PyNoteFlow

GPU Detection & Smart Setup

Connect PyNoteFlow to a local Jupyter server on a GPU machine or to Google Colab and GPU detection activates automatically — device name and VRAM appear in the status bar instantly (e.g. ⚡ RTX 4090 · 24 GB). Works identically for local and cloud GPUs.

Click the badge to open the GPU panel: see routing status, re-apply CUDA routing, or download a ready-made Jupyter GPU Launcher (.bat) that sets all CUDA environment variables before starting Jupyter.

Smart GPU Setup in the AI panel reads your notebook, identifies your framework (PyTorch / TensorFlow / JAX), and inserts the exact setup code. For PyTorch it auto-patches every relevant cell — adding .to(device) on models, inputs, and labels so nothing silently runs on CPU.

Built-in GPU Usage Guide with framework-specific tabs.

AI Notebook Builder

Describe a machine learning task in plain English. PyNoteFlow sends it to the AI and generates a complete, ready-to-run notebook — problem analysis cell, code cells, and markdown explanations — structured in logical chunks from imports to evaluation.

Agentic execution loop: the AI automatically runs each generated cell in sequence. If a cell errors, the AI reads the traceback and fixes it (1 auto-retry per cell). You watch the live progress log as it runs.

GPU-aware from the start — the AI sees your detected GPU and writes optimised code. No boilerplate hunting. No copy-pasting setup snippets.

Works with Gemini, Groq, and OpenAI-compatible APIs.

Smart Viz Suggestions

After every cell execution, PyNoteFlow scans your outputs for tables, arrays, and numeric data — then suggests the right chart type automatically. Click any suggestion to insert a ready-to-run matplotlib snippet. No typing required.

One File for Everything

Notebook cells, pen annotation strokes, Writing Pad sketches & cards, and embedded PDF state are all bundled into a single .ipynf or .yaml file. Reopen it anywhere in PyNoteFlow and every stroke, card, and PDF page is exactly where you left it.

The .yaml format is plain readable text — ideal for editing notebooks with AI assistants like Copilot or Gemini. Because it's structured text rather than binary JSON, AI models consume far fewer tokens reading and modifying it, keeping you well inside your context window.

Standard .ipynb files (Jupyter) and plain .py scripts are also fully supported for both opening and saving — keeping your workflow compatible with JupyterLab, VS Code, and any Python editor.

Reveal Mode — Live Teaching

Enter presentation mode with one click — all code is hidden. Press Page Down to reveal one line at a time, cell by cell, with a natural typewriter animation. Students see code build up as you explain it. Projector Only mode opens a separate window you drag to the classroom screen, while your laptop shows the complete code.

Dark · Light · Tint — Three Modes per Theme

Every colour theme has three background modes: ☀️ Light for bright rooms, 🌓 Tint to wash the accent colour across cell backgrounds, and 🌙 Dark for low-light environments. Toggle with one click in the header — no reload, instant switch.

5 Export Formats — Accurate & Lossless

Every export is a faithful, optimally structured transformation — no copy-paste, no lost formatting, no manual cleanup. HTML — self-contained with a dark/light toggle, all outputs inline. PDF — print-ready, respects your current layout. LaTeX .tex — compilable with pdflatex; code blocks, math equations, and tables map faithfully. Word .docx — Consolas code blocks, Cambria Math equations, correct heading hierarchy. Image ZIP — every chart and figure output as a PNG, bundled in one archive.

Pinch-to-Zoom

Two-finger pinch scales the entire notebook 0.5× – 3×. Zoom persists across sessions in localStorage. The annotation canvas resizes correctly after every pinch so strokes stay anchored to their cells.

Embedded PDF Viewer

Attach PDF files directly inside your notebook and browse pages without switching tabs. Create quick-access links between any PDF page, a specific notebook cell, or a Writing Pad card — perfect for referencing lecture slides, research papers, or textbooks alongside live code. Free plan: up to 10 PDF pages.

50-cell notebook as .ipynb

{
  "nbformat": 4,
  "nbformat_minor": 5,
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    ...
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 1,
      "id": "a1b2c3d4",
      "metadata": {},
      "outputs": [ { "output_type": "stream",
        "name": "stdout", "text": ["Hello\n"] } ],
      "source": [ "print('Hello')" ]
    },
    ...
  ]
}