You have a course handbook on your phone and need one answer, not another long scroll. OGAM (Off Grid AI Mobile) lets you add the PDF to a project and ask questions about its text. With a downloaded local chat model, the answer can be generated on your iPhone without uploading the PDF to a cloud AI service.

Get OGAM for iPhone

Current mobile release

---

What would you like to do with Off Grid AI?

Have a feature or use case you would like us to support? Tell us what you want to do and which device you use.

Write to support@offgridmobileai.co, join our Slack community, or talk to us on Reddit.

Start with one PDF and a question whose answer you can check. Once that works, you can reuse the project for follow-up questions instead of attaching the same file each time.

What do you need for offline PDF questions?

Use a supported iPhone 12 or newer with iOS 17 or later, the current app, a downloaded on-device text model, and a PDF with readable text. Project document search is a free feature. A separate vision model is not needed to extract an ordinary PDF text layer.

In the iPhone Files picker, choose a PDF that is already downloaded. A file shown in iCloud Drive can still need internet before its contents are available. Download it first or keep a local copy under On My iPhone.

Choose a text model that fits the phone’s available memory. The app minimum does not mean every model fits. Complete installation and model downloads before the offline check.

How do you add the PDF to a project?

Open Projects, select New, give the project a clear name, and select Save. Open the project and use Add in Knowledge Base to choose your PDF. Wait until indexing finishes and the file appears in the list.

1. Create a project such as “Manual questions” or “Course handbook.”
2. In that project, select Knowledge Base > Add.
3. Choose the local PDF in the file picker.
4. Wait for Indexing to finish. Check that the file’s Use switch is on.
5. Under the project’s Chats section, select New.

OGAM reads the PDF text and prepares it for local search. The app then finds relevant passages when you ask a question. The released PDF extraction code uses the phone’s native PDF text extractor.

What should you ask first?

Ask one question about a specific task or detail. Request the filename or supporting passage so you can compare the reply with the document.

For example:

What does the handbook say I need to submit before the first class? Name the source and say when the document does not answer.

The expected result is an answer based on relevant text from the project. Open the PDF and check the passage yourself. A generated citation is a useful lead, not proof that every sentence is correct.

If the model gives a general answer, narrow the request:

Use only this PDF. Quote the short sentence that supports your answer. If you cannot find it, say that.

The instruction states what you want; it cannot guarantee that the model follows it. Check important instructions, dates, and numbers in the source.

Does this work with scanned PDFs?

A scanned PDF can be a collection of pictures with no text layer. This project importer does not perform OCR on those pages. If it cannot extract text, it reports that limit instead of creating a useful searchable document.

Use a text version from the original source, or create and check a searchable PDF before importing. Merely selecting a vision model does not add OCR to this PDF import path. The released project importer makes that limit explicit.

What should you check if the answer misses something?

Problem	Check	Action
The file cannot open offline	File location	Download it to the phone before the test.
No text can be extracted	PDF text layer	Use a searchable PDF or checked text version.
The file is listed but ignored	Project and Use switch	Start the chat inside the correct project and enable the file.
The answer is broad or incomplete	Question scope	Ask about one section or fact, then inspect the cited material.
The reply needs a connection	Active chat model	Choose a downloaded on-device model instead of a remote server.

Each answer uses a limited selection of passages. It does not review every page for every question. Very large documents can also exceed the import text limit. Break large collections into useful sections if you need systematic coverage.

How do you check privacy and offline use?

After the file imports and the local model is ready, enable airplane mode and switch Wi-Fi off. Ask a different question about the same PDF. You should still get a local answer from the prepared project.

The local model and document search run on the phone. Remote model choices and configured device sync are separate. Paired-device sync shares project data. If the PDF must remain only on this phone, keep it outside a paired shared workspace; there is no per-project privacy switch in this sync flow.

Download OGAM for iPhone, add one readable PDF, and ask one question you can verify in the original. Then repeat with internet access off.