Ramiyar MD Engine

This folder is the engine: a single-file Python server that renders any .md in the browser with real math, diagrams, charts and hot reload — plus the notes it renders and the guides that explain it. What you are reading is the engine rendering its own home page.

Local time, from the built-in widget:

In one line. mddash <folder> serves any folder on the spot; right-click a .md in Finder → Open With → Ramiyar MD Engine does the same for one file. Nothing runs in the background on this Mac — the always-on copy is the Raspberry Pi.

Read this first

Getting started

The mental model (engine vs content folder), the one command, folder layout, the style/widget override rule, and what you can write in a note. Start here if you are new.

Projects Tracker

Everything currently open — five tracks, one page each, with live tick boxes. The map of what is being worked on.

Machine Learning Notes

The main body of writing this engine exists to render: six notes in order, from linear regression to decision trees, plus a four-part neural-network reference. Has its own hub.

Reading on the Pi

The always-on copy: how to push notes with mdpush, list them with mdlist, the Finder Quick Actions, and how pictures travel with a note.

What is in this folder

	What
The engine	`dashboard_server.py` (server + renderer + the whole HTML page), `style.css`, `neuralflow.js` / `neuralflow.css`, `fonts/`, `widgets/`, `node_modules/`, `.venv/`. These must stay flat here — the engine finds them as siblings of `dashboard_server.py`.
Launchers	`mddash` · `mdopen` · `mdpush` · `mdlist` (symlinked into `~/.local/bin/`), `Ramiyar MD Engine.command`, `Ramiyar MD Engine.app`, `com.mddash.server.plist`, `make-quick-actions.py`.
`Guides/`	Everything written about the engine — see below.
`0_My Projects Tracker/`	The map of open work — one page per track, live tick boxes, own hub. Its pages carry no copied numbers: widgets count them out of the content folders at render time.
`Machine Learning Notes/`	13 notes across four tracks, from linear regression to deep Q-learning. Own hub, own `CLAUDE.md`, own `status.json`.
`5G Systems Notes/`	5G NR, the RF front end, Open RAN and AI-RAN, written from the 3GPP and O-RAN specifications. Four layers, own hub and `CLAUDE.md`.
`Smart Devices/`	What actually runs on the Pi, one folder per device. The code is on the Pi and is the source of truth.
`Courses (RAW)/`	The course archive, 3 GB — lecture videos, Jupyter labs, slide decks, 3GPP specs. No `.md`, nothing served, never searched. Never `mdpush` or `mdexport` this folder's parent (the engine root) or you ship all of it.
`Raspberry Pi Access Guide/`	How to reach the Pi — SSH, Tailscale, the quick-access commands.
`FFT/`	A signal-processing note and its chart generator.
`templates/`	Copy-paste NeuralFlow starting points: topologies, fault states, decision branches.
`neuralflow-2/`	A standalone NeuralFlow sandbox with its own copy of everything. Not what the engine serves — keep it in sync by hand, or ignore it.
`archive/`	Frozen: the old locked guides, a scratch prompt, an old snapshot of the server. Nothing links to them.
`CLAUDE.md`	The engineering reference for the engine itself — routes, render pipeline, gotchas. Written for whoever edits `dashboard_server.py`.

The guides

getting-started.md

How to run it and what a note may contain.

AUTHORING-TEMPLATE.md

The house style, as a working page: front matter, auto-numbered headings, caption blocks for equations/figures/tables, a NeuralFlow pair, a KaTeX chart, callouts. Copy it to start a note.

neuralflow-guide.md

The diagram language in full — nodes, edges, groups, colours, sizing, layout, and flowchart mode for decision charts.

neuralflow-showcase.md

Live demos of every NeuralFlow feature, side by side.

charts.md

Chart.js recipes: line, scatter, bar, and the KaTeX-labelled variants.

showcase.md

The big one — every feature the engine has, on one very long page. Slow to load; useful when you want to see whether something is possible.

Running it

mddash                                  # serve the folder you are standing in
mddash ~/Documents/md-engine            # serve this folder (port 8031)
mddash -p 8090 "Machine Learning Notes" # any folder, any port
mdopen note.md                          # start/reuse a server for that file's folder
mdpush "Machine Learning Notes"         # copy a folder to the Pi
mdlist                                  # list what the Pi is serving

pkill -f dashboard_server.py            # stop every engine server
pgrep -fl dashboard_server.py           # see what is still running

After editing dashboard_server.py, restart the server. style.css and neuralflow.js are read from disk per request, but the page's inline JavaScript is baked into the Python process when it starts — so a server left running serves new CSS with old JS. CLAUDE.md gotcha 6 has the details.

Browse every file the server can see →