STARBEST — telegrapher.work, offline copy
=========================================
© 2026 Xtra Ad Service, Inc. · x@telegrapher.work

Local use only. You may run this on your own machines. You may not
republish, resell, or host it publicly.


WHAT OPENS WITH NO SETUP
------------------------
Double-click any of these:

  index.html          the front page
  graph.html          the Telegrapher studio — the main event
  manual.html         the studio manual
  unsonic/index.html  the recordings, then on to the findings
  unsonic/whirr.html  the Unsonic Project research page

FIRST, WHILE YOU STILL HAVE INTERNET
-----------------------------------
graph.html loads Three.js and a few other libraries from a CDN. With no
network, the studio will not start. Run this once, from this folder:

    python make-offline.py

It downloads those libraries into ./vendor and rewrites graph.html to
use the local copies. After that the studio runs with the network off.

Once that is done: text, emoji, equations, SVG-to-3D, kaleidoscope,
voxel paint, and STL/GLB export all run in the browser, no server.

Two studio features need a real https:// address and stay quiet on a
local file: screen-share relief capture, and camera entropy. Browsers
gate those on a secure context, not on this software.


THE ACADEMY
-----------
  academy/index.html               the two tracks
  academy/wordsmith-academy-hosted.html    50 lessons
  academy/music-theory-academy.html        70 lessons

Opened straight from the folder, the lessons read and run, but nothing
is saved between sessions. Progress, submissions, ZIP export, and
grading all need the server.


RUNNING THE SERVER (this is what saves your progress)
-----------------------------------------------------
You need Python 3.9 or newer. In a terminal, from this folder:

    cd academy
    pip install -r academy-requirements.txt
    export ACADEMY_SECRET="pick-any-long-random-string"
    python academy-server.py

Then open http://127.0.0.1:5000 in your browser. Register a name and a
four-digit PIN and your work is archived on your own machine, in
./academy-data. Nothing leaves your computer.

On Windows, use `set ACADEMY_SECRET=...` instead of `export`.

Grading: without a model configured, submissions get an effort-and-
receipt check, which is honest about what it did not verify. To turn on
real grading, either run LM Studio and set:

    export ACADEMY_LLM=local
    export LMSTUDIO_URL=http://127.0.0.1:1234/v1

or set ANTHROPIC_API_KEY. Details in academy-README.md.


GETTING CAUSTIC (for the Musicology track)
------------------------------------------
Caustic 3 is not included here. It belongs to Single Cell Software and
is not mine to hand out. Get it from the developer:

    https://singlecellsoftware.com/caustic3.html

That page has a 64-bit Android build the developer recompiled himself.
It is not a new version -- it is what was last on Google Play, made to
run on modern devices. You will need to side-load the APK.

READ THIS BEFORE LESSON 1: the demo runs with no time limit, but SAVE
AND EXPORT ARE DISABLED. Every assignment in the Musicology track asks
you to submit a .caustic project file, and some ask for a MIDI or WAV
export. None of that is possible in demo mode. You need the separate
"Caustic Unlock Key" to save and to export WAV, OGG, and MIDI.

Two lessons (65 and 70) accept a written capture journal with no
project file. Everything else needs a saved project.


THE STUDIO SERVER (optional)
----------------------------
    python tvobj_server.py

Serves the studio at http://127.0.0.1:8080 and turns on short telegram
codes. Without it the studio still shares objects — the whole recipe
rides in the link instead of a five-and-five code.


FILES
-----
  index.html, graph.html, manual.html, gallery.html, archive.html
  academy/    the two tracks and academy-server.py
  unsonic/    the recordings, the Silence score, and the Whirr findings
  objects/    renders used by the front page
  tvobj_server.py
  make-offline.py   run once, online, to vendor the libraries
  LICENSE.txt
  START-HERE.txt

Built with the help of multiple AI models. Shared so you can use it.
