kritrim-Smriti 1.3.2

An offline-capable, multi-provider AI study assistant for PDF books with VARK learning modes

PDF Tutor

---

I read a lot of technical PDFs — textbooks, research papers, course materials. The problem is that reading them passively doesn't really work for me. I'd finish a chapter and retain almost nothing.

I tried highlighting, re-reading, watching videos on the same topic — still slow. What actually helped was having someone explain it to me in different ways. A diagram here, a quick quiz there, hearing it out loud.

So I built this. You drop a PDF in, pick a chapter, and it explains it to you — using whichever style actually works for your brain. Diagrams if you're visual, audio if you're auditory, structured notes, or hands-on commands if you learn by doing. It can also run completely offline if you don't want your study material leaving your machine.

---

Screenshots

Main window	Mind map in the diagram viewer (zoom · copy · save)

Any AI-generated diagram opens in a built-in viewer where you can zoom, fit, save as PNG, or copy to the clipboard — and mind maps render offline via a local graphviz fallback when there's no internet.

---

How it works

The app reads the table of contents from your PDF and lets you pick a chapter. It sends that chapter's text to whichever AI you've set up and asks it to explain things in the style you want.

The learning modes are based on VARK — Visual, Auditory, Read/Write, Kinesthetic. There's also a quiz that figures out which style fits you, and an Omni mode that does all four at once.

Visual — mind maps, flowcharts, comparison tables Auditory — conversational explanation + reads it out loud (TTS) Read/Write — structured notes, definitions, writing prompts Kinesthetic — terminal commands and code experiments you can actually run

You can export notes as Markdown or HTML, generate Anki flashcards, or save interactive mindmaps as HTML files.

---

AI providers

It supports four providers — use whichever you already have access to:

Provider	Cost	Where it runs
Ollama	Free	Your machine (fully offline)
Google Gemini	Free tier	Cloud
Groq	Free tier	Cloud
OpenRouter	Free tier	Cloud

I mostly use Gemini because its 1M token context window handles entire chapters without truncation. Ollama is great when I'm on a plane or don't want data leaving my machine.

---

Installation

git clone https://github.com/Ashut90/pdf-tutor.git
cd pdf-tutor
pip install -r requirements.txt
python -m pdf_tutor

That's enough to get started. Two optional system packages unlock extra features:

sudo apt install graphviz   # diagram fallback when offline
sudo apt install espeak-ng  # offline TTS (otherwise falls back to gTTS)

On macOS, replace apt with brew. On Windows, graphviz has an installer at graphviz.org/download and TTS uses SAPI5 built-in.

---

Setting up a provider

Ollama (offline):

# Install from https://ollama.com, then:
ollama pull qwen2.5-coder:7b
ollama serve

Select Ollama in the app — no key needed.

Gemini / Groq / OpenRouter: Get a free key from aistudio.google.com/apikey, console.groq.com, or openrouter.ai, paste it into the app, done.

---

Architecture

The UI loads a chapter via core/pdf.py, sends the text with a mode-specific prompt (learning/modes.py) to the selected provider (ai/client.py), and renders any diagrams or charts the AI produces (rendering/visuals.py).

---

Project structure

pdf-tutor/
├── pdf_tutor/
│   ├── __main__.py          # entry point
│   ├── config.py            # theme, fonts, providers
│   ├── core/pdf.py          # TOC extraction, text/page rendering
│   ├── ai/client.py         # unified client for all 4 providers
│   ├── rendering/visuals.py # diagram / chart rendering
│   ├── learning/modes.py    # VARK prompts and teaching modes
│   └── ui/app.py            # Tkinter GUI
├── tests/
├── requirements.txt
└── pyproject.toml

---

Running tests

pip install pytest
pytest -v

---

What's next

Things I want to add but haven't gotten to yet:

• Conversation history that persists across sessions
• Multi-PDF library with search
• EPUB and DjVu support
• Configurable prompt templates per subject

If any of these interest you, feel free to open a PR.

---

Contributing

Open an issue first if it's a bigger change — happy to discuss direction before you spend time on it. For small fixes, just open a PR directly.

---

License

MIT — see LICENSE.

---

Credits

• PyMuPDF for PDF parsing
• Ollama, Groq, Google AI Studio, OpenRouter for model access
• VARK model by Neil Fleming