” refers to Ollama on that worker, not your device.
Is my document sent to ChatGPT, Claude, or Gemini?
No. The current smart splitting and naming steps use Ollama on Papmil's processing worker. Document content is not sent to an external AI API.
How long are files kept?
Source, intermediate, and result files are automatically deleted 30 days after the workflow finishes. Non-file workflow history remains.
What is live today?
One Standard workflow with four operations: resize, OCR, smart split, and smart rename.
Can I build my own workflow?
Not yet. The engine can represent longer and different chains, but the public self-service builder is aspirational. The Aspirational Roadmap is explicit about that.
How many steps can the engine contain?
The data model and orchestrator do not impose a four-step product concept. In practice, the only public workflow today contains four operations.
Which files and languages are supported?
The free proof of concept currently accepts one PDF up to 15 MB per workflow. OCR is live, but explicit language selection and detection are not wired; the worker relies on Tesseract's configured default.
How many free workflows do I get?
New free accounts currently receive two workflow credits. Those initial credits expire after two weeks.
Is Papmil a document archive?
No. Papmil prepares documents for filing elsewhere and deletes its file copies after 30 days. For a real archive, Paperless-ngx is an excellent choice.
Why are you so positive about competitors?
Because they are excellent. Papmil is early and narrow. Trust is more useful than pretending one experimental workflow should replace mature PDF products.