Inside LlamaBox 1.1.0

Ask your PDF offline after setup

Prepare a readable PDF, a suitable local chat model and the document search model. Then ask questions and inspect sources without internet.

Inside LlamaBox 1.1.0. Version 1.1.0 is available on Google Play. Install the update to use document chat, model benchmarks and the refreshed toolkit. Availability depends on your device and region.

First check the file

Try selecting text in your PDF. A scanned image-only PDF cannot be read by the document text extractor. Use a short readable document for your first attempt.

Prepare on Android

  1. Install LlamaBox and check that your build includes the document Library.
  2. Choose a compatible chat model that fits your phone. Document mode recommends 3B+; smaller models may be experimental or unsuitable.
  3. Download the ~334 MB search model when prompted.
  4. Import the PDF into Library, wait until it is ready and attach it to your chat.
  5. Ask a question about the file. Tap a source to inspect the original passage.
  6. Turn off Wi-Fi and mobile data, then ask another question to verify your own offline setup.
Workshop handbook · PDFIllustrated example

What equipment should I bring?

Bring a laptop, its charger and headphones.

Printed worksheets are supplied at the venue. [1]

Open source 1 · Page 2

Workshop handbook / Equipment

Participants should bring a laptop, its charger and headphones. Printed worksheets are supplied at the venue. The workshop room opens at 9:30 am. The first session begins at 10:00 am.

A fictional document illustrates the interaction. This browser example does not run AI. Actual answers depend on your local model; check the source.

Keep the question concrete

Try “What equipment should I bring?” or “Where does it explain the schedule?” Check important details against the passage. A model can still make mistakes or answer something the file does not establish.

If something is missing

Check model compatibility, the file’s text layer, available storage and indexing status. Photo input is a separate vision feature; it does not add OCR to scanned PDFs.

Get it on Google Play